Aditya Gulati (Mar 24 2018 at 20:25): Aditya Gulati (Mar 24 2018 at 20:25):Hey,
I wanted to ask if no one yet is assigned the task python-brlcad task and wanted to add a proposal for the same. I am a former Code in finalist and have some experience with working with brlcad and for other open source organizations as well. Hope I am not too late and the task hasn't been assigned yet.
Shubham Rathore (Apr 25 2018 at 10:57):Hello, @Jaipal Singh . congrats on being a part of the community ! For the start, can we have your database updated on http://brlcad.org/wiki/Google_Summer_of_Code/2018 . Plus it is a good practice to maintain regular dev logs, and we've been following this for long. I suggest doing this first and then coming up with a prototype for immediate next task you are planning to do.
Jaipal Singh (Apr 25 2018 at 11:01):Oops, posted in the wrong stream! :(
Hey @Shubham Rathore Thanks a ton!
Yea sure thing, it'll be up and updated soon.
Would you recommend hosting the devlogs on BRL-CAD's website or would you recommend using an external page (maybe a blog or a github.io page)
I'm on the verge of completion of my sessionals at university in a couple of days. I'll start work on formalizing the project, latest by the 1st of May. :)
Shubham Rathore (Apr 26 2018 at 04:43):all we need are dev logs ! Up to you :)
Jaipal Singh (May 07 2018 at 11:43):Sure thing! @Shubham Rathore
Jaipal Singh (May 07 2018 at 12:39):Hey @Shubham Rathore , I've pushed introduce a new primitive (ell) and also given an example for the same. https://github.com/Killthebug/python-brlcad-tcl/commit/a0973c5b42615e0ef53ba9c876ee14992e1df937
If this looks good, I can start working on integrating the other primitives as well :)
Sean (May 07 2018 at 17:35):Hey @Shubham Rathore , I've pushed introduce a new primitive (ell) and also given an example for the same. https://github.com/Killthebug/python-brlcad-tcl/commit/a0973c5b42615e0ef53ba9c876ee14992e1df937
If this looks good, I can start working on integrating the other primitives as well :)
@Jaipal Singh you mentioned it earlier but can you explain again the difference with the different approaches being taken? also, do you have a write-up of your project plan posted up somewhere?
Jaipal Singh (May 08 2018 at 08:59):Hey @Sean , you can find a write-up of the project plan at : https://brlcad.org/wiki/User:Schindler97/GSoC18/Project
Jaipal Singh (May 08 2018 at 09:02):So, the original version of python brlcad (as developed by Kanzure in 2014) involved using ctypesgen. What ctypesgen does is that it allows you to load a DLL, dynamic-link library on Windows, shared libraries .so on Linux (or a dylib on macOS) and call functions from these libraries, directly from Python.
Jaipal Singh (May 08 2018 at 09:08):Now there are a couple of issues with using ctypesgen :
1. It sometimes has platform specific code
2. These bindings are brlcad version dependent, we figured this out when we saw errors in working with the original bindings on the latest SVN builds.
Jaipal Singh (May 08 2018 at 09:11):The other approach as developed by nmz787 (the one I propose we use) is fairly straight forward. You can have individual files that creates a tcl file (of either a simple primitive or a complicated shape involving multiple primitives), then sends the tcl file to mged. This then creates and populates a .g file database, then the .g file is converted with g-stl to produce an STL file.
Jaipal Singh (May 08 2018 at 09:12):Introducing newer primitives and also creating newer shapes in this approach is less complicated than the original approach. One advantage of this approach is that it does not need to interact with the brlcad binary, hence also making it (in a way) version independent.
Jaipal Singh (May 08 2018 at 09:13):Also, we use TCL just to write intermediate conversions
Jaipal Singh (May 08 2018 at 09:25):@Shubham Rathore , procedural geometry objects sounds like a good idea. Could you please elaborate more on that? I have a feeling that the tcl based approach does solve the problem of creating database objects that are described at run-time.
Sean (May 09 2018 at 04:00):@Jaipal Singh thanks for the background, very helpful
Jaipal Singh (May 09 2018 at 04:01):Sure thing :)
Sean (May 09 2018 at 04:02):so it sounds like there are several considerations to make then, especially with procedural objects being a primary desirable feature
Jaipal Singh (May 09 2018 at 04:03):Yea, though I'm still a little in the dark about how procedural objects work.
Sean (May 09 2018 at 04:04):a couple observations first though -- you said the ctypesgen bindings are version dependent but that is also true of calling mged -- it's just as likely that a ged command syntax will change
Jaipal Singh (May 09 2018 at 04:04):Oh! I wasn't aware of that
Sean (May 09 2018 at 04:04):the main difference being that with one you'll get a compile-time error about the change and the other, you just get bad runtime behavior
Jaipal Singh (May 09 2018 at 04:05):How often would the ged command syntax change (is there a way to predict this?)?
Sean (May 09 2018 at 04:05):Yea, though I'm still a little in the dark about how procedural objects work.
so they currently don't -- this is a feature that doesn't yet exist but has been on our todo for years, incredibly powerful
Sean (May 09 2018 at 04:06):there's no way to predict ged command changes -- they change all the time just like API changes all the time
Jaipal Singh (May 09 2018 at 04:06):Uh okay!
Sean (May 09 2018 at 04:07):so for understanding procedural objects, have you gone through the tutorials?
Sean (May 09 2018 at 04:07):created some geometry
Jaipal Singh (May 09 2018 at 04:07):Yea, I have
Jaipal Singh (May 09 2018 at 04:08):Though not for "understanding procedural objects", I was just trying to get a feel of how mged works and what all it can do
Sean (May 09 2018 at 04:08):suggest taking a few hours to get through all of the Introduction to MGED tutorials -- they take less than a day but introduce you to basic terminology and concepts used throughout BRL-CAD
Sean (May 09 2018 at 04:08):it's at http://brlcad.org/wiki/Documentation
Jaipal Singh (May 09 2018 at 04:10):Sure thing!
I've gone through Chapters 1-3 at http://brlcad.org/VolumeII-Introduction_to_MGED.pdf
Sean (May 09 2018 at 04:11):excellent, if you can make it through all 16 that will help a fair bit
Sean (May 09 2018 at 04:11):they're not the most exciting, but part of learning the terminology is through use, exposure, and repetition
Jaipal Singh (May 09 2018 at 04:12):I could definitely do that. Probably make some fun stuff along the way as well :P
Sean (May 09 2018 at 04:13):modeling your own thing after going through the tutorials is something I've found very effective as a next step
Sean (May 09 2018 at 04:14):so back to procedural objects -- check out the section in the TODO file for starters
Sean (May 09 2018 at 04:15):it might not all make sense if you've only gotten through lessons 1-3, but it's a short section
Jaipal Singh (May 09 2018 at 04:16):I get the part until "similar to shader, material, and other object constructs", after that it's difficult to grasp
Jaipal Singh (May 09 2018 at 04:17):So, I imagine something like an interactive shape creator?
Sean (May 09 2018 at 04:17):nods
Sean (May 09 2018 at 04:17):sort of
the general idea is to have an object that is fully defined by a script
Jaipal Singh (May 09 2018 at 04:17):Where in the user could request a cube and then the system prompts the user asking him for the dimensions of the cube.
Similarly, he can then add more and more primitives to create a complicated shape.
Jaipal Singh (May 09 2018 at 04:18):So this script is also a template of sorts in which you can fill up different values at runtime right?
Jaipal Singh (May 09 2018 at 04:19):I mean, the script would generate placeholders and get the user to fill them in
Sean (May 09 2018 at 04:19):it could take inputs
Sean (May 09 2018 at 04:19):so yes, sort of
Sean (May 09 2018 at 04:20):it could also take no inputs and just generate something via a script
Sean (May 09 2018 at 04:20):do you know any shell scripting?
Jaipal Singh (May 09 2018 at 04:20):Oh okay
And the script could be a tcl script?
Jaipal Singh (May 09 2018 at 04:20):Yea, I can work my way around a linux system fairly decently
Sean (May 09 2018 at 04:20):have you seen this -- http://brlcad.org/wiki/SGI_Cube ?
Sean (May 09 2018 at 04:21):ignore everything after RENDER THE CUBE
Sean (May 09 2018 at 04:23):basically that script is just some simple shell scripting that invokes mged commands (not unlike what nmz was doing in python) to create geometry
Sean (May 09 2018 at 04:25):now you can hopefully easily imagine that script in python form (in fact it might be an interesting exercise to replicate the script in python)
Sean (May 09 2018 at 04:26):as all it is doing is creating a bunch of rcc's and sph's using the ged 'in' command, and then grouping it all together
Jaipal Singh (May 09 2018 at 04:26):Yea, the script makes sense!
Jaipal Singh (May 09 2018 at 04:27):what does the -c "classic console mode" do?
Sean (May 09 2018 at 04:27):that just tells mged to not even worry about starting up a GUI
Sean (May 09 2018 at 04:27):classic mode aka command mode
Jaipal Singh (May 09 2018 at 04:28):Ah! Alright
Sean (May 09 2018 at 04:28):now here's the trick
Sean (May 09 2018 at 04:29):INSTEAD of a file outside of mged run through an interpreter that invokes mged ...
Sean (May 09 2018 at 04:29):imagine the script inside a .g file as an object, and that is run by an interpreter mged (or librt) invokes
Jaipal Singh (May 09 2018 at 04:30):I'm trying hard to wrap my head around this
Sean (May 09 2018 at 04:30)::)
Sean (May 09 2018 at 04:31):maybe let's try a simple example .. I'm going to create a procedure that makes a unit sphere, that's all it does
Jaipal Singh (May 09 2018 at 04:31):Alright
This is a bash script, let's assume that for now?
Sean (May 09 2018 at 04:31):doesn't really matter -- it can be a python script
Jaipal Singh (May 09 2018 at 04:31):Alright
Sean (May 09 2018 at 04:33):if it's a #!/bin/bash script, it would be a one-liner that said something like mged -c temp.g in sph sph 0 0 0 0 0 1 1000
Sean (May 09 2018 at 04:33):not very clever, it just makes a sphere at the origin (0,0,0) of size 1000mm (1m)
Sean (May 09 2018 at 04:34):in python, what would that look like currently?
Jaipal Singh (May 09 2018 at 04:34):Uhm, we could access mged using subprocess
Sean (May 09 2018 at 04:35):that's under the hood -- what would the python look like?
Sean (May 09 2018 at 04:35):the script
Sean (May 09 2018 at 04:35):there's currently no wrapper or encapsulation for shell script, so we call mged directly
Jaipal Singh (May 09 2018 at 04:35):import subprocess
bashCommand = "mged -c temp.g in sph sph 0 0 0 0 0 1 1000"
process = subprocess.Popen(bashCommand.split(), stdout=subprocess.PIPE)
output, error = process.communicate()
Jaipal Singh (May 09 2018 at 04:35):Oh! Oh alright!
Sean (May 09 2018 at 04:37):we had to directly invoke mged for shell script -- but I could totally create a wrapper like nmz did for python, so it's something simpler, so my shell script might look like: sphere create 0 0 0 1000
Jaipal Singh (May 09 2018 at 04:37):Yea! That could be easily done
Jaipal Singh (May 09 2018 at 04:38):Coupled with some primitive specific checks so that we don't run into errors
Sean (May 09 2018 at 04:38):sure
Sean (May 09 2018 at 04:39):so now that is run on a unix command line e.g., ./make_my_sphere.sh
Jaipal Singh (May 09 2018 at 04:39):So I could do something like python3 script.py temp.g
create sphere 0 0 0 1 1 1 1000
create ellipsoid 0 0 0 10 0 0 0 10 0 0 0 10
exit
Sean (May 09 2018 at 04:40):perfect, so we literally could do the same thing in python or shell
Jaipal Singh (May 09 2018 at 04:40):Uh-huh! yup
Sean (May 09 2018 at 04:40):put it into a text file and the shell environment invokes the appropriate interpreter and the script is run
Sean (May 09 2018 at 04:41):now instead of putting it into a file, we want to store that procedure in a .g file
Jaipal Singh (May 09 2018 at 04:41):Ok. How do we do that?
Sean (May 09 2018 at 04:41):not only store it there, we want to be able to run it as needed too
Sean (May 09 2018 at 04:42):a .g file is simple key=value database where the key is an object name and the value is the object's data
Jaipal Singh (May 09 2018 at 04:42):But the script is working to save the text file into a .g file right
Jaipal Singh (May 09 2018 at 04:43):Not literally the text file as is, but what the textfile wants to create as a .g file
Jaipal Singh (May 09 2018 at 04:43):This .g can then be loaded whenever it needs to be used?
Sean (May 09 2018 at 04:43):consider when you make a sphere, it creates an object with that name and stores it's parameters into the .g file
Jaipal Singh (May 09 2018 at 04:44):Ah ok! But the .g file is not human readable
Sean (May 09 2018 at 04:44):you did that with the "in" command to "input" the object
Sean (May 09 2018 at 04:44):in my_sphere sph 0 0 0 0 0 1 1000
Sean (May 09 2018 at 04:44):something close to that
Sean (May 09 2018 at 04:44):sph is the object type
Sean (May 09 2018 at 04:45):everything that follows are parameters
Jaipal Singh (May 09 2018 at 04:45):alright
Sean (May 09 2018 at 04:45):imagine being able to do something that puts a script into the .g, like
Sean (May 09 2018 at 04:46):in my_script script my_file.py
Jaipal Singh (May 09 2018 at 04:47):Our python script could then create tcl file after taking in the user input. Send it to mged which creates a .g file
Sean (May 09 2018 at 04:47):and then being able to "draw my_script" to see its wireframe, or "comb something u my_script - foo"
Sean (May 09 2018 at 04:48):forget about Tcl, there is no Tcl ... in fact, there isn't even necessarily an mged
Jaipal Singh (May 09 2018 at 04:49):So "in" and "draw" are bash commands that we define? What does my_file.py do then?
Sean (May 09 2018 at 04:50):no, sorry -- in and draw are GED commands
that's where getting through the rest of the tutorials would be helpful ;)
Jaipal Singh (May 09 2018 at 04:50):Ah okay!
Sean (May 09 2018 at 04:50):my_file.py didn't do anything in that example, it just contained a script (e.g., our one-liner) that was read into the my_script object
Sean (May 09 2018 at 04:51):so anytime I do something with the my_script object, it runs the script and uses the geometry that results
Jaipal Singh (May 09 2018 at 04:51):and the "script" keyword is defined or it needs to be defined?
Sean (May 09 2018 at 04:51):would need to be defined
Sean (May 09 2018 at 04:51):that's the "procedural object" bits mentioned in the TODO
Sean (May 09 2018 at 04:52):that keyword to the "in" command is the object type
Jaipal Singh (May 09 2018 at 04:52):Ah! Alright! So the script command would tell the in command what's coming in
Jaipal Singh (May 09 2018 at 04:52):Cool cool
Jaipal Singh (May 09 2018 at 04:52):*script keyword
Sean (May 09 2018 at 04:53):*object type ;)
Jaipal Singh (May 09 2018 at 04:53):Ugh! My bad :P
Sean (May 09 2018 at 04:53):it is a keyword obviously, but it directly refers to a registered object type
Jaipal Singh (May 09 2018 at 04:54):Again, this would not be python right? I'd have to work in C to manipulate the ged commands?
Sean (May 09 2018 at 04:54):what is "this"?
Jaipal Singh (May 09 2018 at 04:55):Introducing "script"
Sean (May 09 2018 at 04:55):we're talking about a lot of things...
Sean (May 09 2018 at 04:56):ah, right so that bit would be core infrastructure code that someone else would probably write -- I could probably implement the basic necessary bits in a couple days
Sean (May 09 2018 at 04:56):if you want to see what it looks like, look in the src/librt/primitives directory in a brlcad source checkout
Jaipal Singh (May 09 2018 at 04:57):Sure thing, I'll have a look at that
Sean (May 09 2018 at 04:57):the table.c file is where all object types are registered and each implementation lives in the corresponding src/librt/primitives/OBJ subdir
Sean (May 09 2018 at 04:57):so the idea is that there would be a new "script" object type that would hold the actual script (instead of it being in a text file)
Sean (May 09 2018 at 04:58):and there'd be an embedded interpreter to run the script
Jaipal Singh (May 09 2018 at 04:59):So this script would parse a text object at run time and use the appropriate interpreter to create valid shapes?
Sean (May 09 2018 at 05:00):almost got it
Sean (May 09 2018 at 05:00):LIBRT would parse a script object at run time and use the appropriate interpreter to create geometry
Jaipal Singh (May 09 2018 at 05:01):But the script object can only exist if it is fed in valued from an external textfile?
Sean (May 09 2018 at 05:01):LIBRT is the main library in BRL-CAD that is used by essentially everything
Sean (May 09 2018 at 05:01):not necessarily
Sean (May 09 2018 at 05:02):imagine creating a script in mged or archer, and when you run "edit myscript", it simply pops open a text editor
Sean (May 09 2018 at 05:02):not reading from an external text file, but from the object
Jaipal Singh (May 09 2018 at 05:02):Ohhhhhhh!
Jaipal Singh (May 09 2018 at 05:03):It's almost like reverse engineering objects and converting them into plain-text descriptions
Jaipal Singh (May 09 2018 at 05:03):_or I got this completely wrong_
Jaipal Singh (May 09 2018 at 05:04):A script in mged would be?
Sean (May 09 2018 at 05:05):what do you mean?
Jaipal Singh (May 09 2018 at 05:06):the mged script could be either a bash one or something for the python interpreter
Jaipal Singh (May 09 2018 at 05:07):Also, is edit a ged command?
Sean (May 09 2018 at 05:07):yes, edit is a ged command
Sean (May 09 2018 at 05:07):I feel like we're still missing something... :)
Jaipal Singh (May 09 2018 at 05:07):I think going through the tutorial is going to fairly help
Sean (May 09 2018 at 05:07):yes, probably
Sean (May 09 2018 at 05:08):and then maybe going through an example without shell or python mixed in
Jaipal Singh (May 09 2018 at 05:08):How about this? I'll go through the tutorial and then go through our conversation again and maybe come up with a possible interaction-diagram of the different components and we can modify that to get a clearer picture?
Sean (May 09 2018 at 05:09):sure, sounds like a good start
Sean (May 09 2018 at 05:09):I don't intend you to get too caught up on the C/C++ side of things in LIBRT since that's outside of scope for your project
Jaipal Singh (May 09 2018 at 05:09):As far as I understand, there are multiple parts involved in being able to execute
in my_sphere sph 0 0 0 0 0 1 1000
Sean (May 09 2018 at 05:10):it just influences what will work best architecturally
Jaipal Singh (May 09 2018 at 05:10):Sounds fair, but I don't mind playing around there as well (though I might take a while to figure things out)
Jaipal Singh (May 09 2018 at 05:10):Uh-huh
Sean (May 09 2018 at 05:11):what it implies is that you have C/C++ code running an embedded python interpreter
Sean (May 09 2018 at 05:11):that logic is running from inside of "mged" and/or other command-line tools (e.g., "rt", "archer", "gqa", ...)
Sean (May 09 2018 at 05:12):what we're wanting to end up with with procedural objects is not just the ability to source scripts, but create complex geometry like the SGI cube shell script
Sean (May 09 2018 at 05:13):notice in that shell script how the cube dimensions are hard-coded
Jaipal Singh (May 09 2018 at 05:13):Yea
Sean (May 09 2018 at 05:13):radius=100 for example
Jaipal Singh (May 09 2018 at 05:14):We'd want that to be interactive?
Sean (May 09 2018 at 05:14):say I change the script, remove 100 and specify $1 instead (argv[1] or sys.argv[1] in python terms)
Jaipal Singh (May 09 2018 at 05:14):Yea
Sean (May 09 2018 at 05:15):so instead of ./sgi.sh ... I'd run ./sgi.sh 100 and get a 100 radius shape
Sean (May 09 2018 at 05:15):or ./sgi.sh 10000 and it's bigger
Jaipal Singh (May 09 2018 at 05:15):True true
Sean (May 09 2018 at 05:15):as a procedural object, we'd want similar flexibility
Sean (May 09 2018 at 05:16):maybe something like: in mysgi script sgi.py 1000
Sean (May 09 2018 at 05:16):or, in mysgi script sgi.py
Jaipal Singh (May 09 2018 at 05:17):In the second case sgi.py could have some hardcoded values inside of it (or maybe fall back to some pre-defined value)
Sean (May 09 2018 at 05:17):followed by some command that specifies parameters like "in mysql1000 exec mysgi 1000"
Sean (May 09 2018 at 05:18):that way, you could stash the script and evaluate different parameters without making copies of the script each time
Jaipal Singh (May 09 2018 at 05:19):So, if you don't specify the "1000" here, would that raise an error?
Sean (May 09 2018 at 05:20):would probably depend on the script
Sean (May 09 2018 at 05:20):maybe it has a default, maybe it requires 3 parameters
Jaipal Singh (May 09 2018 at 05:20):Sure
Jaipal Singh (May 09 2018 at 05:20):So, exec here, that would need to be defined too or is it already defined?
Jaipal Singh (May 09 2018 at 05:23):Also, my initial tasks would be wrapping the primitives and then getting the python interpreter to work?
Sean (May 09 2018 at 05:23):would need to be defined -- that's why this concept is in the TODO file
Sean (May 09 2018 at 05:23):it's a somewhat advanced CAD feature, very powerful
Sean (May 09 2018 at 05:23):but not terribly hard to implement
Sean (May 09 2018 at 05:24):have you ever played with openscad?
Jaipal Singh (May 09 2018 at 05:24):Nope, never have :/
Jaipal Singh (May 09 2018 at 05:25):I'll brb! _AFK_
Sean (May 09 2018 at 05:25):ah, well might still be helpful example -- openscad is a geometry system that essentially uses procedures everywhere
Sean (May 09 2018 at 05:25):if you check out https://en.wikibooks.org/wiki/OpenSCAD_User_Manual/User-Defined_Functions_and_Modules and jump down to "Example 2", you'll see something very similar to what we were talking about
Jaipal Singh (May 09 2018 at 05:34):Example two is just like python's range function
Jaipal Singh (May 09 2018 at 05:34):Wait, sorry example 2 in modules?
Sean (May 09 2018 at 05:36):yes, sorry -- example two under the Object modules section -- little houses
Sean (May 09 2018 at 05:36):module house
Sean (May 09 2018 at 05:38):another consideration to test is performance .. how long it takes with the two python methods to create XXXX objects
Jaipal Singh (May 09 2018 at 05:38):Ah, I get the picture
Sean (May 09 2018 at 05:39):alright, times up for me -- good talk
Jaipal Singh (May 09 2018 at 05:40):Thanks a ton for everything! I've definitely got a lot to think about and discover :)
I'll come back to you with a design for the procedural geometry in a couple of days?
Sean (May 09 2018 at 05:40):give the tutorials a go, revisit the discussion, check out TODO, talk some more with the other mentors too, test performance
Sean (May 09 2018 at 05:40):sounds good
Jaipal Singh (May 09 2018 at 05:40):Awesome! Thanks
Jaipal Singh (May 10 2018 at 17:47):Hey @Sean, @Shubham Rathore! I'm travelling home and will be AFK for the next two days (until 12/10/18), I'll resume work on the 13th and will make up for lost time then! :)
Shubham Rathore (May 13 2018 at 09:23):Hey @Jaipal Singh with the coding period about to start, I hope you are ready with a plan for the deliverables. I think Sean cleared most of your doubts, still if you are stuck with anything, feel free to ask. Wish you a happy coding period. Cheers !
Jaipal Singh (May 13 2018 at 09:58):Hey @Shubham Rathore! Yea the conversation did clear out a ton of things (also introduced a ton of doubts). I'm working on completing the mged tutorial series, should be done by tonight.
The first immediate goal is start wrapping up relevant primitives while designing the procedural geometry architecture parallely. I'll be furthering nmz's implementations and will keep you posted with my daily updates on the dev log. Cheers! :D
Jaipal Singh (May 14 2018 at 08:24):When working with arb4-arb8, are there any restrictions on the how the points are placed in a 3d space? From https://brlcad.org/wiki/BRL-CAD_Primitives, I gather that it's fine for points to overlap when working with arb8 (not be unique), are there any other such restrictions or constraints?
Jaipal Singh (May 14 2018 at 08:24):arb4-arb7 should have 4-7 unique vertices respectively
Shubham Rathore (May 15 2018 at 18:19):Not actually, its just the overlap of the points that make the difference.
Jaipal Singh (May 15 2018 at 18:21):Ah. Figured that :)
Shubham Rathore (May 15 2018 at 18:27):Went through your dev logs. The primitives look good. they run fine on my system. It'll be good if you document the export instructions for the mged as well. Will save users from some compile time errors ;)
Sean (May 16 2018 at 13:16):Hey @Sean, @Shubham Rathore! I'm travelling home and will be AFK for the next two days (until 12/10/18), I'll resume work on the 13th and will make up for lost time then! :)
@Jaipal Singh thanks for letting people know
Sean (May 16 2018 at 13:24):When working with arb4-arb8, are there any restrictions on the how the points are placed in a 3d space? From https://brlcad.org/wiki/BRL-CAD_Primitives, I gather that it's fine for points to overlap when working with arb8 (not be unique), are there any other such restrictions or constraints?
yes there are restrictions but I don't recall exactly what they are -- probably best to refer to the source code
Jaipal Singh (May 16 2018 at 18:22):Thanks @Shubham Rathore :) I'll get the documentation in place
Jaipal Singh (May 16 2018 at 18:23):@Sean , yea that's what I've been doing and it seems good so far!
Sean (May 18 2018 at 04:29):@Jaipal Singh we need to talk some more about how this will all get leveraged, especially to @Shubham Rathore's earlier point about what is needed to get script objects to work in librt
Sean (May 18 2018 at 04:30):I'm concerned that continuing down the "in" command route will be problematic because it relies on mged
Sean (May 18 2018 at 04:31):mged will not necessarily be available within a librt context (because mged depends on librt, so it would create a cyclic dependency) -- even libged has that problem in a way, so there are considerations to be made...
Jaipal Singh (May 18 2018 at 05:18):@Sean sure. Is there some documentation where I can read about librt / libged and how they work with mged? That'll help me get a clearer idea of the problems we might have.
Jaipal Singh (May 18 2018 at 05:22):Also, I'm posting daily updates to the dev logs here : https://brlcad.org/wiki/User:Schindler97/GSoC18/Log :)
Sean (May 18 2018 at 11:56):not much in the way of architecture docs other than the source code that you have in front of you, the doxygen API docs, and README files scattered throughout (read those....) , that's why I said we should talk more
Sean (May 18 2018 at 11:57):a brl-cad .g file is a "geometry database" filled with named objects. those objects are of a particular registered type. we want to get to a point where we have a "script" type that can be Python.
Sean (May 18 2018 at 11:59):all the logic for reading/writing objects will eventually be in a low-level "libg" library, but right now they live in librt (see src/librt/primitives/table.c to see where they are registered and each subdir for their implementation)
Sean (May 18 2018 at 12:02):you can see librt's current dependencies in its build file (src/librt/CMakeLists.txt), but essentially it just depends on libbu/libbn/libbg and the libraries the primitives themselves require (e.g., libnmg, libbrep, opennurbs)
Sean (May 18 2018 at 12:03):mged is an application that sits on top of libged for all its commands, and libged sits on top of all the other libraries (e.g., librt)
Sean (May 18 2018 at 12:05):so the question / problem is how to make a script primitive that will eventually belong in libg/librt that doesn't depend on mged/libged, but has access to everything in libged...
Sean (May 18 2018 at 12:07):at least not a strict linkage dependency -- the solution probably lies in dynamic runtime loading, or a different runtime object implementation that is somehow loaded by a calling context
Jaipal Singh (May 18 2018 at 13:03):@Sean , can I have time till Sunday to go through all this, think the architecture through and then have a conversation with you about how we would proceed?
Jaipal Singh (May 18 2018 at 13:34):so the question / problem is how to make a script primitive that will eventually belong in libg/librt that doesn't depend on mged/libged, but has access to everything in libged...
But, we want to use the in command and that is a ged command. How would we do away with that?
Also, I went through our conversation from earlier this week and it makes more sense (I think I understand the problem better now)
Jaipal Singh (May 18 2018 at 13:37):a brl-cad .g file is a "geometry database" filled with named objects. those objects are of a particular registered type. we want to get to a point where we have a "script" type that can be Python.
This python script would then be parsed by an embedded interpreter? Which would be called from within the C/C++ code?
Jaipal Singh (May 18 2018 at 13:48):all the logic for reading/writing objects will eventually be in a low-level "libg" library, but right now they live in librt (see src/librt/primitives/table.c to see where they are registered and each subdir for their implementation)
Yes, I'm going through the individual primitives, I did that before/during the community bonding period as well. I'm not able to figure out though how to see "where primitives are registered"
Cezar (May 18 2018 at 13:50):i think he's talking about struct rt_functab OBJ in table.c, regarding where primitives are registered
Jaipal Singh (May 18 2018 at 13:53):Ah ok!
Jaipal Singh (May 18 2018 at 14:09):at least not a strict linkage dependency -- the solution probably lies in dynamic runtime loading, or a different runtime object implementation that is somehow loaded by a calling context
So the "script" would be a primitive defined in librt.
It would parse a python script (myscript.py) which would consist on instructions on how to make complicated objects at runtime.
The "script" code in librt would invoke a python interpreter to understand the input script.
Current Approach : nmz uses his code to create a .tcl -> .g (using mged)
Our Goal? : use the instructions from myscript.py to create a .g file by working with the primitive definitions internally.
Jaipal Singh (May 18 2018 at 14:50):libged seems to have the source code for the commands used by mged, but I can't find anything for the 'in' command. I'm looking for an 'in.c', because for commands like 'ls', there exists an 'ls.c' and I was assuming it would be the same for 'in' as well. (sorry if these doubts come across as extremely trivial)
Cezar (May 18 2018 at 14:55):it's in libged/typein.c
Jaipal Singh (May 18 2018 at 15:03):Thanks a ton @Cezar :D
Jaipal Singh (May 18 2018 at 15:04):@Cezar , how do you look this up? Did you explore through the file listing or is there somewhere you would look this up?
Cezar (May 18 2018 at 15:07):mm... i figured the command strings must be somewhere, so i ran grep -R "\"ls\"" . in the mged folder, and there was an array of objects which contain command strings and their associated functions. so then i ran the search for "in", and it looked like its function was ged_in. so next i went into libged and ran grep -R "^ged_in(" ., and it was found in typein.c
Cezar (May 18 2018 at 15:08):with the last grep there, if you're wondering, i searched for ged_in( at the beginning of a line because that's the coding style brl-cad uses for function definitions
Jaipal Singh (May 18 2018 at 15:10):Aye, great
Sean (May 19 2018 at 03:36):This python script would then be parsed by an embedded interpreter? Which would be called from within the C/C++ code?
Yes, you got it!
Sean (May 19 2018 at 03:37):libged seems to have the source code for the commands used by mged, but I can't find anything for the 'in' command. I'm looking for an 'in.c', because for commands like 'ls', there exists an 'ls.c' and I was assuming it would be the same for 'in' as well. (sorry if these doubts come across as extremely trivial)
The 'in' command is in the src/libged/typein.c file
Sean (May 19 2018 at 03:38):with the last grep there, if you're wondering, i searched for
ged_in(at the beginning of a line because that's the coding style brl-cad uses for function definitions
And that is an outstanding explanation for how to find mged command guts in general!
Daniel Rossberg (May 19 2018 at 11:39):src/mged/setup.c contains the dictionary which translates the Tcl command strings to C functions for mged.
Shubham Rathore (May 20 2018 at 09:44):libged/typein.c is the place ! There holds a structure for a primitive that contains the list of input strings to asked from the user during the runtime. Later you may find functions to load the input to the database.
Jaipal Singh (May 21 2018 at 08:48):@Daniel Rossberg , to create a .g file, is it necessary to have mged to be a part of the pipeline?
Cezar (May 21 2018 at 09:07):not sure, but i don't think so. there are librt functions (see include/rt/db_io.h, for example) which read/write .g files
Daniel Rossberg (May 21 2018 at 11:03):It depends on where you are. If you can link with the BRL-CAD core libraries e.g. db_open(fileName, "rw") opens a .g file for writing.
Daniel Rossberg (May 21 2018 at 11:04):And if this doesn't work db_create(fileName, 5) creates a .g database.
Jaipal Singh (May 21 2018 at 12:04):Ah okay!
I just want to understand better how to complete this work flow :
tcl -> .g . src/mged/setup.c (converts tcl to c commands) -> ??? -> include/rt/db_io.h -> .g
Jaipal Singh (May 21 2018 at 12:07):I have a feeling that I've abstracted a lot of what might go on in the middle.
Jaipal Singh (May 21 2018 at 12:47):src/conv seems to have the relevant scripts to carry out conversions from *format-> .g
Jaipal Singh (May 21 2018 at 12:49):asc2g.c makes a GED database from an ASCII GED data file.
On page 34 : https://brlcad.org/w/images/5/5f/Geometry_Conversions.pdf
From what I understand, tt says that asc2g.c can be used to convert a TCL script to a .g file
Daniel Rossberg (May 21 2018 at 15:56):The mged command to create a new database is opendb, which calls the f_opendb() function defined in src/mged/mged.c.
Daniel Rossberg (May 21 2018 at 16:06):asc2g.c makes a GED database from an ASCII GED data file.
On page 34 : https://brlcad.org/w/images/5/5f/Geometry_Conversions.pdf
From what I understand, tt says that asc2g.c can be used to convert a TCL script to a .g file
Page 34 says something about Pro/Engineer.
In addition, the TCL dialect of asc2g isn't the one used in mged. You can however use mged as a TCL script converter as well. See the mged manual page for instructions .
Jaipal Singh (May 21 2018 at 16:26):Sure! Thanks
Jaipal Singh (May 22 2018 at 06:34):@Daniel Rossberg , we want to convert tcl to .g without getting mged into the pipeline.
Jaipal Singh (May 22 2018 at 06:36):so the question / problem is how to make a script primitive that will eventually belong in libg/librt that doesn't depend on mged/libged, but has access to everything in libged...
@Sean had recommended this and hence we were thinking of alternate possible ways
Sean (May 22 2018 at 11:45):@Jaipal Singh asc2g embeds a tcl interpreter that has just an essential set of commands registered
Sean (May 22 2018 at 11:58):Think I see 3 options...1 call librt/libwdb api , 2 dynamically load libged , 3 create new libg library and call that
Jaipal Singh (May 22 2018 at 12:14):Dynamically load libged?
Jaipal Singh (May 22 2018 at 12:15):Working with libged gives us abstraction over librt and libwdb right?
Jaipal Singh (May 22 2018 at 12:17):Also, @Sean , @Shubham Rathore , I'm done wrapping most of the primitives and have pushed them with some example for you to try out each of them, let me know what you think and where I should head to next.
I'm going through librt and libged to get a coarse understanding of how things are dealt with by them.
Jaipal Singh (May 24 2018 at 16:49):@Shubham Rathore , I'm starting off on documenting the python brlcad on how parts of the code work and how complicated shapes can be drawn. Do you have any particular suggestions that I should keep in mind? :D
Jaipal Singh (May 24 2018 at 17:01):Also, I could only work for a couple of hours yesterday, I have a research paper deadline coming up on the first of June and ended up spending a lot of time with it. I'll make up by working extra hours today and tomorrow :)
Shubham Rathore (May 25 2018 at 00:52):Documentation is indeed required but that is something you should do after the primary tasks are over. Currently the main focus is to convert tcl to geometry files bypassing mged. I was traveling in the past days so couldn't be much active. But we need to figure this out soon.
Jaipal Singh (May 25 2018 at 07:51):In the Hacking-BRL-CAD.pdf, there's a section on procedural geometry, which states that : "BRL-CAD provides a comprehensive procedural geometry interface as a means for creating models algorithmically instead of manually". What exactly is this referring to?
Jaipal Singh (May 25 2018 at 08:19):@Shubham Rathore , I'd agree that's the major priority for the project as of now. I'm a little in the dark here about what could be done to bypass mged and hence thought I could use this time for documenting the project and continue working on the code once I have a clearer picture :)
Jaipal Singh (May 29 2018 at 15:49):Hey @Shubham Rathore , @Sean Should I start working on the script primitive? So far, I've not been able to figure out a way to get procedural geometry to work without using libged. Would you have any leads in that direction?
Sean (May 30 2018 at 05:32):@Jaipal Singh the hacking book is referring to the LIBWDB api.
Sean (May 30 2018 at 05:34):That’s one of the binding options I mentioned last week. We could see what all if anything is missing for what you need.
Sean (May 30 2018 at 05:37):Start there and check out some of the C examples in src/proc-db and src/shapes .. I suspect binding wdb.h is super easy
Jaipal Singh (May 30 2018 at 06:13):@Sean On it!
Shubham Rathore (Jun 05 2018 at 03:30):Hey @Jaipal Singh Can I get some leads on the current work. Did you go through the libwdb ?
Sean (Jun 05 2018 at 04:17):@Sean On it!
how did that go? be sure to update your log please
Jaipal Singh (Jun 05 2018 at 05:43):Hey @Shubham Rathore , @Sean , I was caught up with the conference deadline up until yesterday. :/
I've resumed work on the project from today. Will share updates regularly and make up for lost time ASAP.
Sean (Jun 05 2018 at 11:59):@Jaipal Singh what conference? I don’t see mention of that in your plans. Totally supportive but you need to tell us things like that in advance if they are going to have any impact on your work..I have to imagine more time if you’re accepted?
Jaipal Singh (Jun 05 2018 at 12:00):Hey @Sean , yea, it was careless on my part to not inform you of this a little earlier. This is the conference : http://www.conll.org/
Jaipal Singh (Jun 05 2018 at 12:00):Nope, nothing until the GSoC period ends as any further work related to the conference will not only happen in late August
Jaipal Singh (Jun 05 2018 at 12:01):The submission here was a part of my masters thesis at university and now I'm officially done with all work for the summer so I can focus purely on GSoC
Jaipal Singh (Jun 05 2018 at 12:02):Also, I didn't know I was going to submit a paper to this conference while I was applying for GSoC, hence hadn't included it in the schedule.
Jaipal Singh (Jun 05 2018 at 12:02):But you can be rest assured that I'll finish up everything in time!
Sean (Jun 05 2018 at 12:08):Well that’s good you can focus now, because looking at your plan schedule, we’re way off track...things still being figured out. There needs to be more coding going on by this point.
Jaipal Singh (Jun 05 2018 at 12:11):I've been through libwdb and it seems to be the right thing to handle our problem. It has functions for writing all the primitives. Also went through some of the examples in proc-db and src/shapes and getting a hang of how the shapes are being drawn. I plan to go through all of them to have a clearer idea about how libwdb functions
Jaipal Singh (Jun 05 2018 at 12:11):Yea, I'm sorry for that @Sean :/
Sean (Jun 05 2018 at 12:12):You didn’t know when applying, but you certainly knew after and should have mentioned it once you knew, could have mentioned it in your plan.
Jaipal Singh (Jun 05 2018 at 12:13):Yea, my bad absolutely!
Sean (Jun 05 2018 at 12:17):So what is the next steps coding?
Jaipal Singh (Jun 05 2018 at 12:18):So, first off would be introducing the script primitive in librt to read and parse the scripts
Jaipal Singh (Jun 05 2018 at 12:19):Before that I'd have to define the syntax for the scripts
Jaipal Singh (Jun 05 2018 at 12:21):But what I'm confused about is with libwdb in place now, what's the need for a python interpreter?
Jaipal Singh (Jun 05 2018 at 12:24):Also, how does one interact with the libwdb api? How/where to run src/shapes/bolt
Jaipal Singh (Jun 05 2018 at 12:35):How/where to run src/shapes/bolt
Got this.
Jaipal Singh (Jun 05 2018 at 13:03):But what I'm confused about is with libwdb in place now, what's the need for a python interpreter?
The current python interpreter creates .tcl scripts and then interacts with mged to create .g scripts.
Jaipal Singh (Jun 05 2018 at 13:15):But now that we're using the libwdb API, there's no need for an interim .tcl file.
Though we could have a templating engine written in python to create files similar to some in proc-db
Jaipal Singh (Jun 05 2018 at 13:41):@Sean , @Shubham Rathore , is there a guide/tutorial of sorts for introducing new primitives?
Shubham Rathore (Jun 05 2018 at 13:46):The existing primitives should serve as a reference.
Shubham Rathore (Jun 05 2018 at 13:49):Check out files in librt primarily. An extensive search using the defined primitives would tell you the relevant files to be modified.
Jaipal Singh (Jun 05 2018 at 16:27):Alright @Shubham Rathore
Also, I've updated the dev log.
Shubham Rathore (Jun 05 2018 at 16:31):I'll go through it. Let me know if you get stuck while defining the primitive.
Jaipal Singh (Jun 05 2018 at 16:35):Sure thing! I'm exploring librt to figure out the right things needed to start.
Jaipal Singh (Jun 05 2018 at 17:34):@Sean Do you think introducing a new primitive is going to be a redundant task? We could work with the 'in' command. Personally I think primitives would be great if we were creating something new, but that's not the case here.
Sean (Jun 05 2018 at 17:57):@Jaipal Singh redundant how? There is currently no way to have geometry define itself — this would provide that capability.
Sean (Jun 05 2018 at 17:58):You could even use this to have completely user-defined geometry.
Sean (Jun 05 2018 at 17:58):Instead of the predefined set we support.
Jaipal Singh (Jun 05 2018 at 18:09):@Jaipal Singh redundant how? There is currently no way to have geometry define itself — this would provide that capability.
What do you mean by "the geometry define itself"?
Why I think it's redundant is because the in command already has the ability to create primitives with parameters given to it real time. Hence a parametric approach to buildings things. We could convert this into a procedural approach by modifying it to understand scripts (procedures) and use the currently present architecture to draw the shapes.
Jaipal Singh (Jun 05 2018 at 18:10):Instead of the predefined set we support.
What exactly are you referring to here?
Shubham Rathore (Jun 06 2018 at 12:59):I too didn't understand what redundancy here means . Are you trying to direct towards using some inheritance for the functionalities? If so, then it won't be possible and is again a big project to handle. I think there's a bit of confusion erupting here. Will you please clarify a bit?
Jaipal Singh (Jun 06 2018 at 13:10):So, the redundancy I'm talking about is with regarding the to what the 'in command' already offers.
If a new script primitive is introduced it's job will be to parse the scripts (SC from now) and draw shapes as defined in them. A script usually would define how to create a certain shape using the primitives defined in librt. To create this shape (combination of primitives) we will use the libwdb api. Hence when creating the script primitive we would have to write code to interact with libwdb for each primitive.
Now consider the 'in command'. It currently takes single primitives and arguments to draw them. It already has to code to construct all primitives. What I was recommending is we use these functions. A new script primitive could be introduced to work with the 'in command'. It would parse the SC and construct the primitives one by one and create the bigger shape. This was we can reduce our problem to adding a functionality to the 'in command' to create multiple primitives given a SC. We wouldn't have to write interactions with the libwdb API as there already exists in typein.c code to create these primitives.
Jaipal Singh (Jun 06 2018 at 13:14):Think I see 3 options...1 call librt/libwdb api , 2 dynamically load libged , 3 create new libg library and call that
Last week @Sean had mentioned that we could use the librt or libwdb api for constructing the primitives. Typein.c uses the librt approach to construct primitives already. I'm suggesting we leverage this existing functionality and add to it.
Jaipal Singh (Jun 06 2018 at 13:35):Also, I'm completely down for creating a new primitive but just want your insights on what you think about this idea!
Jaipal Singh (Jun 07 2018 at 03:31):@Sean @Shubham Rathore What do you think?
Sean (Jun 07 2018 at 04:50):What do you mean by "the geometry define itself"?
A main purpose for script objects is having user-defined parametric geometry (e.g., "make a person, male, 6' tall"). They define a shape but are an object, thus they define themselves. All primitives technically do this, so the distinction is really that they define an arbitrary/unknown shape.
Sean (Jun 07 2018 at 05:13):A script usually would define how to create a certain shape using the primitives defined in librt. To create this shape (combination of primitives) we will use the libwdb api. Hence when creating the script primitive we would have to write code to interact with libwdb for each primitive.
This is all good observations. Also realize that several of the primitives are quite flexible/arbitrary (e.g., bot and brep) and will even permit hooking into 3rd party applications. Like imagine a script that reads a Blender file on the fly.
Now consider the 'in command'. It currently takes single primitives and arguments to draw them. It already has to code to construct all primitives. What I was recommending is we use these functions. A new script primitive could be introduced to work with the 'in command'. It would parse the SC and construct the primitives one by one and create the bigger shape. This was we can reduce our problem to adding a functionality to the 'in command' to create multiple primitives given a SC. We wouldn't have to write interactions with the libwdb API as there already exists in typein.c code to create these primitives.
I'm not following. You recommend we use those functions ... but there aren't functions -- there's just one, ged_typein() and binding to ged is a problem. If I'm understanding you correctly, the issue I think is that you see object creation code in typein.c that you want to use -- and you are absolutely right that it should be reusable (and currently is not). I'm not seeing how you could possibly have a script get parsed within libged's 'in' command, so were you thinking to move that logic to librt/libwdb as functions?
Jaipal Singh (Jun 07 2018 at 09:14):My bad, I meant "use this functionality" and not "these functions".
Jaipal Singh (Jun 08 2018 at 06:20):binding to ged is a problem
What does that mean? Is it a problem in general or will it be a problem for me to write code to bind with ged?
how you could possibly have a script get parsed within libged's 'in' command
Yes, I had two ideas. One of them as you mentioned was to move the logic to librt/libwdb as functions. The second was to embed the logic in typein.c (of course this approach is extremely non-modular but serves as a hack). I think including the logic in libwdb is going to be a good choice.
Jaipal Singh (Jun 08 2018 at 06:25):@Sean, is it fair to assume all work from now on will be in C and not python? The script parse can just be written in C
Sean (Jun 08 2018 at 16:12):What does that mean? Is it a problem in general or will it be a problem for me to write code to bind with ged?
It's an architectural problem. libged depends on librt, so you can't introduce code into librt that depends on libged or you create a cyclic dependency. that was the whole discussion a week or so ago when we talked about what the options were (architecturally). however, you note a solution ...
Yes, I had two ideas. One of them as you mentioned was to move the logic to librt/libwdb as functions. The second was to embed the logic in typein.c (of course this approach is extremely non-modular but serves as a hack). I think including the logic in libwdb is going to be a good choice.
Moving the logic to librt/libwdb would work and avoids the cycle. The typein hack won't be adequate for a variety of reasons.
So if you move the typein logic to librt/libwdb, what are your thoughts on having multiple functions for creating objects (essentially different constructors)?
Jaipal Singh (Jun 08 2018 at 17:13):Yes, have multiple constructors (functions) and an array of constructors would be called to create the primitives sequentially as mentioned in the script.
Jaipal Singh (Jun 08 2018 at 17:20):There's still some doubts that I have. Libwdb has functionality to draw the primitives (the mk_primitive() functions). An example where I've seen these being to draw objects are in files like proc-db/wdb_example.c .
The doubt is, at run time, our script logic (present in libwdb) would then invoke these mk_primitive() functions. Will these be rendered in mged in real-time?
Also, where and how does python do you think can help us in this project now? The python-BRLCAD that I had worked on in the first two weeks is completely independent of what we're planning to do now. I personally feel that we can just continue with C.
Jaipal Singh (Jun 08 2018 at 17:32):I also want to re-frame my plan for the next two months given the new tasks at hand and include hence asking all these questions before I get started coding.
I've also been looking for the code responsible for uploading files using mged, but I can't find any. Though the problem could be tackled in two ways, we can maybe not upload the file but read it at run time. That raises the problem of passing the absolute or relative path that we would have to then tackle.
Sean (Jun 08 2018 at 17:33):Yes, have multiple constructors (functions) and an array of constructors would be called to create the primitives sequentially as mentioned in the script.
I'm not understanding this... you'd only need to call one constructor per statement
Sean (Jun 08 2018 at 17:33):what's sequential?
Jaipal Singh (Jun 08 2018 at 17:34):Yea, it would be a constructor per statement. Sequentially, as in, calling the constructors in the same order as mentioned in the script.
Jaipal Singh (Jun 08 2018 at 17:45):I'm thinking the immediate task for me right now is to figure out the file upload or file-read part. I'll try and find the code mged uses for uploading files and revert back to you.
Sean (Jun 08 2018 at 17:59):So I was referring to different constructors for the same entity, like specifying a sphere with a "point and a radius" or a "point and a direction vector" or "point and three orthogonal vectors" ... all three make a sphere but the orientation is technically different for each
Sean (Jun 08 2018 at 17:59):s/technically/potentially/
Sean (Jun 08 2018 at 18:00):I'm thinking the immediate task for me right now is to figure out the file upload or file-read part. I'll try and find the code mged uses for uploading files and revert back to you.
to be honest, this feels like a distraction as it's not coding and you've gone several weeks without coding now, no?
Sean (Jun 08 2018 at 18:01):you really need to be submitting code daily by this point, something even if it's not perfect and you're learning as you go
Sean (Jun 08 2018 at 18:01):you'll learn more and we'll have more to talk about.
Sean (Jun 08 2018 at 18:04):you mentioned a while back that there were two python wrappers you'd looked at -- one that used mged directly and another. can you explain what exactly the other one was?
Sean (Jun 08 2018 at 18:07):The doubt is, at run time, our script logic (present in libwdb) would then invoke these mk_primitive() functions. Will these be rendered in mged in real-time?
as an object, absolutely yes -- one of the functions that every primitive defines is how to draw itself. for a script object like this, it will just tell it to draw the geometry it created.
Also, where and how does python do you think can help us in this project now? The python-BRLCAD that I had worked on in the first two weeks is completely independent of what we're planning to do now. I personally feel that we can just continue with C.
my thinking is that the basic object structure would get stubbed in a day or two (I could probably do it today) and you'd spend most of your time defining the C-to-Python binding and then all your remaining time on the Python side creating a more seamless full API
Sean (Jun 08 2018 at 18:09):in essence, do whatever you need to do on the python side so that you can do what wdb_example is doing from python -- which may be already done if the non-mged approach (ctypesgen?) was complete enough
Jaipal Singh (Jun 09 2018 at 07:11):you really need to be submitting code daily by this point, something even if it's not perfect and you're learning as you go
Starting today, I shall do this.
Jaipal Singh (Jun 09 2018 at 07:23):in essence, do whatever you need to do on the python side so that you can do what wdb_example is doing from python -- which may be already done if the non-mged approach (ctypesgen?) was complete enough
I think the ctypesgen approach followed the mged-approach and was complete enough. It though was designed to work with brlcad-7.24.0 and raises a lot of errors due to some changed header files when working with the latest version. I'll fix these and push changes right away.
Cezar (Jun 09 2018 at 07:25):try the latest version from github first. i had submitted a pull request that fixed those errors and was merged. if it works, you don't have to fix those problems yourself
Jaipal Singh (Jun 09 2018 at 07:25):Yea, just saw the patch you had submitted. It should fix the issues. :)
Jaipal Singh (Jun 09 2018 at 13:35):@Cezar , seems to be working, though there's still error logs about stdio.h, plo3.h and nmg.h. I tried a fix similar to yours and tried to include the header files in include/bn but that doesn't seem to fix it.
Cezar (Jun 09 2018 at 13:49):what/where are those error logs? :-?
Cezar (Jun 09 2018 at 13:49):if it's during building, i think it's fine
Cezar (Jun 09 2018 at 13:50):i think i do remember some problems there, but i think i wrote them off for some reason
Jaipal Singh (Jun 09 2018 at 13:50):Yea, it's during building. Guess it shouldn't be an issue.
https://pastebin.com/YPpP2LkJ
Cezar (Jun 09 2018 at 13:52):i think it's because during preprocessing, some headers are not included in other headers. so for example, nmg.h uses FILE but does not include stdio.h, so when looking at the header, ctypesgen (?) thinks it's an error. but nmg.c might #include <stdio.h> #include "nmg.h", and when nmg.c is compiled, no problem arises
Jaipal Singh (Jun 09 2018 at 13:58):Uh-huh makes sense. I guess this should never be an issue at run-time then.
Jaipal Singh (Jun 09 2018 at 14:01):my thinking is that the basic object structure would get stubbed in a day or two
@Sean , I didn't get this
Sean (Jun 09 2018 at 18:06):i think it's because during preprocessing, some headers are not included in other headers. so for example, nmg.h uses FILE but does not include stdio.h, so when looking at the header, ctypesgen (?) thinks it's an error. but nmg.c might
#include <stdio.h> #include "nmg.h", and when nmg.c is compiled, no problem arises
That might actually be a good testing framework for making sure headers are appropriate encapsulated... or it's missing some critical defines (quite likely) as I believe stdio.h is included by everything via common.h
Cezar (Jun 09 2018 at 18:09):That might actually be a good testing framework for making sure headers are appropriate encapsulated... or it's missing some critical defines (quite likely) as I believe stdio.h is included by everything via common.h
hmm... i was thinking of rob pike's notes on programming in c (here), where he says that including header files in header files slows down compilation
Sean (Jun 09 2018 at 18:29):looking at a few of them, it's not clear if this is a false positive or not -- some of them clearly don't include stdio.h directly, but at least one does indirectly and it still reported
Sean (Jun 09 2018 at 18:30):it can slow down compilation, especially for big complicated headers where you only need a type or something simple
Sean (Jun 09 2018 at 18:31):that said, performance-wise, it's pretty much an insignificant blip on development unless people pathologically include too many headers needlessly
Sean (Jun 09 2018 at 18:32):headers should be self contained, including and declaring what they need so that header stands on it's own. otherwise, you're just trading a tiny fraction of development cpu time with human time (which is FAR FAR FAR more expensive), especially when they fail to include a required header and spend an hour or three debugging
Sean (Jun 09 2018 at 18:56):Yea, it's during building. Guess it shouldn't be an issue.
https://pastebin.com/YPpP2LkJ
@Jaipal Singh I just committed a bunch of header additions that should take care of them (r71053). would you try again and report back?
Jaipal Singh (Jun 10 2018 at 05:01):@Sean , the build fails due to some missing header files : https://pastebin.com/YyhhdYMG
Jaipal Singh (Jun 10 2018 at 05:02):The weird part is when I run make -j 3, I get the error that I shared above,
When I just run make on one core, it gives a different error https://pastebin.com/Q9uX0h26
Jaipal Singh (Jun 10 2018 at 05:06):I'll pull the stable version from the trunk and introduce the changes (r71053) and see if that works
Sean (Jun 10 2018 at 05:14):huh, that's interesting -- @Jaipal Singh you apparently have a system-installed poly2tri that is different than the one BRL-CAD is using. since it's installed in /usr/local, that takes precedence over ours in src/other/poly2tri ... and your system one looks misconfigured or has a bug.
Sean (Jun 10 2018 at 05:15):couple workarounds possible -- easiest being: sudo mv /usr/local/include/poly2tri /usr/local/include/poly2tri.backup
Sean (Jun 10 2018 at 05:17):alternatively, commenting out the block in the top-level CMakelLists.txt that adds /usr/local (around "# We want to check /usr/local by default, so add it if it exists")
Sean (Jun 10 2018 at 05:19):the difference with one core is just in how the dependencies are resolved and what order they are compiled. doesn't mean anything other than possibly another issue that we'll see later with the -j3 build when it gets to that point.
Sean (Jun 10 2018 at 05:20):pulling the stable from trunk won't likely fix the poly2tri issue because it's an issue on your system
Jaipal Singh (Jun 10 2018 at 05:20):pulling the stable from trunk won't likely fix the poly2tri issue because it's an issue on your system
Yes, moving it out to poly2tri,backup seems to have solved the issue for now
Jaipal Singh (Jun 10 2018 at 05:21):my thinking is that the basic object structure would get stubbed in a day or two
@Sean Some clarity on this?
Sean (Jun 10 2018 at 05:22):the whole script object portion is not dependent upon what you need to sort out with ctypesgen and/or librt and/or libwdb and/or moving code from libged/typein.c
Sean (Jun 10 2018 at 05:22):but if we want scripting objects, they obviously need to exist
Sean (Jun 10 2018 at 05:24):well that work is really dead simple .. I can probably do it in a day if/when needed, at least the most important bits. you could probably do it in just a few days at most (but don't have to, would rather you focus closer to the python-specific needs)
Jaipal Singh (Jun 10 2018 at 05:24):Alright!
So, the build fails again : https://pastebin.com/8rb6q2iX
Sean (Jun 10 2018 at 05:26):would you switch to https://pastebin.ca or the debian pastebin or literally anything other than pastebin.com ... they're a really obnoxious site often with maliciousness going on
Sean (Jun 10 2018 at 05:26):and inaccessible to some devs (blocked for those same reasons)
Jaipal Singh (Jun 10 2018 at 05:26):Sure, didn't know that
Sean (Jun 10 2018 at 05:27):so that lempar issue is really odd
Sean (Jun 10 2018 at 05:27):what version of cmake and make are you using?
Jaipal Singh (Jun 10 2018 at 05:27):3.10.3
Jaipal Singh (Jun 10 2018 at 05:28):3.11.3 is the latest
Sean (Jun 10 2018 at 05:28):hrm, so then I think full logs are in order on a clean build
Sean (Jun 10 2018 at 05:28):yes, that version is fine
Sean (Jun 10 2018 at 05:29):anything 3.3+ should be fine
Jaipal Singh (Jun 10 2018 at 05:29):Ok..
Sean (Jun 10 2018 at 05:31):oh you know what, I bet it's a similar issue
Sean (Jun 10 2018 at 05:31):you probably have an old system lemon installed somewhere
Jaipal Singh (Jun 10 2018 at 05:31):A prior messed up installation?
Sean (Jun 10 2018 at 05:32):no, not likely
Sean (Jun 10 2018 at 05:32):what's your cmake line?
Jaipal Singh (Jun 10 2018 at 05:32):cmake .. -DCMAKE_INSTALL_PREFIX=/usr/local
Sean (Jun 10 2018 at 05:33):ok
Sean (Jun 10 2018 at 05:33):so run: rm -rf CMake*
Jaipal Singh (Jun 10 2018 at 05:33):Done and now cmake again?
Sean (Jun 10 2018 at 05:34):then: cmake .. -DCMAKE_INSTALL_PREFIX=/opt/brlcad -DBRLCAD_BUNDLED_LIBS=ON
Sean (Jun 10 2018 at 05:34):recommend against installing into /usr/local or other common system dir until you get more experienced
Sean (Jun 10 2018 at 05:35):there are a variety of issues that can come up, naming conflicts
Sean (Jun 10 2018 at 05:36):the important part is bundled libs
Jaipal Singh (Jun 10 2018 at 05:37):Yea, what does it do?
Sean (Jun 10 2018 at 05:37):BRL-CAD uses about 2 dozen external libraries and tools like libpng, libz, tcl/tk, etc
Jaipal Singh (Jun 10 2018 at 05:37):Also, should I go ahead and make it now?
Sean (Jun 10 2018 at 05:38):we bundle them so everything will build out of the box whether you have them installed or not, whether you have rpm/dpkg/homebrew/apt/whatever package manager
Sean (Jun 10 2018 at 05:38):and this makes things much simpler on platforms like Windows that have nothing
Jaipal Singh (Jun 10 2018 at 05:38):Alright, so they're pre-compiled binaries?
Sean (Jun 10 2018 at 05:39):but it is just for convenience ... you can download all 20+ dependencies yourself, compile/install them, use rpms/apt/whatever
Sean (Jun 10 2018 at 05:39):not pre-compiled, we still compile them as needed
Jaipal Singh (Jun 10 2018 at 05:40):but it is just for convenience ... you can download all 20+ dependencies yourself, compile/install them, use rpms/apt/whatever
True, but there's a fair chance for something messing up here.
Sean (Jun 10 2018 at 05:40):so by default, we test and see if you have a given external dep, like 'tcl/tk' and ... 'lemon' and if it finds it, we don't build ours -- we assume the system one we found is good to use
Jaipal Singh (Jun 10 2018 at 05:40):Uh-huh
Jaipal Singh (Jun 10 2018 at 05:41):But then what about version compatibility? How often do you have to update the bundled libs?
Sean (Jun 10 2018 at 05:42):that varies from dependency to dependency, but where it matters we typically test and only use a system one if it's a version we require (or newer)
Jaipal Singh (Jun 10 2018 at 05:43):Cool
Sean (Jun 10 2018 at 05:43):the bundled libs are updated as we need / want / care to .. those are the ones that typically always work because we made sure they work
Sean (Jun 10 2018 at 05:43):so when you compile, the default is "Auto" where it'll detect what you have and use it if it seems to be version-compatible
Sean (Jun 10 2018 at 05:44):the other modes are Bundled == Off and Bundled == On which tell cmake to either ONLY use system dependency libs or NEVER use system dependency libs
Jaipal Singh (Jun 10 2018 at 05:45):Does it fall back to using bundled libs if system dependencies are missing?
Sean (Jun 10 2018 at 05:45):for Auto, sure, it has to -- that's the whole point
Sean (Jun 10 2018 at 05:46):for Bundled, it doesn't care what you have on the system, it's not going to use it
Jaipal Singh (Jun 10 2018 at 05:46):Alright
Jaipal Singh (Jun 10 2018 at 05:47):Sean, could you help me chalk out how I should progress with the project now. The current python BRLCAD uses ctypesgen that allows us to interact with the libwdb api from python.
Sean (Jun 10 2018 at 05:48):to be pedantically accurate, the cmake variable is BRLCAD_BUNDLED_LIBS=AUTO or =SYSTEM or =BUNDLED (which is same as =ON)
Jaipal Singh (Jun 10 2018 at 05:48):Ok yea.
Sean (Jun 10 2018 at 05:48):so it does use libwdb?
Sean (Jun 10 2018 at 05:48):that much wasn't clear/known
Sean (Jun 10 2018 at 05:49):who were the main authors of the two different versions ... there was nmz, pedro, and .... ?
Jaipal Singh (Jun 10 2018 at 05:49):I've been going through the code figured most of it out yesterday (there's still some parts that throw me off)
Jaipal Singh (Jun 10 2018 at 05:49):kanzure
Sean (Jun 10 2018 at 05:49):and who did which?
Jaipal Singh (Jun 10 2018 at 05:51):nmz was in favour of the mged approach, the one I worked on in the first two weeks of the project. That's currently loaded with functionality to create all the primitives.
kanzure worked on the ctypesgen project. I was in touch with him before the coding period began to get some help in choosing which project to continue with, the ctypesgen one or the mged one.
Sean (Jun 10 2018 at 05:56):okay, so where does that code seem to stand?
Jaipal Singh (Jun 10 2018 at 05:56):Which one are you referring to?
Sean (Jun 10 2018 at 05:57):you said the mged-based one was loaded with functionality to create (nearly) all the primitives ... so where's the ctypesgen one at?
Jaipal Singh (Jun 10 2018 at 05:57):The ctypesgen one is missing on a few primitives. Once those are in place it should be at part with the mged one
Sean (Jun 10 2018 at 05:57):how does the python API compare between the two, how many primitives are hooked up?
Jaipal Singh (Jun 10 2018 at 06:00):The mged approach technically does not use any API, it makes a system call using to mged and generates the.g file
Jaipal Singh (Jun 10 2018 at 06:01):The ctypesgen approach usues ctypes to interact with the libwdb api and create the .g files
Sean (Jun 10 2018 at 06:01):do you think you could easily show what the equivalent to wdb_example looks like in each (mged and ctypesgen versions)? how long might that take you?
Jaipal Singh (Jun 10 2018 at 06:02):I could give it to you in a couple of hours from now?
Jaipal Singh (Jun 10 2018 at 06:02):I think for the ctypesgen version, there's already an wdb_example equivalent, 1 second
Sean (Jun 10 2018 at 06:02):that would be great to see
Jaipal Singh (Jun 10 2018 at 06:02):https://github.com/kanzure/python-brlcad/blob/master/examples/wdb_example.py
Sean (Jun 10 2018 at 06:05):so yeah, then see what that looks like with the other one would be great for comparison
Sean (Jun 10 2018 at 06:06):lots of little choices being made there -- method on a class, sets of global classes, using intrinsic types or not, etc
Jaipal Singh (Jun 10 2018 at 06:07):I still have to figure out how to execute operations like "union" or "intersection" via the mged approach. I'll get those sorted and revert back to you with an equivalent script for the mged version?
Jaipal Singh (Jun 10 2018 at 06:07):Yea!
Sean (Jun 10 2018 at 06:07):sure .. it's succinct enough that I can't imagine it'll take too long
Jaipal Singh (Jun 10 2018 at 06:09):Yes! I have to quickly run some errands outside (sunday chores) and I'll be back and update you with the script? Won't be very long for sure!
Sean (Jun 10 2018 at 06:10):then to answer your question, I think the next step will be to simply add a primitive to the wdb one that seems missing, just to get familiar with the process and code, then create a spreadsheet to itemize what's there and what's missing, and then work on a second slightly more complex primitive (again for experience) while I get the script object set up, and then you test running an embedded python in C to run one of the scripts successfully, and then we pull it all together!
Sean (Jun 10 2018 at 06:11):sounds good
Shubham Rathore (Jun 10 2018 at 07:08):@Jaipal Singh are you done with the installation errors? I suggest you to go through the "INSTALL" file. lines from 190 onwards, will clear the picture for the installation flags. As far as the ctypesgen approach is considered, that code is totally broke, with some dependency issues, import issues, version issues, platform issues etc. Basically the stuff is incomplete, and the progress sounds hazy , that was the reason to shift to the tcl approach. We should keep these things in mind before we think forward. Looking forward to a code snippet from you. Cheers !
Jaipal Singh (Jun 10 2018 at 11:45):@Sean , the errors still show up when setting python brlcad (ctypesgen) due to the header files even after including your patch
Jaipal Singh (Jun 10 2018 at 11:50):https://gist.github.com/Killthebug/99d5aa4d6254e1dce083c784fedd7c25
Jaipal Singh (Jun 10 2018 at 16:02):@Sean , here's the script for wdb_example using the tcl approach.
https://github.com/Killthebug/python-brlcad-tcl/blob/master/examples/wdb_example.py
Jaipal Singh (Jun 14 2018 at 13:28):@Sean , here's the list of primitives already incorporated into the ctypesgen code. https://docs.google.com/spreadsheets/d/1XeeI1mpLlPFZtc7i5GeOmjLFE-s8JMsYpwVszIPsrmQ/edit?usp=sharing
There's individual files for most of the primitives and the others are inherited from their parent class.
I've gone through all the primitives and am fairly familiar with the code. As for the complicated primitive, which one would you recommend?
Also, what would be a good time to get started on setting up the script object? I'd like to give it a shot myself but am a little in the dark about how I should go about doing it.
Jaipal Singh (Jun 19 2018 at 05:09):Hey, anything on this front?
Shubham Rathore (Jun 19 2018 at 21:05):I think you can include hrt primitive as well. Might be useful in procedural geometry.
Shubham Rathore (Jun 19 2018 at 21:09):And start working on the script primitive structure. That is something you can do in very less time. Just read the code containers for the other primitives, and move along the template. The basic primitive definition lies in librt. I suggest, start from here and ping for further clarifications. You need to start doing stuff, that will make things more clear. You need to work smartly here, as many fields may remain empty for this primitive. Please ensure the updation of your daily logs.
Shubham Rathore (Jun 19 2018 at 21:11):Later our task remains to incorporate the primitive container and the ctypes bindings, but before that the primitive needs to be well defined. librt is the place where you should focus initially. Pick up a primitive as a reference and follow accordingly. Cheers !
Shubham Rathore (Jun 21 2018 at 15:12):@Jaipal Singh any updates? Are you stuck on anything?
Jaipal Singh (Jun 21 2018 at 18:00):Hey! I've been working on the script parser! I'll finish that up by tomorrow, I have to add some more primitive functions. Will move onto the script primitive then.
Jaipal Singh (Jun 22 2018 at 05:37):You can have a look at the script parser here : https://github.com/Killthebug/python-brlcad/blob/master/examples/script.py
Shubham Rathore (Jun 22 2018 at 07:09):Some questions for you. Weren't we going the ctypes way? All I can figure out is tcl-> .g in this script. and how is mged being bypassed here?
Jaipal Singh (Jun 22 2018 at 10:00):Hey Shubham! This is is the ctypesgen approach.
First off, I've chosen tcl as the scripting language here because tcl is already used in the brlcad ecosystem and hence I don't have to introduce a new script BNF.
Jaipal Singh (Jun 22 2018 at 10:00):Secondly, this is the ctypegen approach. mged is being bypassed here.
We use ctyepgen to generate bindings for libwdb. We then use the libwdb API in python to create primitives.
Jaipal Singh (Jun 22 2018 at 10:01):Until now, you would have to write separate code if you wanted to make a shape comprising of multiple primitives, something like : https://github.com/Killthebug/python-brlcad/blob/master/examples/wdb_example.py
Jaipal Singh (Jun 22 2018 at 10:04):With the scripting option that I've not introduced, you can give it tcl scripts like : https://github.com/Killthebug/python-brlcad/blob/master/examples/temp.tcl to create the same objects.
Jaipal Singh (Jun 22 2018 at 10:04):The scripting option is not completely ready. It can make all the primitives for now but still lacks functionality to deal with variables, primitive rotation, etc. I'm working on these additions right now.
Jaipal Singh (Jun 22 2018 at 10:05):As a part of procedural geometry, we can also have loops in the script to create multiple instances of the same objects with different variations. Let me know what you think about that and I'll get started on it once this part is completed.
Shubham Rathore (Jun 22 2018 at 11:20):Great !
Shubham Rathore (Jun 22 2018 at 11:25):Looks good ! I'll be waiting for it to run this way.
Shubham Rathore (Jun 22 2018 at 11:27):And what about the script primitive? Have you started working on that ?
Jaipal Singh (Jun 22 2018 at 13:47):So, the script primitive will essentially be responsible for invoking the python script parser that I've shared above. It will be python embedded in C. Once I am done perfecting the parser, I'll start work on the script primitive. Sound good?
Shubham Rathore (Jun 24 2018 at 15:24):okay, make sure you update your logs ! I find a lag there. Cheers !
Jaipal Singh (Jun 26 2018 at 05:25):Yes, I'll keep them updated @Shubham Rathore :)
Jaipal Singh (Jun 26 2018 at 05:26):@Sean , here's a sample script which the script parser will work with to create the procedural geometry. https://paste.debian.net/1030745/
Jaipal Singh (Jun 26 2018 at 05:27):This is inspired from the sgicube.sh (https://brlcad.org/wiki/SGI_Cube)
Jaipal Singh (Jun 26 2018 at 05:28):As of now the script parser can work with variables and creating primitives. Right now, I'm working on introducing procedure parsing (function parsing) in to the script. Do you think the procedure structure in the sample script I've shared looks fine?
Jaipal Singh (Jun 26 2018 at 07:53):Also, could you have a look at the workflow of the script parser and certain assumptions regarding the scripts that would be parsed?
https://github.com/Killthebug/python-brlcad/commit/b2bb4eb9ba4c830e17c80743310af31fa544205a
Jaipal Singh (Jun 29 2018 at 04:15):@Sean , @Shubham Rathore Any inputs on this? ^
Shubham Rathore (Jun 30 2018 at 14:39):@Jaipal Singh Looks ok on the first view. have you tested it ?
Jaipal Singh (Jun 30 2018 at 14:42):Yup! I test each part before I commit it! :)
Shubham Rathore (Jun 30 2018 at 14:47):How do you exactly plan to introduce function parsing into the script ?
Shubham Rathore (Jun 30 2018 at 14:47):Though the procedure structure looks fine in the script
Shubham Rathore (Jun 30 2018 at 14:50):Regarding the workflow, we'll have to see its feasibility at the run time . Fine for now ! Please try to complete this before your second evals. The only part left after the second eval should be the script primitive. I have a basic frame ready for it, so that shouldn't take much time. Plus linking different parts of the pipeline will comprise of many errors, we need time for that too. I hope you'll follow
Jaipal Singh (Jul 02 2018 at 05:10):I am creating a parser that will work with the tcl scripts based on the procedure structure that i've shared with you. It's somewhat like designing a compiler (not as complicated) where I'll be calculating values of local variables based on arguments passed and global variables available, every time a procedure is called.
Sean (Jul 04 2018 at 04:25):I've gone through all the primitives and am fairly familiar with the code. As for the complicated primitive, which one would you recommend?
bot would probably be best to start with since it's somewhat more complicated than the other primitives, but a concern -- can we make changes on the C/C++ side so that there doesn't need to be primitive-specific bindings on the Python side?
Also, what would be a good time to get started on setting up the script object? I'd like to give it a shot myself but am a little in the dark about how I should go about doing it.
If you'd like to give it a shot, start with src/librt/primitives/table.cpp -- notice how it's a bunch of ID's and functions. look at one of the newest IDs at the bottom, like ID_DATUM, and then find all references to it throughout the source tree (grep -r ID_DATUM include src). I can certainly help you with this and/or implement it for you, but glad to give you a shot if you want to try.
Jaipal Singh (Jul 04 2018 at 04:55):bot would probably be best to start with since it's somewhat more complicated than the other primitives, but a concern -- can we make changes on the C/C++ side so that there doesn't need to be primitive-specific bindings on the Python side?
Yea, we can make changes on the C/C++ side without impacting the bindings on the python side.
Jaipal Singh (Jul 04 2018 at 05:17):If you'd like to give it a shot, start with src/librt/primitives/table.cpp -- notice how it's a bunch of ID's and functions. look at one of the newest IDs at the bottom, like ID_DATUM, and then find all references to it throughout the source tree (grep -r ID_DATUM include src). I can certainly help you with this and/or implement it for you, but glad to give you a shot if you want to try.
I'll be moving on to do this once I've completed the script parser that I'm writing. It'll be done by the second eval. Just yesterday I completed an syntax parser which can parse mathematical expressions in procedural scripts :D (Got to be one of the most exciting things I've learnt and built so far)
Jaipal Singh (Jul 04 2018 at 05:18):Procedural scripts have expressions like : [exp{2+4{$i+$j/$old}*22}]. I created a syntax parser to parse such expressions.
Sean (Jul 04 2018 at 05:37):As of now the script parser can work with variables and creating primitives. Right now, I'm working on introducing procedure parsing (function parsing) in to the script. Do you think the procedure structure in the sample script I've shared looks fine?
@Jaipal Singh I may have missed a major detail here ... it should just read Python code -- what's the custom parsing??
Sean (Jul 04 2018 at 05:41):I am creating a parser that will work with the tcl scripts based on the procedure structure that i've shared with you. It's somewhat like designing a compiler (not as complicated) where I'll be calculating values of local variables based on arguments passed and global variables available, every time a procedure is called.
I really don't think we want that -- if we did, we should just use Tcl... there's super complicated syntax and constructs possible even with Tcl that would not be productive to implement. keeping the scope limited to Python is perfectly fine for now. we don't need multiple points of failure and a special purpose language pretending to be a standard lanaguage. :)
Sean (Jul 04 2018 at 05:44):Yea, we can make changes on the C/C++ side without impacting the bindings on the python side.
That's not exactly answering what was asked... :) Can I make changes on the C/C++ side, like add a new parameter to an ELL primitive that is made available on the Python side without changing a ell.py file or something similar? Basically, some sort of automatically reflective binding or discovered bindings or some other mechanism.
Sean (Jul 04 2018 at 05:50):I'll be moving on to do this once I've completed the script parser that I'm writing. It'll be done by the second eval. Just yesterday I completed an syntax parser which can parse mathematical expressions in procedural scripts :D (Got to be one of the most exciting things I've learnt and built so far)
It's awesome that you've been having fun. Implementing parsers can be SUPER exciting when they start working. One area where that may come in handy down the road is in ascertaining the security of a given script before it's passed to a parser. For now, though, I do think we need to keep it focused on Python and how object creation is exposed as an API -- what changes are needed to make that better.
Jaipal Singh (Jul 04 2018 at 05:54):I really don't think we want that -- if we did, we should just use Tcl... there's super complicated syntax and constructs possible even with Tcl that would not be productive to implement. keeping the scope limited to Python is perfectly fine for now. we don't need multiple points of failure and a special purpose language pretending to be a standard lanaguage.
I've followed the tcl syntax for writing expressions as [exp....], similar to what is defined here : http://wiki.tcl.tk/583.
Jaipal Singh (Jul 04 2018 at 05:55):That's not exactly answering what was asked... :) Can I make changes on the C/C++ side, like add a new parameter to an ELL primitive that is made available on the Python side without changing a ell.py file or something similar? Basically, some sort of automatically reflective binding or discovered bindings or some other mechanism.
No that'll be an issue, but only at the argument level. So, the ellipsoid primitive is called using this file : https://github.com/Killthebug/python-brlcad/blob/master/brlcad/primitives/ellipsoid.py.
Jaipal Singh (Jul 04 2018 at 05:56):If you change something within the working of the primitive in the C code, it shouldn't be an issue, but if you change the number of parameters being passed to it, then that has to be manually updated here.
Jaipal Singh (Jul 04 2018 at 05:57):^ Does this make sense?
Jaipal Singh (Jul 04 2018 at 05:59):Let's say you change how an ell is drawn in the c code. In that case the code on the python side can still continue to work as the internal working of the primitive are abstracted
Jaipal Singh (Jul 04 2018 at 06:03):Bindings are automatically generated according to the version of brlcad you're using when you setup python brlcad. Though the functions that work with these bindings are already coded like the ellipsoid.py file that I had shared with you above. As long as how you interact with the primitives (arguments passed) doesn't change, there isn't going to be an issue.
Jaipal Singh (Jul 04 2018 at 06:22):@Sean , here's a sample script which the script parser will work with to create the procedural geometry. https://paste.debian.net/1030745/
This is inspired from the sgicube.sh (https://brlcad.org/wiki/SGI_Cube).
This (https://imgur.com/a/Kzi8CkF) is what the "script parser" does. It's made to handle procedural scripts like the one I shared above.
Shubham Rathore (Jul 09 2018 at 08:56):Let's have something for the reflectance here. Manual updation sounds like a pain. Once you finish your primary work, I think we can have a script that continuously updates the changes made on one side to the other.
Shubham Rathore (Jul 09 2018 at 09:00):and one more thing-- DEVLOGS ;)
Jaipal Singh (Jul 10 2018 at 03:53):@Shubham Rathore YES! I think we can brew something up! Primary work is almost done. Just doing some finishing touches and making sure that the parser is robust enough. I'll also document the code soon, so far I've not done any major documentation other than code comments.
Yes, Devlogs, DONE! :D
Sean (Jul 10 2018 at 03:58):No that'll be an issue, but only at the argument level.
Can you think of any ways to eliminate that issue? even at the argument level? For example, most of the primitive bindings including the ellipsoid.py you linked are straight pass-through parameters (a = data.a, b = data.b, etc). What would be some strategies/options for doing that automatically?
Sean (Jul 10 2018 at 04:01):@Sean , here's a sample script which the script parser will work with to create the procedural geometry. https://paste.debian.net/1030745/
This (https://imgur.com/a/Kzi8CkF) is what the "script parser" does. It's made to handle procedural scripts like the one I shared above.
@Jaipal Singh I read this all earlier as that is what prompted the "what is the custom parsing" question...
Sean (Jul 10 2018 at 04:02):my comments stand that we really should be focusing exclusively on python, not tcl
Sean (Jul 10 2018 at 04:07):The reason being is that while what you wrote looks a lot like Tcl, it's technically not Tcl. It's fun to implement, but there are severe usability implications... I'm sure I could very easily add a few lines of valid Tcl that would break it ... but more importantly, I would wonder why wouldn't we just use a Tcl interpreter.
Anyone can implement the SGI cube script in a #/opt/brlcad/bin/btclsh script today, call any ged command, and use any Tcl construct. What's the value of having a different script interpreter? More importantly ... what aren't we getting done on the python side because of the detour? :)
Jaipal Singh (Jul 10 2018 at 04:09):@Sean , I'll get back to you regarding the arguments soon.
So how would a script on the python side look like?
Sean (Jul 10 2018 at 04:12):So how would a script on the python side look like?
that is what this project is principally concerned with figuring out, no? :)
Sean (Jul 10 2018 at 04:12):you have one example at https://github.com/Killthebug/python-brlcad/blob/master/examples/wdb_example.py and that's a start but I think quite inadequate
Jaipal Singh (Jul 10 2018 at 04:13):Yea :P
So, this is what the wdb_example.py looks like : https://github.com/Killthebug/python-brlcad/blob/master/examples/wdb_example.py
But there's no procedural elements in this
Sean (Jul 10 2018 at 04:13):do you think you could take a stab at implementing SGI cube in "ideal python"
Jaipal Singh (Jul 10 2018 at 04:14):Yup. I think I can do that!
Sean (Jul 10 2018 at 04:14):if you figure out what that looks like, then the next step would be to figure out how to make that happen
Sean (Jul 10 2018 at 04:14):and there are several decision points along the way
Jaipal Singh (Jul 10 2018 at 04:14):Uh-huh
Sean (Jul 10 2018 at 04:14):it could be straight up procedural like Tcl
Sean (Jul 10 2018 at 04:14):or object-oriented
Sean (Jul 10 2018 at 04:15):what's currently bound through ctypesgen looks like a hybrid
Sean (Jul 10 2018 at 04:15):could maybe build a OO version on the ctypesgen hybrid or change the interface it's being generated off of on the C header side
Sean (Jul 10 2018 at 04:16):i.e., have some low-level version that is essentially a straight up binding to libwdb/librt, but then a nicer OO python layer on top
Jaipal Singh (Jul 10 2018 at 04:18):change the interface it's being generated off of on the C header side
That would almost be like scraping off ctypesgen and starting from scratch, because ctypesgen creates the bindings for libwdb (I'm still thinking of a way of how we can go lower than this)
Jaipal Singh (Jul 10 2018 at 04:18):But, um, can you give me some time to research and experiment a little bit? I don't want to drop ideas without testing out a few things :P
Sean (Jul 10 2018 at 04:19):so yeah, looks like kanzure's exisiting code was an attempt at exactly what I'm suggesting, so the next step is probably to write a "pure" version without focusing on the backend
Sean (Jul 10 2018 at 04:20):That would almost be like scraping off ctypesgen and starting from scratch, because ctypesgen creates the bindings for libwdb (I'm still thinking of a way of how we can go lower than this)
we don't need to go lower -- what's there is probably perfectly adequate binding-wise. we can even change or add to the C side if it helped clean something up, but that gives the lowest level access without creating a cyclic dependency.
Sean (Jul 10 2018 at 04:21):But, um, can you give me some time to research and experiment a little bit? I don't want to drop ideas without testing out a few things :P
can you try implementing the "clean python" version of that script first?
Jaipal Singh (Jul 10 2018 at 04:21):But the advantage of using ctypesgen is that it generates bindings automatically, you don't have to manually write a lot of code (unlike something like "swig")
Sean (Jul 10 2018 at 04:22):I really don't think there's any time left to research and experiment. there's a month remaining and we're not really any closer than when you started to a feature a user could put into use yet...
Jaipal Singh (Jul 10 2018 at 04:22):Yea sure, just for clarity, what would a "clean python" version be?
Sean (Jul 10 2018 at 04:23):I would think it's the least complicated, most straightforward easy-to-understand python that lets you create geometry
Sean (Jul 10 2018 at 04:23):no "wdb" (wtf is wdb)
Jaipal Singh (Jul 10 2018 at 04:24):I've spent a major time of last month working on the script parser, so you can now use tcl procedural scripts to draw complicated objects. (genuinely hope it comes into use sometime) :P
Sean (Jul 10 2018 at 04:24):probably no files, but tbd... they really just want to create geometry
Jaipal Singh (Jul 10 2018 at 04:25):Wait, no wdb, then create geometry? How?
Sean (Jul 10 2018 at 04:25):I know you've spent at least a solid two weeks...
Sean (Jul 10 2018 at 04:25):that's why I interjected last week to try and get you to take a step back
Jaipal Singh (Jul 10 2018 at 04:26):Yea, also sorry I deviated from it! :/
Jaipal Singh (Jul 10 2018 at 04:26):I was moving ahead with the goal of having some way to properly implement procedural geometry
Sean (Jul 10 2018 at 04:26):I can already use tcl scripts to draw complicated objects .... I'm really having a hard time understanding the motivation
Sean (Jul 10 2018 at 04:28):most of mged and archer are implemented in tcl...
an "mged script" is a Tcl script
Sean (Jul 10 2018 at 04:28):that's not to say what you've done isn't potentially useful -- it could be
Sean (Jul 10 2018 at 04:29):for example, if we implement script objects, I can't just drop in an mged script for the same reason you couldn't just call mged -- it'd create a cyclic dependency
Sean (Jul 10 2018 at 04:30):so it's conceivable that something like your parser could be used to parse script objects in a variety of languages
Sean (Jul 10 2018 at 04:30):tcl, python, etc
Sean (Jul 10 2018 at 04:32):but that's why I said we really need to just focus on python -- we do not need added complexity or broad options. we need 1 that works, is exposed to users cleanly, lets us store procedural geometry scripts in .g files, works with all the expected functions like plotting and ray tracing, etc.
Sean (Jul 10 2018 at 04:33):I suppose from your devlog that you didn't make any progress trying to stub in a new 'script' primitive like we talked about?
Jaipal Singh (Jul 10 2018 at 06:50):but that's why I said we really need to just focus on python -- we do not need added complexity or broad options. we need 1 that works, is exposed to users cleanly, lets us store procedural geometry scripts in .g files, works with all the expected functions like plotting and ray tracing, etc.
Alright!
I suppose from your devlog that you didn't make any progress trying to stub in a new 'script' primitive like we talked about?
Yea, I've not worked on that yet.
Jaipal Singh (Jul 10 2018 at 09:50):@Sean : https://github.com/Killthebug/python-brlcad/blob/master/examples/sgi.py
Jaipal Singh (Jul 10 2018 at 09:51):I'm facing a little issue with the shader, everything else seems to be fine. Figuring it out now.
Sean (Jul 10 2018 at 15:43):Yea, I've not worked on that yet.
Okay, then it may make more sense to @Shubham Rathore to try or I can if it takes him more than a day. ;)
Sean (Jul 10 2018 at 15:57):@Sean : https://github.com/Killthebug/python-brlcad/blob/master/examples/sgi.py
well done @Jaipal Singh !
Sean (Jul 10 2018 at 15:58):so that's obviously a nearly direct translation of the shell script code .. but is pretty ugly python, don't you think?
Sean (Jul 10 2018 at 16:00):so ignoring how it was done in shell, fundamentally what is going on is we're making 18 spheres and 18 cylinders, grouping them all together, and creating a region ... what's the best way to do that?
Shubham Rathore (Jul 11 2018 at 04:13):I'll get back from the office and put a night into the primitive. It'll be done.
Sean (Jul 11 2018 at 04:48):thanks @Shubham Rathore ! if you just grep on one of the more recent primitives, it's usually pretty easy to find all the bits needed to add a new one (e.g., grep -i -r superell .) and let me know if you get stuck.
Sean (Jul 11 2018 at 04:48):feel free to commit incrementally, might be able to help things along
Jaipal Singh (Jul 11 2018 at 04:52):so ignoring how it was done in shell, fundamentally what is going on is we're making 18 spheres and 18 cylinders, grouping them all together, and creating a region ... what's the best way to do that?
Yea, not the cleanest of python scripts. What's a better way to do it ... hmmm..
Sean (Jul 11 2018 at 05:03):you could maybe map out the commands actually needed, and then see how you'd reduce it and eliminate the duplication
Jaipal Singh (Jul 11 2018 at 05:41):I'll try and make a condensed version. I think this script looks very "long" because there are multiple procedures involved in it (6 to be precise). Each procedure is almost the same (but that's only in this scenario, procedures will usually be fairly different I think). Also, the name of the primitives (rcc.100) created are "calculate/composed" when the procedure is called. This takes up a few lines in each procedure too.
I'm going to try and see how much I can reduce it to soon! :)
Jaipal Singh (Jul 11 2018 at 05:43):Would it be fine if I created a script with the goal of creating the same object as sgi.sh but didn't map it exactly to python (this script is a literally a translation into python)?
Sean (Jul 11 2018 at 05:52):don't feel limited to how the original one was implemented, what names were used, the order and manner it was constructed
Jaipal Singh (Jul 11 2018 at 05:53):Alright! Yea, the previous one was completely constraint by how the original was implemented.
Sean (Jul 11 2018 at 06:02):all that matters is making the same resulting shape (probably still with 18 spheres and cylinders)
Shubham Rathore (Jul 11 2018 at 07:22):@Sean Yess... did something similar last summer. Thanks for the help !
Jaipal Singh (Jul 11 2018 at 10:34):@Sean , tried to get it under 100 lines and also not look very complicated. https://github.com/Killthebug/python-brlcad/blob/master/examples/sgi.py
Sean (Jul 11 2018 at 15:06):@Jaipal Singh good reduction but not sure it helped. lets step it back more. I think we need something a lot simpler to start with. How would you make a simple cylinder with rounded edges?
Jaipal Singh (Jul 11 2018 at 15:09):Maybe draw a cylinder and two semi-spheres at it's ends. Or maybe define an object that is a cylinder with rounded edges and draw that multiple times?
Sean (Jul 11 2018 at 15:13):Not the whole end, just the edges.
Sean (Jul 11 2018 at 15:16):A sphere would round the entire end cap.
Sean (Jul 11 2018 at 15:18):Say I have a cylinder from 0,0,0 to 100,0,0 ... radius is, say, 25
Sean (Jul 11 2018 at 15:20):I want the top and bottom edges to be rounded to a radius of, say, 10
Sean (Jul 11 2018 at 15:21):How?
Sean (Jul 11 2018 at 16:33):@Jaipal Singh here’s an example: http://www.f-lohmueller.de/pov_tut/all_shapes/shapes340f.htm
Jaipal Singh (Jul 11 2018 at 17:12):Sorry was AFK.
Jaipal Singh (Jul 11 2018 at 17:12):So a cylinder with two Tori on the ends?
Sean (Jul 11 2018 at 17:13):try it
Jaipal Singh (Jul 11 2018 at 17:13):I'm trying to figure out the examples you've shared.
Sean (Jul 11 2018 at 17:17):in the examples, that's using a Round_Cylinder macro someone wrote for povray, which looks like it takes a point, a direction, a radius of the cylinder, a radius on the edge, and one more parameter we can ignore
Jaipal Singh (Jul 11 2018 at 17:18):In the rounded cylinder example, the cylinder in the middle is a merge(cylinder + 2 spheres) but the first one seems to be a merge (cylinder + 2 Tori)
Jaipal Singh (Jul 11 2018 at 17:19):One second, I'll check it out
Sean (Jul 11 2018 at 17:19):ignore the middle one and the merge parameter
it's a total distraction
Sean (Jul 11 2018 at 17:20):I suggest trying to create one by hand using brl-cad primitives and boolean operations (in mged/archer)
Sean (Jul 11 2018 at 17:21):once you figure it out, then write a python function that makes one for you (taking just a point, a direction, a cylinder radius, and an edge radius)
Jaipal Singh (Jul 11 2018 at 17:22):I suggest trying to create one by hand using brl-cad primitives and boolean operations (in mged/archer)
The first cylinder in the example you'd shared?
Sean (Jul 11 2018 at 17:22):sure
Sean (Jul 11 2018 at 17:23):you can use the values I mentioned, or duplicate the povray example, or whatever -- the point is figuring out how to construct it correctly (there are choices)
Jaipal Singh (Jul 11 2018 at 17:23):How did you deduce that it's a point and a direction? They're both 3 dimensional vectors
Sean (Jul 11 2018 at 17:25):because that's all that makes sense in this context ;)
Jaipal Singh (Jul 11 2018 at 17:25):http://www.f-lohmueller.de/pov_tut/all_shapes/shapes3_80e.htm : Here it says A and B are two points :P
Sean (Jul 11 2018 at 17:26):it's modeled at the origin, so the second parameter is equivalent
Jaipal Singh (Jul 11 2018 at 17:26):Makes sense! :thumbs_up:
Sean (Jul 11 2018 at 17:26):the first can only be a point because 0,0,0 is degenerate
Sean (Jul 11 2018 at 17:28):either way, make it a point+vector or point+point, that part doesn't matter -- pick either but after you figure out a construction method by hand :)
Jaipal Singh (Jul 11 2018 at 17:28):Yup! On it.
Sean (Jul 11 2018 at 17:29):I think this will lay better ground work for what we're trying to get at
Jaipal Singh (Jul 11 2018 at 17:31):Aye! I think I'm starting to understand the importance of procedural geometry now. (Hopefully the right thing)
Sean (Jul 11 2018 at 17:32):@Sean Yess... did something similar last summer. Thanks for the help !
@Shubham Rathore any progress yet?
Shubham Rathore (Jul 11 2018 at 17:43):@Sean just back from the office . About to start working.
Jaipal Singh (Jul 11 2018 at 17:47):Just wanted to know, is it possible to round the edges of an arb8? (This old email is the only clue I had https://sourceforge.net/p/brlcad/discussion/362510/thread/2cdbd563/)
Shubham Rathore (Jul 12 2018 at 03:57):@Jaipal Singh I'm unaware of any such thing. Never ever tried rounding those edges. @Sean might give you a better insight.
P.S. - if not a feature, try implementing it through procedural geometry ;)
Sean (Jul 12 2018 at 04:40):@Jaipal Singh of course it’s possible. There are very few shapes that are not possible without resorting to explicit surfaces. Rounded edges on a box is a basic intro shape, only slightly more complicated than rounded rcc. Speaking of which, did you figure it out?
Sean (Jul 12 2018 at 04:42):If you need a hint, rounded rcc requires at least 4 shapes.
Shubham Rathore (Jul 12 2018 at 04:47):@Sean What should the " rt_script_internal "look like. What params should it comprise other that the uint32_t magic ?
Sean (Jul 12 2018 at 05:03):@Shubham Rathore we don’t know that yet so can just stub it empty.
Sean (Jul 12 2018 at 05:07):I’d expect it’ll probably need a text buffer, something describing what kind of script it is, and maybe instantiation info
Shubham Rathore (Jul 12 2018 at 05:08):Was thinking the same... Probably a bu_vls and a magic should be a part currently
Sean (Jul 12 2018 at 05:10):Sounds like a start
Shubham Rathore (Jul 12 2018 at 06:31):@Jaipal Singh What are you working on currently ?
Jaipal Singh (Jul 12 2018 at 06:34):@Shubham Rathore , I just started work on modeling a round edged cylinder. Will then implement it in python once I know how to construct it.
Shubham Rathore (Jul 12 2018 at 07:07):The primitive will be ready soon, the part to link your code to the primitive stands next. You should start thinking of ways to encounter all the issues that we might have in the coming days.
Jaipal Singh (Jul 12 2018 at 07:10):Sure
Jaipal Singh (Jul 12 2018 at 11:34):Screen-Shot-2018-07-12-at-5.02.07-PM.png
Getting there.
Sean (Jul 12 2018 at 17:01):@Shubham Rathore , I just started work on modeling a round edged cylinder. Will then implement it in python once I know how to construct it.
Figure it out yet?
Sean (Jul 12 2018 at 18:54):@Jaipal Singh ? this can't be an all-day activity. it was supposed to be something quick that just takes an hour at most. there needs to be more communication, more coding, please.
Sean (Jul 12 2018 at 18:54):if you can't figure out the caps, that's okay, say something! it's not a test.
Jaipal Singh (Jul 12 2018 at 19:04):Spent a ton of time, just not able to figure it out!
Jaipal Singh (Jul 12 2018 at 19:07):I drew an rcc (rcc1) with two tori(t1, t2) on the ends, then tried to subtract another rcc(rcc2) of the same height but lesser radius from the tori. After this fill the space with rcc2 and render. On paper it makes sense, but I just can't visualize it
Sean (Jul 12 2018 at 19:10):so @Jaipal Singh you have the right idea with torii on the ends. you just need to fill the divets they create (with another cylinder).
Sean (Jul 12 2018 at 19:10):a rounding is about tangency, so you need something tangent with the side and top of the torus
Sean (Jul 12 2018 at 19:28):which method do you think would be easier to implement in code?
Jaipal Singh (Jul 12 2018 at 19:28):Method 1 seems the simplest
Jaipal Singh (Jul 12 2018 at 19:29):The only motivation I see behind method3 is that it could allow the user to have two different types of rounded edges (different on top and bottom)?
Sean (Jul 12 2018 at 19:30):that's a great point regarding the differences!
Sean (Jul 12 2018 at 19:31):however, that would also imply not using canonical torus
Jaipal Singh (Jul 12 2018 at 19:31):canonical as in isometric?
Sean (Jul 12 2018 at 19:32):but that's a great point, one could use an elliptical torus to taper sides differently or arb8's to create hexagonal pattern or ...
Jaipal Singh (Jul 12 2018 at 19:35):The only issue with the arb8 is that it might overflow right. then we'll have edges that are beyond the tori
Sean (Jul 12 2018 at 19:35):canonical == ring torus with a circular cross-section
Sean (Jul 12 2018 at 19:36):if you unioned them, sure, but that wouldn't be the way to go -- you'd want to subtract them
Sean (Jul 12 2018 at 19:36):chamfers instead of rounds
Jaipal Singh (Jul 12 2018 at 19:38):We'd have to have restrictions on the relationship between the tube radius and the width of the arb8 as well...
Sean (Jul 12 2018 at 19:39):so for the script, just keep it as simple as possible -- the parameters we talked about yesterday were perfect: cylinder start/end points, cylinder radius, and rounding radius
Sean (Jul 12 2018 at 19:40):see if you can implement a function that takes those four and produces a method1 result
Jaipal Singh (Jul 12 2018 at 19:40):Ok sure
Sean (Jul 12 2018 at 19:41):note my comment about needing to reduce the principal cylinder's length by the rounding radius on both ends
Sean (Jul 12 2018 at 19:44):so if you specify a cylinder 100 long (e.g., from 0,0,0 to 0,0,100) with rounding 10, you'll create a cylinder 80 long (because you have to +10 rounding on each end), and then create your interior cylinder 100 long, but with a radius reduced by the rounding too.
Sean (Jul 12 2018 at 19:44):make sense?
Jaipal Singh (Jul 12 2018 at 19:45):Yea does!
Jaipal Singh (Jul 12 2018 at 19:46):So, only the z co-ordinates would be impacted right
Jaipal Singh (Jul 12 2018 at 19:48):No no!
Sean (Jul 12 2018 at 19:49):heh, not at all if the user is allowed to specify two random points
Jaipal Singh (Jul 12 2018 at 19:50):Yes, I'm just thinking my way around it (i'm poor at 3d visualization)
Sean (Jul 12 2018 at 19:51):since the point of this isn't figuring out math, you can keep it simple and now support two points -- you could support a cylinder length instead and just model it at the origin, up the Z axis
Sean (Jul 12 2018 at 19:51):if you want to figure it out, I can talk you through what's needed
Sean (Jul 12 2018 at 19:51):I suggest keeping it simple
Sean (Jul 12 2018 at 19:52):can always come back and make it better :)
Jaipal Singh (Jul 12 2018 at 19:52):since the point of this isn't figuring out math, you can keep it simple and now support two points -- you could support a cylinder length instead and just model it at the origin, up the Z axis
I'll do this for now, but I'd really appreciate if you could talk me through the math too :P (I want it to be able to handle everything)
Jaipal Singh (Jul 12 2018 at 19:52):Yea!
Jaipal Singh (Jul 12 2018 at 19:52):I'll give the math a thought while I'm creating the simple version, will maybe hit you up later if I don't get it
Sean (Jul 12 2018 at 19:53):so a python function then will take a cylinder length, cylinder radius, and rounding radius
Jaipal Singh (Jul 12 2018 at 19:54):Can we keep it to be two vertices, cylinder radius and rounding radius? We'll just pass it arguments such that only the z-coordinate changes
Jaipal Singh (Jul 12 2018 at 19:54):That way if we change the math later, we won't have to change a lot of the code
Sean (Jul 12 2018 at 19:54):that way you can follow exactly what's in method1 in the sample geometry, which essentially is the result of "function(1100, 250, 50)" iirc
Sean (Jul 12 2018 at 19:55):I'd keep it simple arguments that are valid, not plan for future but be wrong in the meantime.. there's not going to be a lot of code to change
Jaipal Singh (Jul 12 2018 at 19:56):Ok sure
Sean (Jul 12 2018 at 19:56):this whole thing is probably going to be < 40 lines with validation and bounds checking
Jaipal Singh (Jul 12 2018 at 19:56):Yea. Sean, why do you use "no tree"?
Sean (Jul 12 2018 at 19:56):the geometry is made in six lines
Sean (Jul 12 2018 at 19:57):what do you mean "no tree"?
Jaipal Singh (Jul 12 2018 at 19:59):... comb region no tree {u {u ... . Is it because you don't want it to print the tree structure?
Sean (Jul 12 2018 at 20:00):heh
you're reading the .asc file ... just because it's text doesn't mean you should be reading it like it's text ;)
Sean (Jul 12 2018 at 20:01):that's the serialization format of a comb in .asc format
Sean (Jul 12 2018 at 20:01):those are key value pairings in Tcl list form
Jaipal Singh (Jul 12 2018 at 20:01):Oooooh!
Sean (Jul 12 2018 at 20:02):region=no
Sean (Jul 12 2018 at 20:02):tree="{u {u ..."
Jaipal Singh (Jul 12 2018 at 20:02):Makes a lot more sense :|
Sean (Jul 12 2018 at 20:02):if you compile, you'll have a rounds.g in your share/db folder, or directly run asc2g rounds.asc rounds.g
Sean (Jul 12 2018 at 20:02):then open that .g in archer or mged
Jaipal Singh (Jul 12 2018 at 20:03):Alright
Sean (Jul 12 2018 at 20:04):that's a v5 asc file .. if I had uploaded it as a v4 asc, you probably wouldn't have tried to read it ;)
Sean (Jul 12 2018 at 20:04):if you look at the other .asc files, you'll notice how they're different
Jaipal Singh (Jul 12 2018 at 20:07):Haha! Just saw some of those, I don't think I can really spend time deducing them. :sweat:
Sean (Jul 12 2018 at 20:58):basically, you're not supposed to read them ;)
Jaipal Singh (Jul 12 2018 at 21:28):I'm a little stuck with respect to the geometry. For the end of the cylinder away from the vertex, I'm able to achieve a perfect rounded edge.
Jaipal Singh (Jul 12 2018 at 21:29):On the end of the cylinder near the origin, I make the outer cylinder with : in outer rcc 0 0 10 0 0 90 40 and the filler cylinder with : in filler rcc 0 0 0 0 0 100 30
Jaipal Singh (Jul 12 2018 at 21:30):r/sphere/cylinder
Jaipal Singh (Jul 12 2018 at 21:31):Despite having different inital vertices both the cylinders seem to have the same start point. But if I change to : in outer rcc 0 0 0 10 0 0 80 40, I can see the outer cylinder moving away from the origin
Sean (Jul 12 2018 at 21:32):your filler cylinder shouldn't have a radius larger than your outer...
Jaipal Singh (Jul 12 2018 at 21:33):Filler had a radius of 30 and the outer of 40, that's not an issue right
Sean (Jul 12 2018 at 21:33):if you create them in mged, you can highlight them with "sed outer", followed by "reject"
Sean (Jul 12 2018 at 21:34):I think you may be misinterpreting the wireframe
Jaipal Singh (Jul 12 2018 at 21:34):Could be, I'll figure it out.
Sean (Jul 12 2018 at 21:35):it can be confusing to read them right if you never have before. turning on "depth queuing" in the Misc menu may help a little
Jaipal Singh (Jul 12 2018 at 21:51):Done!
@Sean : https://github.com/Killthebug/python-brlcad/blob/master/examples/rounded_rcc.py
Sean (Jul 13 2018 at 16:06):excellent! let's see some examples...
Sean (Jul 13 2018 at 16:07):@Shubham Rathore excellent too, can you commit what you have? I can probably finish is up today.
Jaipal Singh (Jul 13 2018 at 19:25):@Sean , what's the plan for the next month then?
Sean (Jul 13 2018 at 19:37):@Jaipal Singh you tell me first :)
Sean (Jul 13 2018 at 19:37):and where are those examples? :)
Jaipal Singh (Jul 13 2018 at 19:39):Just saw the chat, will make em soon and share. :)
Jaipal Singh (Jul 13 2018 at 19:40):Also, should I include some sort of validation/check in the script to prevent possibly incorrect values being passed
Sean (Jul 13 2018 at 19:40):sure, sanity checks for what actually works is always a good idea
Jaipal Singh (Jul 13 2018 at 19:41):I think the final part of the project would be to pair python-brlcad with the script primitive
Sean (Jul 13 2018 at 19:42):can't have a bend radius greater than the cylinder radius, can't be more than half the cylinder length, ...
Jaipal Singh (Jul 13 2018 at 19:42):Sure! I'll think of anything else if that might be an issue
Sean (Jul 13 2018 at 19:42):I think the final part of the project would be to pair python-brlcad with the script primitive
I'm hoping we can get further than this, but I fully agree with this as a goal as it puts python into user hands
Sean (Jul 13 2018 at 19:43):one thing we need to figure out quickly is how this all gets used
Jaipal Singh (Jul 13 2018 at 19:43):True, that's what I wanted to know. Once we couple with with the script primitive what happens?? What's the workflow like? When is it used?
Sean (Jul 13 2018 at 19:43):what the actual mechanism looks like and how it's hooked up via script objects, whether we need separate objects to instantiate them or not
Jaipal Singh (Jul 13 2018 at 19:44):I'm hoping we can get further than this, but I fully agree with this as a goal as it puts python into user hands
I think the project will have a lot more thing to be worked on, I was just setting the month's target as perfecting the pairing between them
Sean (Jul 13 2018 at 19:45):so I think we need to think about end-user design
Sean (Jul 13 2018 at 19:46):what we want that to look like since that will drive some of the implementation decisions
Jaipal Singh (Jul 13 2018 at 19:46):Uh-huh.. So let's say I want to construct a lot of rounded cylinders
Jaipal Singh (Jul 13 2018 at 19:46):And let's also assume, there isn't a script for rounded cylinders.
Sean (Jul 13 2018 at 19:46):k
Jaipal Singh (Jul 13 2018 at 19:46):So I write a script in python (like the one I had shared above)
Sean (Jul 13 2018 at 19:47):first question is whether that script can go into the .g or if it MUST live in an outside file
Sean (Jul 13 2018 at 19:48):I think it should be able to go into the .g since we do want a .g to fully function if someone wants it to be portable by itself, but not limited to being in the .g
Sean (Jul 13 2018 at 19:49):which would be similar to how we handle image data like textures, they can be embedded or in external files
Jaipal Singh (Jul 13 2018 at 19:50):Yea, first off I don't know how it can be inside a .g file and secondly I don't think it should be because it locks up a lot of flexibility
Sean (Jul 13 2018 at 19:51):you can put anything into a .g, it's just a collection of objects -- you can stash a microsoft word document in there, for example
Jaipal Singh (Jul 13 2018 at 19:51):Uh-huh
Jaipal Singh (Jul 13 2018 at 19:51):Woah, didn't know that (too)!
Jaipal Singh (Jul 13 2018 at 19:52):Once I make the script, I can pass it to the script primitive and some params with it to construct my objects in mged or archer..
Sean (Jul 13 2018 at 19:52):and you can pull them out, use external tools on them, etc
Sean (Jul 13 2018 at 19:52):right, so script exists (somewhere)
Jaipal Singh (Jul 13 2018 at 19:53):Woah, is .g a universal format or something that was made by brlcad?
Jaipal Singh (Jul 13 2018 at 19:53):Yea, somewhere (we've got to figure out how paths are decided then)
Sean (Jul 13 2018 at 19:54):Woah, is .g a universal format or something that was made by brlcad?
yes ;)
Jaipal Singh (Jul 13 2018 at 19:54):Or we could have a predecided directory, but that would be a pain.
Jaipal Singh (Jul 13 2018 at 19:54):Damn, sweet!
Sean (Jul 13 2018 at 19:54):it was designed, has a formal specification, we've talked about maybe submitting it to ISO, but it's pretty flexible as-is
Sean (Jul 13 2018 at 19:55):we could even require that the script be embedded and if a user wants it to be external, they'd #import or whatever in the embedded source
Sean (Jul 13 2018 at 19:56):so options there, I don't see any problems
Jaipal Singh (Jul 13 2018 at 19:56):Amazing. (Just out of curiosity regarding the .g files, is there some documentation about how it was made?)
Sean (Jul 13 2018 at 19:56):so the script exists, and we need some way to instantiate what it creates
Jaipal Singh (Jul 13 2018 at 19:57):Be embeded?
Sean (Jul 13 2018 at 19:57):there's an old out-of-date spec file here https://brlcad.org/w/images/2/2d/BRL_CAD_g_format_V5.pdf
Sean (Jul 13 2018 at 19:58):that's a conversion from the real source with lots of sections removed, that was some 15 years ago
Jaipal Singh (Jul 13 2018 at 20:00):Oh embeded as in, in the script primitive? So we can have some important procedures be embeded?
Sean (Jul 13 2018 at 20:00):the documentation on how it was made is in old mailing list archives, not easy to get to (might be some details in the source repo doc dir)
Sean (Jul 13 2018 at 20:02):embedded as in inside the .g file, we could require that there be at least a snippet of code to import and call a function \
Jaipal Singh (Jul 13 2018 at 20:02):Great, I'll try and gather all that I can! (It's very sad so much documentation get's lost over the years, but again, can't do anything about that)
Sean (Jul 13 2018 at 20:02):the only reason for doing that would be to simplify the usage -- wouldn't have to have options for specifying whether the script is in an object or in a file
Jaipal Singh (Jul 13 2018 at 20:02):Alright alright
Sean (Jul 13 2018 at 20:03):that wasn't in response to gathering docs :)
Jaipal Singh (Jul 13 2018 at 20:03):Haha yea! :P
Sean (Jul 13 2018 at 20:03):the only reason for requiring a script be in the .g would be to simply usage
Sean (Jul 13 2018 at 20:04):you're welcome to gather docs or not -- I'm not sure what you'd do with them, and you definitely don't have access to all of them :)
Sean (Jul 13 2018 at 20:04):definitely not helpful for your project, but spelunking is fun :)
Jaipal Singh (Jul 13 2018 at 20:05):the only reason for requiring a script be in the .g would be to simply usage
True. I hadn't thought of it like this. I always assumed that we'd have a directory where we'd have a bunch of scripts that could be used.
Sean (Jul 13 2018 at 20:05):so ...
Sean (Jul 13 2018 at 20:05):well that's a good point
Sean (Jul 13 2018 at 20:05):we'll likely end up shipping a set of standard scripts that are available to users, and those definitely won't be embedded in the .g
Jaipal Singh (Jul 13 2018 at 20:06):And that way people could modify any of the scripts depending on how they might want it to
Sean (Jul 13 2018 at 20:06):do you know how python's import mechanism works?
Sean (Jul 13 2018 at 20:06):I mean how it searches, how that's controlled?
Jaipal Singh (Jul 13 2018 at 20:07):Not very specifically, but I have vague ideas. I could learn by tomorrow
Sean (Jul 13 2018 at 20:07):don't worry about it then, just wondering
Sean (Jul 13 2018 at 20:07):back to instantiation...
Jaipal Singh (Jul 13 2018 at 20:07):I know it can import modules and objects and usually looks up in a standard directory (where all packages are present) or the same directory if you a __init__.py in it.
Jaipal Singh (Jul 13 2018 at 20:08):Yea..
Jaipal Singh (Jul 13 2018 at 20:08):*you have a
Sean (Jul 13 2018 at 20:09):so the 'in' command is the canonical entry point for creating objects by hand
Sean (Jul 13 2018 at 20:09):in mysphere sph ...
Jaipal Singh (Jul 13 2018 at 20:09):Uh-huh
Jaipal Singh (Jul 13 2018 at 20:09):Yea
Sean (Jul 13 2018 at 20:10):so first thought that comes to my mind is whether we want to propagate scripts up to being first-class primitives or not
Sean (Jul 13 2018 at 20:10):in myroundedrcc rounded_rcc ...
Sean (Jul 13 2018 at 20:10):probably not that low, but it's an interesting thought
Jaipal Singh (Jul 13 2018 at 20:10):Ohh! I don't think we should do something like that
Jaipal Singh (Jul 13 2018 at 20:11):But then that way if I introduce a new script, I'll have to recompile the code so that it reflects in the in command
Sean (Jul 13 2018 at 20:11):next would be whether instantiations are separate from the scripts themselves if they're embedded
Jaipal Singh (Jul 13 2018 at 20:11):How about something life : in myroundedrcc script rounded_rcc
Sean (Jul 13 2018 at 20:13):yes, that's more likely if we don't propagate to first-class
Sean (Jul 13 2018 at 20:13):good
Sean (Jul 13 2018 at 20:13):so there's specification of the script source and specification of the instantiation
Jaipal Singh (Jul 13 2018 at 20:13):Yea..
Sean (Jul 13 2018 at 20:14):they could be one in the same, but I'm not sure they should
Jaipal Singh (Jul 13 2018 at 20:14):Now that I think of it, it might be different..
Sean (Jul 13 2018 at 20:14):for example, both together might be something like: in myrcc script FILE FUNCTION ARGS
Jaipal Singh (Jul 13 2018 at 20:15):Alright..
Sean (Jul 13 2018 at 20:15):in myrcc script /path/to/rounded_rcc.py rounded_rcc 100 50 10
Jaipal Singh (Jul 13 2018 at 20:15):Perfect
Sean (Jul 13 2018 at 20:16):with that, the script lives in a file, we specify the entry point, and arguments to pass -- does that work for all argument types a script might want to pass?
Sean (Jul 13 2018 at 20:16):I think so ...
Jaipal Singh (Jul 13 2018 at 20:17):Uhm, so I don't know about things sketch.. but for the primitives that we're dealing with for now, I think it should all go through
Sean (Jul 13 2018 at 20:18):well basically it's asking if there's anything we might need to pass to a python function
Sean (Jul 13 2018 at 20:18):that isn't easiliy described as an int/float/string
Jaipal Singh (Jul 13 2018 at 20:18):Nothing other than those would be required
Sean (Jul 13 2018 at 20:20):so let's call that "option 1" .. script objects that specify source and instantiation
Sean (Jul 13 2018 at 20:20):other options?
Jaipal Singh (Jul 13 2018 at 20:24):_thinking_
Jaipal Singh (Jul 13 2018 at 20:25):I've always had only the option1 in my mind when I've thought of including the scripts..
Jaipal Singh (Jul 13 2018 at 20:25):How about re-using instantiations?
Jaipal Singh (Jul 13 2018 at 20:26):No, that's just complicating some simple things..
Sean (Jul 13 2018 at 20:27):think about what you originally said -- say you have a bunch of instantiations
Sean (Jul 13 2018 at 20:27):say 100
Sean (Jul 13 2018 at 20:29):easy enough to create them all. now we decide to change the name of our function from rounded_rcc to rounded_cylinder ...
Jaipal Singh (Jul 13 2018 at 20:29):Uh-huh..
Sean (Jul 13 2018 at 20:29):wham, 100 broken objects .. what can we do about that
Jaipal Singh (Jul 13 2018 at 20:30):Taking a step back, what are instantiations here?
Jaipal Singh (Jul 13 2018 at 20:33):Even if you change the function name, you need to re-compile all the code
Jaipal Singh (Jul 13 2018 at 20:34):Now, talking about the instances we had created earlier, they're independent of the function_name right cause what matters to us is the object name (myrcc) which contains commands on how to build the object we want
Sean (Jul 13 2018 at 20:34):instantiations are separate instances of that script being run
Sean (Jul 13 2018 at 20:34):so in that first example, it's a rounded cylinder length 100, radius 50, rounding 10
Jaipal Singh (Jul 13 2018 at 20:34):OK
Sean (Jul 13 2018 at 20:35):that's 1 instantiation, an object in the database with name 'myrcc'
Sean (Jul 13 2018 at 20:35):I can copy that, rename it, delete it, scale it, translate it, etc
Jaipal Singh (Jul 13 2018 at 20:35):Yea..
Sean (Jul 13 2018 at 20:35):so instantiation == geometry
Sean (Jul 13 2018 at 20:36):geometry resulting from running the script
Jaipal Singh (Jul 13 2018 at 20:36):Alright, yea. So the instantiation is independent of the function name in the script right?
Sean (Jul 13 2018 at 20:36):in "option 1" yes ...
Sean (Jul 13 2018 at 20:37):I suppose we could run the script right then and there, and be done with it, so renaming the function wouldn't break the objects
Jaipal Singh (Jul 13 2018 at 20:37):Yea.. So this way even changing the function name doesn't become an issue
Sean (Jul 13 2018 at 20:37):let's call that "option 2"
Sean (Jul 13 2018 at 20:38):downside being say instead of changing the function name, I change the implementation ... now I have to re-run all my instantiations
Jaipal Singh (Jul 13 2018 at 20:38):Yea, that's almost like creating a new script in it's own.
Sean (Jul 13 2018 at 20:39):basically, it's not much different than just running the script, importing the result .. no association
Sean (Jul 13 2018 at 20:40):I was thinking something more dynamic that stays in sync and gets evaluated on-demand
Jaipal Singh (Jul 13 2018 at 20:40):Hmm.. So we track changes in the script and then modify all the instances of a script before it was changed?
Sean (Jul 13 2018 at 20:41):I don't need script objects if all i'm going to do is run "exec ./rounded_rcc 100 50 10 ; dbconcat file.g"
Sean (Jul 13 2018 at 20:42):I don't think we could reliably track changes in the script
Sean (Jul 13 2018 at 20:42):but we shouldn't need to
Sean (Jul 13 2018 at 20:42):if it's on-demand, the script can be evaluated as needed
Jaipal Singh (Jul 13 2018 at 20:42):I suppose we could run the script right then and there, and be done with it, so renaming the function wouldn't break the objects
Like you said
Sean (Jul 13 2018 at 20:42):hey, so I see the time for you -- is this something to pick up later?
Sean (Jul 13 2018 at 20:43):or keep going, i'm fine either way :)
Jaipal Singh (Jul 13 2018 at 20:43):I've got 10 minutes before it's bedtime so I'd like to continue a little more :P
Sean (Jul 13 2018 at 20:44):okay
Jaipal Singh (Jul 13 2018 at 20:44):I mean, if that's okay with you?
Sean (Jul 13 2018 at 20:44):like I said, I'm fine either way :)
Jaipal Singh (Jul 13 2018 at 20:44):I like the option of running the script then and there and finishing it off
Sean (Jul 13 2018 at 20:45):my point earlier is that if we just run the script and are done with it -- that's not a very helpful feature
Sean (Jul 13 2018 at 20:45):users could do that now just by running the script outside of mged for example
Sean (Jul 13 2018 at 20:45):there's not really much added value
Sean (Jul 13 2018 at 20:45):we save them 1 step, importing what they created (dbconcat)
Jaipal Singh (Jul 13 2018 at 20:46):Alright..
Sean (Jul 13 2018 at 20:46):that's where my original thinking is with scripts that live in the database, I want to edit code on the fly and see geometry change :)
Sean (Jul 13 2018 at 20:47):like what if while you were writing your rounded_rcc.py script, you could see the geometry pop up in the window, add one more line to create a torus and you see it
Sean (Jul 13 2018 at 20:47):instead of re-running, opening the geometry file, running ls, drawing the geometry object, etc
Sean (Jul 13 2018 at 20:48):edit, see it
Sean (Jul 13 2018 at 20:48):I think that's what we want
Jaipal Singh (Jul 13 2018 at 20:48):Ahhh!
Jaipal Singh (Jul 13 2018 at 20:48):I get it
Jaipal Singh (Jul 13 2018 at 20:48):So, are .g files that are being rendered refreshed constantly?
Sean (Jul 13 2018 at 20:48):Yes!
Sean (Jul 13 2018 at 20:49):or at least we could make them do that
Jaipal Singh (Jul 13 2018 at 20:49):Again, something I just learnt!
Sean (Jul 13 2018 at 20:49):everything goes through the primitive callbacks, so in the case of a script object we can choose to re-evaluate the script every time or only once
Sean (Jul 13 2018 at 20:50):I think it should re-evaluate every time for starters
Jaipal Singh (Jul 13 2018 at 20:50):Alright, alright.. So that'd be something we account for when we're creating the script primitive?
Sean (Jul 13 2018 at 20:50):right, so some basic background there -- every primitive defines a set of functions
Jaipal Singh (Jul 13 2018 at 20:51):Ok..
Sean (Jul 13 2018 at 20:51):one of those functions for example, plot(), creates the wireframe you see when you run "draw someobject"
Sean (Jul 13 2018 at 20:52):when you run "rt someobject", it happens to call prep() once and shot() lots of times to render an image
Sean (Jul 13 2018 at 20:52):when you run "l someobject", it calls describe()
Jaipal Singh (Jul 13 2018 at 20:52):Alright
Sean (Jul 13 2018 at 20:53):so we define all of those functions, and would for a script object
Jaipal Singh (Jul 13 2018 at 20:53):Are these defined in each primitive or are they defined somewhere else and then called for a specific primitive?
Sean (Jul 13 2018 at 20:53):so the basic idea is that we'd call the python interp before most if not all of those functions to get the geometry
Jaipal Singh (Jul 13 2018 at 20:53):Yea
Sean (Jul 13 2018 at 20:53):Are these defined in each primitive or are they defined somewhere else and then called for a specific primitive?
no comprendo .. don't understand the question
Jaipal Singh (Jul 13 2018 at 20:55):So let's say you have : describe(). It serves the same purpose for each primitive. Now, do you define a describe() for each primitive (in each file for a primitive) or is it a single function places somewhere that is called and the primitive is passed to it.
Sean (Jul 13 2018 at 20:56):serves the same purpose, but is different for every one of them, so yeah each primitive defines itself
Sean (Jul 13 2018 at 20:56):every primitive defines this set of callbacks (you can browse them in src/librt/primitives ... 1 subdir per primitive) -- this is what @Shubham Rathore is stubbing for us empty
Jaipal Singh (Jul 13 2018 at 20:56):Yea, I've gone through some of those to get a basic idea about them
Sean (Jul 13 2018 at 20:56):it's not describe(), it's actually rt_ell_describe() and rt_arb8_describe() and ... we'll have an rt_script_describe()
Jaipal Singh (Jul 13 2018 at 20:57):Ahh! Ok.
Sean (Jul 13 2018 at 20:57):there is a general rt_obj_describe where you pass the object, but it automatically calls the rt_*_describe() for that object
Sean (Jul 13 2018 at 20:57):it's all implemented in a form of object-oriented C
Jaipal Singh (Jul 13 2018 at 20:58):Sweet
Sean (Jul 13 2018 at 20:58):so we have options 1 and 2 which are "in obj script file func params..." and then only difference being whether we evaluate file every time or once and forget the file
Jaipal Singh (Jul 13 2018 at 20:59):Hmm yea. Now again, I think it's best if you decide this, cause you'd know what the users would prefer
Jaipal Singh (Jul 13 2018 at 21:01):Can we continue this tomorrow? :) It's 2:30AM and I'm going to spoil my weekend if I sleep into tomorrow afternoon!
Sean (Jul 13 2018 at 21:01):there's an option 3 where we separate the script specification from instantiation, so that might be something like: in rrcc script /path/to/rounded_rcc.py rounded_rcc .... and another command like "script create myrcc rrcc 100 50 10"
Jaipal Singh (Jul 13 2018 at 21:02):Oooh! So if you change the script, you recompile and it then changes the instantiation?
Sean (Jul 13 2018 at 21:02):option 4: or even two separate primitives, one for script, another for instance: in myrcc instance rounded_rcc 100 50 10
Jaipal Singh (Jul 13 2018 at 21:02):But this won't do the real-time rendering thing (option3)
Jaipal Singh (Jul 13 2018 at 21:03):I think option 4 could be reached once we implement option 2 followed by option 3
Sean (Jul 13 2018 at 21:03):Oooh! So if you change the script, you recompile and it then changes the instantiation?
right, there'd be some way to propagate updates or re-evaluate, maybe "script update myrcc"
Jaipal Singh (Jul 13 2018 at 21:04):Yea
Sean (Jul 13 2018 at 21:04):we can do the on-demand evaluation with all but option 2 since that's specifically what it's about
Sean (Jul 13 2018 at 21:04):so I think 2's out
Jaipal Singh (Jul 13 2018 at 21:04):Alright
Sean (Jul 13 2018 at 21:04):okay
Sean (Jul 13 2018 at 21:04):so let's think on it some more
Sean (Jul 13 2018 at 21:05):you working this weekend or not?
Jaipal Singh (Jul 13 2018 at 21:05):Second half of tomorrow and first half of Sunday, yes!
Jaipal Singh (Jul 13 2018 at 21:05):World Cup Finals on Sunday night :D
Sean (Jul 13 2018 at 21:05):okay, so maybe we can continue the discussion tomorrow
Sean (Jul 13 2018 at 21:05):right!
Jaipal Singh (Jul 13 2018 at 21:05):Sure!
Sean (Jul 13 2018 at 21:06):fun
Sean (Jul 13 2018 at 21:06):alright, well good talking -- sleep on it some ;)
Jaipal Singh (Jul 13 2018 at 21:06):Was that sarcastic? :P
Sean (Jul 13 2018 at 21:06):nope :)
Jaipal Singh (Jul 13 2018 at 21:06):Indeed! :D
Jaipal Singh (Jul 13 2018 at 21:06):Sorry! Night!
Sean (Jul 13 2018 at 21:06):I do that often .. think about something late into the night, then sleep on it so I can realize how awful those ideas were ;)
Sean (Jul 13 2018 at 21:07):night!
Jaipal Singh (Jul 13 2018 at 21:08):I do that often .. think about something late into the night, then sleep on it so I can realize how awful those ideas were ;)
hahaha! :P
Shubham Rathore (Jul 14 2018 at 05:25):I do that often .. think about something late into the night, then sleep on it so I can realize how awful those ideas were ;)
hahaha! :P
Actually helps! Specially when you need to devise an algorithm, sleep with the problems and you have a solution or maybe a rough idea the next morning. :)
Shubham Rathore (Jul 14 2018 at 15:33):@Sean the patch for the primitive is uploaded, Do let me know if you find an error : https://sourceforge.net/p/brlcad/patches/500/ . @Jaipal Singh Have a look !
Jaipal Singh (Jul 14 2018 at 15:39):Sure!
Jaipal Singh (Jul 14 2018 at 15:39):Thanks! :)
Sean (Jul 14 2018 at 15:51):@Shubham Rathore that's outstanding work!
Sean (Jul 14 2018 at 15:52):looks pretty good to me, only thing I notice is that you have a couple unrelated files in there modified
Sean (Jul 14 2018 at 15:53):annot/annot.c and src/libdm/dm-X.c have changes that don't look related
Sean (Jul 14 2018 at 15:53):other than that, looks good to commit!
Sean (Jul 14 2018 at 15:54):for what it's worth (it's a minor point, not a problem) -- you don't need to add v4 support for new primitives (db4.h)
Shubham Rathore (Jul 14 2018 at 16:03):Yesss.... i know db4 is of no use now, just added them :).
Shubham Rathore (Jul 14 2018 at 16:03):Yes, there were some indentation issues with annot.c that I rectified
Sean (Jul 14 2018 at 16:38):Yes, there were some indentation issues with annot.c that I rectified
that's fine, they should just be in a separate commit, and the src/libdm/dm-X.c change looked like commented/uncommented code
Shubham Rathore (Jul 14 2018 at 16:40):Ah ! sorry for that
Shubham Rathore (Jul 15 2018 at 06:27):@Sean I've uploaded a revised patch (https://sourceforge.net/p/brlcad/patches/500/) in the comments. Have a look !
Jaipal Singh (Jul 16 2018 at 09:21):Some examples of rounded_rccs Screen-Shot-2018-07-16-at-2.50.02-PM.png
Jaipal Singh (Jul 16 2018 at 09:52):Added basic sanity checks as well.
Sean (Jul 16 2018 at 14:12):@Sean I've uploaded a revised patch (https://sourceforge.net/p/brlcad/patches/500/) in the comments. Have a look !
this looks great, please commit it
Sean (Jul 16 2018 at 14:37):@Jaipal Singh so what's next?
Sean (Jul 16 2018 at 14:47):we talked about 4 options for the user side, eliminated one of the options, and we still have to sort out how to bridge from object to script
Jaipal Singh (Jul 16 2018 at 16:01):there's an option 3 where we separate the script specification from instantiation, so that might be something like: in rrcc script /path/to/rounded_rcc.py rounded_rcc .... and another command like "script create myrcc rrcc 100 50 10"
With regards to option 3, I didn't get how the rrcc object came into picture. Could you maybe elaborate on that a little bit?
Jaipal Singh (Jul 16 2018 at 16:22):Should I start exploring libged/typec.in to understand how we'll be linking the script parser to the script primitive?
Jaipal Singh (Jul 16 2018 at 16:53):I had a look at the patch that @Shubham Rathore had submitted and gather that it's similar to the other primitives. Thanks for this!
Next would be to have functions that can read the file, invoke the python script parser and create the .g file.
Jaipal Singh (Jul 16 2018 at 16:57):Another issue I see with using the 'in' command is that for most primitives, there's a defined list of parameters that it expects from the user. In case of our procedures, we won't be able to know before hand what arguments are required, hence that becomes a problem.
Jaipal Singh (Jul 16 2018 at 16:58):We could move forward with the assumption that the user knows what parameters are required by the procedure and that he would enter only those (but that's a weak assumption)
Sean (Jul 16 2018 at 16:59):there's an option 3 where we separate the script specification from instantiation, so that might be something like: in rrcc script /path/to/rounded_rcc.py rounded_rcc .... and another command like "script create myrcc rrcc 100 50 10"
With regards to option 3, I didn't get how the rrcc object came into picture. Could you maybe elaborate on that a little bit?
the first command (in rrcc script /path/to/script ...) just creates a handle to the code, then we'd need some other step to instantiate the code
Sean (Jul 16 2018 at 16:59):it's a named handle like any other object, so in that example, I named it rrcc, but it could be anything -- in foo script /path/to/myscript.py func
Sean (Jul 16 2018 at 17:00):script create myinst foo 1 2 3
Jaipal Singh (Jul 16 2018 at 17:00):Oh ok ok
Sean (Jul 16 2018 at 17:01):I had a look at the patch that @Shubham Rathore had submitted and gather that it's similar to the other primitives. Thanks for this!
Next would be to have functions that can read the file, invoke the python script parser and create the .g file.
not quite -- one of the details that need to be figured out is not creating .g file (at least not on disk)
Sean (Jul 16 2018 at 17:02):since it's an object, it's not dynamic if we write it out
Jaipal Singh (Jul 16 2018 at 17:11):not quite -- one of the details that need to be figured out is not creating .g file (at least not on disk.
Python BRLCAD uses the libwdb API and that takes a file_pointer whenever it has to construct a primitive. How do we prevent it from writing to a file then?
Jaipal Singh (Jul 16 2018 at 17:13):Also, it has to be saved into some .g file so that the GUI (mged or archer) can read and display it?
Sean (Jul 16 2018 at 17:13):no it doesn't
Sean (Jul 16 2018 at 17:14):there's an in-memory-only form of databases that we can use
Sean (Jul 16 2018 at 17:14):it happens when you open a database, the API has a parameter that tells it to only write to memory, not to disk
Sean (Jul 16 2018 at 17:15):all the other geometry routines remain unaffected
Sean (Jul 16 2018 at 17:15):what I think needs to happen is that the database instance gets passed or made available automatically to the script
Sean (Jul 16 2018 at 17:18):OR .. maybe a default state in 'wdb' that gets set before the script is run so when it runs import wdb, it's going to be an inmem database when it makes a call like wdb.WDB(database_name, "SGI.g")
Jaipal Singh (Jul 16 2018 at 17:19):what I think needs to happen is that the database instance gets passed or made available automatically to the script
Yea, this makes some sense
Sean (Jul 16 2018 at 17:19):looking at https://github.com/Killthebug/python-brlcad/blob/master/examples/rounded_rcc.py ... there's a few ways I could see this working
Sean (Jul 16 2018 at 17:20):when is __name__ not "__main__" in python?
Jaipal Singh (Jul 16 2018 at 17:20):If you import this module into another file and then call a specific function, then __name__ is not __main__
Sean (Jul 16 2018 at 17:21):hmm
Sean (Jul 16 2018 at 17:22):so doable in a couple ways
Sean (Jul 16 2018 at 17:24):one way (A) could be to run the script and use some global variable (like __name__) for the database instance. this would require scripts to be aware of that instance and would probably be incompatible with external execution.
Sean (Jul 16 2018 at 17:26):another way (B) could be to leave the script as-is, but preload/configure all wdb.WDB() calls to be inmem when run from inside and be a regular file when run externally. this should require no changes to the script, but will require figuring out how the script knows about the database/geometry
Jaipal Singh (Jul 16 2018 at 17:27):(C) How about moving forward with creating classes for procedures? That way the database instance can be passed as an argument to a given object otherwise it can fall back to using some default database.
Sean (Jul 16 2018 at 17:27):another could be to only support "import script" form where __name__ is not main, and calling a specified function that takes a WDB as the first argument (or maybe it's stored in a variable akin to option A)
Sean (Jul 16 2018 at 17:27):(C)
Sean (Jul 16 2018 at 17:28):can you elaborate? classes for procedures?
Jaipal Singh (Jul 16 2018 at 17:28):I'm in the dark about how to go about solving (B)
Sean (Jul 16 2018 at 17:30):(B) could be as simple as putting a line like if __name__ != "__main__" in brlcad.wdb that sets a global configuration for in-mem
Jaipal Singh (Jul 16 2018 at 17:30):Yea.. (I'm still thinking about why I said classes as procedures).
Jaipal Singh (Jul 16 2018 at 17:30):Oh yea!
Jaipal Singh (Jul 16 2018 at 17:32):So, if you look at rounded_rcc.py (A), if you import this into another file (B) and try to call it from there, then the variable __name__ takes the value of the name of the module it is being called from (B).
Jaipal Singh (Jul 16 2018 at 17:32):But it doesn't stop there, the interpreter then goes on to execute all the other functions in rounded_rcc.py from the top to bottom.
Jaipal Singh (Jul 16 2018 at 17:33):One second, let me share an example with you..
Sean (Jul 16 2018 at 17:33):k
Jaipal Singh (Jul 16 2018 at 17:34):https://stackoverflow.com/a/419986
Jaipal Singh (Jul 16 2018 at 17:38):What I was hinting is that the functions which are responsible for creating the procedural shapes be re-written as classes
So in rounded_rcc.py : def rounder_rcc -> class rounder_rcc.
Jaipal Singh (Jul 16 2018 at 17:39):It's not a very massive change IMO, just makes things simpler. Also makes using one procedure in another procedure an easier task.
Jaipal Singh (Jul 16 2018 at 17:48):Something like : https://github.com/Killthebug/python-brlcad/blob/master/examples/class_rounded_rcc.py
Sean (Jul 16 2018 at 17:54):mmhmm.. I'm not sure that's any better or worse -- the script would still be entirely inmem aware and run in one mode or another. ideally we need a solution that doesn't require script modification or minimal modification? flexibility would be good for usability
Sean (Jul 16 2018 at 17:54):like not even worrying about making a def or class
Jaipal Singh (Jul 16 2018 at 17:57):Ok
Sean (Jul 16 2018 at 17:58):writing up an example, hold on
Jaipal Singh (Jul 16 2018 at 17:59):Sure
Jaipal Singh (Jul 16 2018 at 18:01):AKF for 15
Jaipal Singh (Jul 16 2018 at 18:01):AFK for 15
Sean (Jul 16 2018 at 18:03):something like https://gist.github.com/brlcad/7de8f35386c9e40075c07a708fc9aed5
Sean (Jul 16 2018 at 18:03):I think I see a way
Sean (Jul 16 2018 at 18:09):we already decided the function must take arguments, so why not make it strictly be anything that could be run as a program, just that the first argument must be the name of the database! then we can override it with a path to a temporary database file and read the contents into an in-mem automatically.
Sean (Jul 16 2018 at 18:09):/me is liking this ...
Sean (Jul 16 2018 at 18:24):that means we can make it work with literally anything that creates geometry and the python driver becomes incredibly simple and extendible to other command-line generators (like our proc-db and shapes examples)
Jaipal Singh (Jul 16 2018 at 18:39):we already decided the function must take arguments, so why not make it strictly be anything that could be run as a program, just that the first argument must be the name of the database! then we can override it with a path to a temporary database file and read the contents into an in-mem automatically.
Perfect!
Jaipal Singh (Jul 16 2018 at 18:40):read the contents into an in-mem automatically.
How does this happen though?
Sean (Jul 16 2018 at 18:41):there can be a wrapper on the "import" callback to run the script and read the contents on the C side
Jaipal Singh (Jul 16 2018 at 18:42):Alright!
Jaipal Singh (Jul 16 2018 at 18:42):So what should be the next steps for me?
Sean (Jul 16 2018 at 18:42):so then the only question is whether to actually embed a python interpreter, or just run python externally as a command
Sean (Jul 16 2018 at 18:43):I think simplifying the WDB interface on the python side is needed
Jaipal Singh (Jul 16 2018 at 18:43):I recommend running python externally as a command.
Sean (Jul 16 2018 at 18:45):what's in the brlcad module currently?
Sean (Jul 16 2018 at 18:45):obviously at least brlcad.wdb and brlcad.primitives classes
Sean (Jul 16 2018 at 18:45):what else?
Jaipal Singh (Jul 16 2018 at 18:46):vmath repo to handle basic maths
Jaipal Singh (Jul 16 2018 at 18:47):code to generate wrappers for libwdb, libged, libbu, librt, libbrep and libbn
Sean (Jul 16 2018 at 18:55):ged?
Jaipal Singh (Jul 16 2018 at 18:55):Yea
Sean (Jul 16 2018 at 18:55):can you elaborate, what does it wrap ... ??
Jaipal Singh (Jul 16 2018 at 18:59):It's a wrapper for ged.h. Depending on the platform (OSX, Windows, Linux) it loads the libged library.
Sean (Jul 16 2018 at 19:00):but I mean what's it look like -- does it expose all of libged's commands?
Jaipal Singh (Jul 16 2018 at 19:05):No, only a few functions are implemented.
All before line 177 in :
https://github.com/Killthebug/python-brlcad/blob/master/brlcad/ged.py#L177
Jaipal Singh (Jul 16 2018 at 19:06):ls, close, ged_3ptarb and ged_in.
Jaipal Singh (Jul 16 2018 at 19:07):There are barebones structures for the other method but they're still to be fixed.
Sean (Jul 16 2018 at 19:41):looks like the GED class is inmem aware
Jaipal Singh (Jul 16 2018 at 19:50):I think I got it!
https://github.com/Killthebug/python-brlcad/blob/master/brlcad/wdb.py#L72
Jaipal Singh (Jul 16 2018 at 19:50):We can change : RT_WDB_TYPE_DB_DISK to _RT_WDB_TYPE_DB_INMEM
Jaipal Singh (Jul 16 2018 at 19:55):Even after changing, it still creates a .g file... there's something I'm missing.
Jaipal Singh (Jul 16 2018 at 19:55):But if we change the mode it shouldn't right?
Jaipal Singh (Jul 16 2018 at 20:08):I'll figure this out..
Shubham Rathore (Jul 16 2018 at 20:12):@Sean I've uploaded a revised patch (https://sourceforge.net/p/brlcad/patches/500/) in the comments. Have a look !
this looks great, please commit it
Done !
Sean (Jul 16 2018 at 23:59):Even after changing, it still creates a .g file... there's something I'm missing.
yes, it's because of the first call to libwdb.db_open()
Sean (Jul 17 2018 at 00:03):there's a separate call for inmem's, db_open_inmem()
Sean (Jul 17 2018 at 00:56):note, the only reason for getting inmem to work on the python side would be to avoid hitting the disk entirely (i.e., for performance) by running in an embedded interpreter
Sean (Jul 17 2018 at 03:03):@Jaipal Singh what can we do about brlcad.wdb.WDB() ... that seems entirely not user friendly. I don't think the python public API should expose "wdb" anywhere, especially as part of a class that fully wraps wdb
Jaipal Singh (Jul 17 2018 at 03:19):@Sean , right now I'm not aware of any way to bypass brlcad.wdb.WDB(). We can maybe encapsulate in some parent class and hide it from plain sight..
Sean (Jul 17 2018 at 03:20):eh? maybe you misunderstand.
Jaipal Singh (Jul 17 2018 at 03:20):I think so..
Sean (Jul 17 2018 at 03:20):of course we can do something about it... we can rename the WDB class to something that isn't alphabet soup
Sean (Jul 17 2018 at 03:20):we can rename the file it's coming from to something more meaningful
Sean (Jul 17 2018 at 03:21):e.g., brlcad.make.Geometry
Sean (Jul 17 2018 at 03:23):or putting something into brlcad.py like db = brlcad.Database(...)
Sean (Jul 17 2018 at 03:23):or many other possible variations
Sean (Jul 17 2018 at 03:23):something more usable and readable than wdb.WDB
Jaipal Singh (Jul 17 2018 at 03:25):Oh alright! Yea we can do that!
Sean (Jul 17 2018 at 03:25):also, for later today -- can you investigate which wdb functions are on the python side and which are currently not (perhaps make a spreadsheet table showing which are in python and which are not)? grep WDB_EXPORT include/wdb.h ... compare with the WDB class implementation, what isn't used/referenced? we don't want to expose everything as some functions in wdb.h need to go away, but it'd be helpful to see a summary of where we're at and whether any functions are missing.
Sean (Jul 17 2018 at 03:26):Oh alright! Yea we can do that!
already kind of liking brlcad.make.Geometry
Sean (Jul 17 2018 at 03:26):but of course open to ideas if you think up something better :)
Jaipal Singh (Jul 17 2018 at 03:27):I'll make the spreadsheet today! And also work on renaming the functions.
Sure thing will keep you posted :)
Jaipal Singh (Jul 17 2018 at 03:34):@Sean when we say that, the python Public API shouldn't expose wdb anywhere, changing function names still exposes the API right? (Just makes it less obvious)
Sean (Jul 17 2018 at 03:34):what do you mean?
Sean (Jul 17 2018 at 03:35):there's the C api, which shouldn't matter to python scripters -- it only matters as an implementation detail
Sean (Jul 17 2018 at 03:36):e.g., the geometry creation functions happen to be in a "wdb" C library, but they don't need to know that (nor should they care)
Jaipal Singh (Jul 17 2018 at 03:36):Alright. So abstract as much information as we can?
Sean (Jul 17 2018 at 03:36):they want to know how to create/modify geometry
Jaipal Singh (Jul 17 2018 at 03:37):Ok ok, makes sense.
Sean (Jul 17 2018 at 03:38):at some point in the near future, we're going to move functions in wdb closer to the primitives in librt ... and we're going to move the primitives out of librt into their own "libg" library along with the other .g related functions
Sean (Jul 17 2018 at 03:39):oh, when you make a spreadsheet, would you add a column for the primitives listed in src/librt/primitives/tables.cpp and itemize which one's don't have an mk_* function ...?
Jaipal Singh (Jul 17 2018 at 03:40):Sure thing, could do that! :)
Sean (Jul 17 2018 at 03:47):at risk of lining up too much, when you get all that done -- try to implement rt_script_describe() in src/librt/primitives/script.c -- you'll need to modify the rt_script_internal in include/rt/geom.h to hold a filename, src/libged/typein.c needs to ask for and safe the filename, and src/librt/primitives/script.c needs to load the filename during import5 and write it during export5, then print the file's contents during describe... then if you get it all working right, it should show a files contents when you run 'l script' after 'in script script yourscript'!
Sean (Jul 17 2018 at 03:48):should hopefully keep you busy all week -- I"m going to be a bit preoccupied
Jaipal Singh (Jul 17 2018 at 03:52):Sure!
Shubham Rathore (Jul 18 2018 at 18:44):Another issue I see with using the 'in' command is that for most primitives, there's a defined list of parameters that it expects from the user. In case of our procedures, we won't be able to know before hand what arguments are required, hence that becomes a problem.
usually we prompt the user in such cases !
Shubham Rathore (Jul 18 2018 at 19:06):at risk of lining up too much, when you get all that done -- try to implement rt_script_describe() in src/librt/primitives/script.c -- you'll need to modify the rt_script_internal in include/rt/geom.h to hold a filename, src/libged/typein.c needs to ask for and safe the filename, and src/librt/primitives/script.c needs to load the filename during import5 and write it during export5, then print the file's contents during describe... then if you get it all working right, it should show a files contents when you run 'l script' after 'in script script yourscript'!
How's it going? Hope you are not stuck
Shubham Rathore (Jul 19 2018 at 07:25):@Jaipal Singh What is the current status? What is the plan, to execute the script using the script primitive that you are sticking to ?
Jaipal Singh (Jul 19 2018 at 09:31):usually we prompt the user in such cases quote
The prompt that is generated (for each primitive) is hardcoded as your can see in libged/typein.c. The issue we face is that we don't have prior knowledge about what primitives are there in a procedure.
Jaipal Singh (Jul 19 2018 at 09:32):at risk of lining up too much, when you get all that done -- try to implement rt_script_describe() in src/librt/primitives/script.c -- you'll need to modify the rt_script_internal in include/rt/geom.h to hold a filename, src/libged/typein.c needs to ask for and safe the filename, and src/librt/primitives/script.c needs to load the filename during import5 and write it during export5, then print the file's contents during describe... then if you get it all working right, it should show a files contents when you run 'l script' after 'in script script yourscript'!
How's it going? Hope you are not stuck
I haven't started on this yet, I'll get started tonight. I was just completing a couple of other tasks :) Will keep you updated!
Shubham Rathore (Jul 19 2018 at 10:49):usually we prompt the user in such cases quote
The prompt that is generated (for each primitive) is hardcoded as your can see in libged/typein.c. The issue we face is that we don't have prior knowledge about what primitives are there in a procedure.
That is absolutely correct. What I meant to say is to ask the user the numerical occurences of all the relevant primitives. This can be a hack till things start working. Once we have a running model we could parse the details of the primitives present through the script and dynamically modify the typein.c ! This wont be a major task to handle "once we have a running model "
Shubham Rathore (Jul 19 2018 at 10:56):So the work that you've done till now, converts a script to a .g file. The script primitive (just for understanding) runs the python script and makes an object. Now this object should be a part of the .g mentioned previously.( The object is a concatenation of various primitives as mentioned by the user). We need to modify things such that we don't get a .g immediately, rather we include the script object and then go for a .g . This is what I thought ! @Sean any insights.
Sean (Jul 22 2018 at 05:33):I haven't started on this yet, I'll get started tonight. I was just completing a couple of other tasks :) Will keep you updated!
yes... and how are those going too?? there's been no update since our talk.
Sean (Jul 22 2018 at 05:36):talked about a symbol mapping spreadsheet, a column for primitives without a mk_ function, and implementing the 'l' command (it's ft_describe() callback)
Jaipal Singh (Jul 22 2018 at 07:11):something more usable and readable than wdb.WDB
Hey @Sean I've completed renaming the libraries and I've updated all other dependent files, examples and tests.
Jaipal Singh (Jul 22 2018 at 07:12):also, for later today -- can you investigate which wdb functions are on the python side and which are currently not (perhaps make a spreadsheet table showing which are in python and which are not)? grep WDB_EXPORT include/wdb.h ... compare with the WDB class implementation, what isn't used/referenced? we don't want to expose everything as some functions in wdb.h need to go away, but it'd be helpful to see a summary of where we're at and whether any functions are missing.
I've completed this too.
Jaipal Singh (Jul 22 2018 at 07:12):at risk of lining up too much, when you get all that done -- try to implement rt_script_describe() in src/librt/primitives/script.c -- you'll need to modify the rt_script_internal in include/rt/geom.h to hold a filename, src/libged/typein.c needs to ask for and safe the filename, and src/librt/primitives/script.c needs to load the filename during import5 and write it during export5, then print the file's contents during describe... then if you get it all working right, it should show a files contents when you run 'l script' after 'in script script yourscript'!
I've almost completed this, will finish it and submit a patch in a couple of hours.
Jaipal Singh (Jul 22 2018 at 07:13):I've been travelling this weekend, and will also be travelling in the week after this one. My communication lapsed because of the same.
I will update the daily logs too right now.
Jaipal Singh (Jul 22 2018 at 07:39):You can check the commits here. https://github.com/Killthebug/python-brlcad/commits/master
Jaipal Singh (Jul 23 2018 at 04:19):Regarding implementing l describe, I've modified rt_script_describe() in script.c to print out the the contents of the files. But there seems to be an issue with typein.c where in mged stops accepting any response after all the parameters for the in command are passed.
Jaipal Singh (Jul 23 2018 at 04:26):Screen-Shot-2018-07-23-at-9.45.48-AM.png
Jaipal Singh (Jul 23 2018 at 04:26):What do you think could be a possible fix for this?
Shubham Rathore (Jul 23 2018 at 07:07):Probably some issue with the args(). Not a major issue. I don't have access to the code right now, it'll be rectified by the evening. Meanwhile you could focus on opening the script.py file and listing out its contents when prompted by the 'l' command (basically the describe function).
Jaipal Singh (Jul 23 2018 at 07:34):Ah yes, I was thinking along the same lines. The args() in typein.c.
Jaipal Singh (Jul 23 2018 at 07:37):So when you do : in scriptobj script myscript.py, what gets stored in the scriptobj "right now"? Does it create a null objects (scriptobj = null)?
Of course later it would hold the procedurally generated object but we've not coupled the python compiler with the C code yet.
Jaipal Singh (Jul 23 2018 at 07:49):So when you do : in scriptobj script myscript.py, what gets stored in the scriptobj "right now"? Does it create a null objects (scriptobj = null)?
Of course later it would hold the procedurally generated object but we've not coupled the python compiler with the C code yet.
Shubham Rathore (Jul 23 2018 at 09:54):It does not make a .s currently. Null object comes into the picture when the .s is created but stubbed empty for some reason.
Jaipal Singh (Jul 23 2018 at 14:52):Alright
Sean (Jul 23 2018 at 16:56):I'll make the spreadsheet today! And also work on renaming the functions.
Sure thing will keep you posted :)
@Jaipal Singh if you have the spreadsheet, please post it somewhere as that will help us figure out our path forward
Jaipal Singh (Jul 23 2018 at 16:58):https://github.com/Killthebug/python-brlcad/blob/master/wdb_functions_python_brlcad.xlsx
Sean (Jul 23 2018 at 17:22):So when you do : in scriptobj script myscript.py, what gets stored in the scriptobj "right now"? Does it create a null objects (scriptobj = null)?
You can see this in src/libged/typein.c in the script_in() function -- it allocates an rt_script_internal object but only sets the script type (see p_script) .. which you specified as "myscript.py" in response. If you run the 'in' command without arguments, it will prompt interactively based on the p_script prompts.
Sean (Jul 23 2018 at 17:24):https://github.com/Killthebug/python-brlcad/blob/master/wdb_functions_python_brlcad.xlsx
That looks good, thanks. Are you still working on adding a column for the primitives with/without a mk_ function?
Sean (Jul 23 2018 at 17:28):They're all listed in src/librt/primitives/tables.cpp or you can look at the subdirectories in src/librt/primitives/* -- two live elsewhere (comb and binunif). Note that some of the mk_ functions map to the same primitive like arb8 and arb5 are both src/librt/primitives/arb8 (not to be confused with src/librt/primitives/arbn which is a completely different 'arbn' object).
Jaipal Singh (Jul 23 2018 at 17:42):So when you do : in scriptobj script myscript.py, what gets stored in the scriptobj "right now"? Does it create a null objects (scriptobj = null)?
You can see this in src/libged/typein.c in the script_in() function -- it allocates an rt_script_internal object but only sets the script type (see p_script) .. which you specified as "myscript.py" in response. If you run the 'in' command without arguments, it will prompt interactively based on the p_script prompts.
I get this, but why does mged not do anything after accepting the in command (in scriptobj script myscript.py) . For other primitives it draws them, but here it just prints (null) as you can see in the screenshot.
Jaipal Singh (Jul 23 2018 at 17:43):https://github.com/Killthebug/python-brlcad/blob/master/wdb_functions_python_brlcad.xlsx
That looks good, thanks. Are you still working on adding a column for the primitives with/without a mk_ function?
Yea, sharing soon!
Sean (Jul 23 2018 at 18:39):The (null) printing is likely coming from code put into the ft_describe() callback. It could also be coming from the ft_draw() callback which is also in src/librt/primitives/script/script.c -- after the 'in' command, that callback is called since most get displayed
Jaipal Singh (Jul 23 2018 at 18:59):https://github.com/Killthebug/python-brlcad/blob/master/wdb_functions_python_brlcad.xlsx
That looks good, thanks. Are you still working on adding a column for the primitives with/without a mk_ function?
I've update the shared file with another sheet containing list of primitives without a mk_primitive() function.
Jaipal Singh (Jul 23 2018 at 19:01):comb though has two mk_primitive function, mk_comb() & mk_comb1()
Sean (Jul 23 2018 at 19:10):excellent, thank you -- that latter sheet is particularly informative. looks like we're missing wdb hooks for 5 entities
Jaipal Singh (Jul 23 2018 at 19:13):Would these be the last 5 rows in sheet 2?
Sean (Jul 23 2018 at 19:42):Would these be the last 5 rows in sheet 2?
no, they are: rec, superell, revolve, pnts, and datum
Jaipal Singh (Jul 23 2018 at 19:43):Uh-huh
Sean (Jul 23 2018 at 19:48):by the way, great job finding not only the libwdb functions but also the wdb functions in librt
Jaipal Singh (Jul 23 2018 at 19:50)::)
Now, I'm trying to get l describe() to work, still haven't figured my way around the (null) printing.
Sean (Jul 23 2018 at 19:53):a classic method, you can add some print statements into the code so you can tell where things are coming from
Sean (Jul 23 2018 at 19:53):have you added any code to the describe() function?
Jaipal Singh (Jul 23 2018 at 19:54):Yes, I have https://hastebin.com/awugozused.cpp
Jaipal Singh (Jul 23 2018 at 19:54):Also, there is no draw function in src/librt/primitives/script.c
Sean (Jul 23 2018 at 19:55):it's the plot() function
Sean (Jul 23 2018 at 19:56):so you're close, but you have a bug
Jaipal Singh (Jul 23 2018 at 19:57):Ok
Yea..
Sean (Jul 23 2018 at 19:58):what you are fopen()'ing is wrong
Jaipal Singh (Jul 23 2018 at 20:00):testing
Jaipal Singh (Jul 23 2018 at 20:11):I tried two things, the first hardcoding the absolute path for a script with in src/librt/primitives/script.c and second removing all the code that I had introduced in src/librt/primitives/script.c and just compiling with almost the original stubb. The same error persists.
Sean (Jul 23 2018 at 20:14):you're pretty new to C I take it?
Jaipal Singh (Jul 23 2018 at 20:15):Not vey familiar :/
Sean (Jul 23 2018 at 20:15):the error in the snippet you showed me is that you printed "Script type: ..." into buf
Sean (Jul 23 2018 at 20:15):then you try to fopen(buf)
Sean (Jul 23 2018 at 20:16):so you're trying to open a file named "Script type: something..."
Sean (Jul 23 2018 at 20:28):another issue is you should be using bu_fgets() instead of fgetc(), to get the file one line at a time
Sean (Jul 23 2018 at 20:29):the loop you have isn't quite right because you used a char when fgetc() returns an int, so your condition (character != EOF) should never be true. a little surprised it's not an infinite loop
Sean (Jul 23 2018 at 20:30):note, vls stands for variable length string -- it's just a string like you might have in python except you access it via functions
Jaipal Singh (Jul 23 2018 at 20:31):Doesn't fgetc() return a character? http://www.cplusplus.com/reference/cstdio/fgetc/
Sean (Jul 23 2018 at 20:32):look at the green line at the top
Sean (Jul 23 2018 at 20:32):it returns an "int"
Sean (Jul 23 2018 at 20:33):in the example code, it uses "int c;" for the character
Jaipal Singh (Jul 23 2018 at 20:33)::|
Sean (Jul 23 2018 at 20:34):in C a character you find in a string like "abc" is not necessarily the "char" data type, it could be a wide variety of types (especially for foreign languages)
Sean (Jul 23 2018 at 20:34):if you see "char a;" ... don't think of it as saying 'a' is a character in a string
Jaipal Singh (Jul 23 2018 at 20:35):Uh-huh.. So does c typecast when it prints what's returned by fgetc()?
Sean (Jul 23 2018 at 20:35):it's saying 'a' is a character of memory, which C simply defines as being 1 "byte" on the host hardware
Sean (Jul 23 2018 at 20:36):you specified an implicit cast from int to char data types when you assigned the result of fgetc (an int) to c (a char)
Jaipal Singh (Jul 23 2018 at 20:37):I will rewrite the file opening and reading snippet.
Though the current error ((null) printing) seems to occur before any code changes I might have made..
Sean (Jul 23 2018 at 20:37):you then test for c != EOF in the loop ... well EOF could be anything but it's typically "(int)-1"
Sean (Jul 23 2018 at 20:37):there very well may be nulls getting printed elsewhere in the code, the two are not exclusive
Jaipal Singh (Jul 23 2018 at 20:38):I doubt that because a (null) gets printed everytime I press 'enter' after completing the in command
Sean (Jul 23 2018 at 20:39):i mean those aren't the only issues in the function -- the "Script is: " printing at the end is also not right :)
Sean (Jul 23 2018 at 20:40):why would you doubt that? that would only matter if you're assuming your describe function is all that's being called when you run the 'in' command and it most certainly is not
Sean (Jul 23 2018 at 20:41):you'd have to read the 'in' command's implementation, trace through the logic to see where describe is called, confirm that is where it's really being called from (e.g., by adding a bu_log() line), then see what other logic or functions get called.
Jaipal Singh (Jul 23 2018 at 20:41):i mean those aren't the only issues in the function -- the "Script is: " printing at the end is also not right :)
Yea, I've scratched away all that. I'll go through typein.c to figure out where it's going wrong.
Jaipal Singh (Jul 23 2018 at 20:42):you'd have to read the 'in' command's implementation, trace through the logic to see where describe is called, confirm that is where it's really being called from (e.g., by adding a bu_log() line), then see what other logic or functions get called.
On it
Sean (Jul 23 2018 at 20:42):you were super close with what you had ...
Sean (Jul 23 2018 at 20:43):another approach would be to start with a function that does nothing -- maybe just prints hello
Shubham Rathore (Jul 24 2018 at 03:43):I doubt that because a (null) gets printed everytime I press 'enter' after completing the in command
@Jaipal Singh This error is because nvals in ged_in() is set as 2, whereas it should be 1 only.
Shubham Rathore (Jul 24 2018 at 03:45):Try setting it 1, at least this error should go away. There will be many more errors in the way. It'll take time to encounter those, so let's work a bit fast.
Jaipal Singh (Jul 24 2018 at 08:53):Hey @Shubham Rathore , yes changed nvals to 1 but I bump into a : Segmentation fault: 11 after I execute the in command.
Shubham Rathore (Jul 24 2018 at 08:54):ah, as expected, new error pops out. I don't have access to the code right now. You familiar with gdb ?
Jaipal Singh (Jul 24 2018 at 08:55):Not very, but I can figure it out! What do I need to do?
Shubham Rathore (Jul 24 2018 at 08:59):Okay, so gdb is a tool that helps to detect seg faults. We need to set -g flag while compiling using gcc.
Shubham Rathore (Jul 24 2018 at 09:00):In this case since we use make, All you need to do is make CFLAGS= -g
Shubham Rathore (Jul 24 2018 at 09:00):Then run gdb mged
Shubham Rathore (Jul 24 2018 at 09:01):Then type "run" , and try making a script object. The moment code breaks, it'll show the line, that is breaking the code
Shubham Rathore (Jul 24 2018 at 09:01):Do you use a mac ?
Jaipal Singh (Jul 24 2018 at 09:01):Yes, I do use a mac
Shubham Rathore (Jul 24 2018 at 09:02):Okay, so you need to use sudo before gdb mged
Jaipal Singh (Jul 24 2018 at 09:02):OK sure!
Shubham Rathore (Jul 24 2018 at 09:02):Otherwise you wont land where you want to
Shubham Rathore (Jul 24 2018 at 09:03):So keeping aside the runtime issues as of now. Do you have the code for decribe() ?
Jaipal Singh (Jul 24 2018 at 09:13):Yea, I had shared it with Sean yesterday, https://hastebin.com/awugozused.cpp, He pointed out a lot of changed but also hinted that I was fairly close to completion. I'll change this and share the updated code soon.
Shubham Rathore (Jul 24 2018 at 09:14):Ah fine, I'll have a look after my office hours. I have restricted access to hastebin :P
Jaipal Singh (Jul 24 2018 at 09:21):It raises the error : During startup program terminated with signal SIGTRAP
This appears to be an OSX issue. I'm fixing it right now..
Jaipal Singh (Jul 24 2018 at 09:23):Seems to be a security issue in Sierra. https://stackoverflow.com/a/40044913
Jaipal Singh (Jul 24 2018 at 09:33):Managed to avoid SIGTRAP by creating a .gdbinit file in my user folder.
Though this has given rise to a new error : During startup program terminated with signal ?, Unknown signal.
Shubham Rathore (Jul 24 2018 at 09:34):Did you use sudo ?
Shubham Rathore (Jul 24 2018 at 09:34):I have sierra, and it works well for me usually
Jaipal Singh (Jul 24 2018 at 09:36):Yes I did use sudo.
Even on the forums online it works for some and for the rest, raises the error that I'm encountering.
Shubham Rathore (Jul 24 2018 at 09:38):Let's not just waste time. Give it another try, if it works... good! otherwise work on the describe(), I'll try to figure it out by this evening.
Jaipal Singh (Jul 24 2018 at 09:41):in /build folder i did : make 'CFLAGS=-g'
then in /build/bin I did : sudo gdb ./mged followed by 'run'
Shubham Rathore (Jul 24 2018 at 09:42):yess.... it should work
Jaipal Singh (Jul 24 2018 at 09:42):(gdb) run Starting program: /Users/Troller/Documents/GSoC/18/brlcad/build/bin/mged [New Thread 0x1603 of process 74940] [New Thread 0x1903 of process 74940] During startup program terminated with signal ?, Unknown signal.
Shubham Rathore (Jul 24 2018 at 09:44):ah.. maybe try LLDB
Shubham Rathore (Jul 24 2018 at 09:44):runs on the same line as GDB
Shubham Rathore (Jul 24 2018 at 09:47):You'll have to handle the clang and gcc stuff ! :(
Jaipal Singh (Jul 24 2018 at 09:57):Sure, will try lldb. Working on the code for l describe now..
Jaipal Singh (Jul 24 2018 at 12:22):Screen-Shot-2018-07-24-at-5.51.12-PM.png It's not complete but I've hacked my way around printing the contents of a file using l describe. Currently I modified /librt/primitives/ell/ell.c to experiment.
Jaipal Singh (Jul 24 2018 at 12:23):I'm modifying typein.c to read the input file too, but that's not working very well for me. Should fix it in a while.
Jaipal Singh (Jul 24 2018 at 12:44):If I have
struct bu_vls x char a[256]
bu_vls_strcpy(*x, a) copies the contents of x to a?
If I want to copy the contents of a to x, how do I do it?
Sean (Jul 24 2018 at 13:17):Okay, so you need to use sudo before gdb mged
er, no you don't :)
Sean (Jul 24 2018 at 13:19):or... maybe you do -- is this a new security feature? I've not run into it yet with lldb
Sean (Jul 24 2018 at 13:20):looking good @Jaipal Singh .. how about a link to the code changes (svn diff > mychanges.diff)
Sean (Jul 24 2018 at 13:22):If I have
struct bu_vls x
char a[256]
bu_vls_strcpy(*x, a) copies the contents of x to a? If I want to copy the contents of a to x, how do I do it?
no, it doesn't. several issues -- first, you should initialize both on declaration (e.g., struct bu_vls x = BU_VLS_INIT_ZERO; char a[256] = {0};)
Sean (Jul 24 2018 at 13:23):to copy from x to a, it would be bu_strlcpy(a, bu_vls_addr(&x), 256);
Sean (Jul 24 2018 at 13:24):to copy from a to x, it would be bu_vls_strcpy(&x, a);
Jaipal Singh (Jul 24 2018 at 13:46): Jaipal Singh (Jul 24 2018 at 13:46):to copy from a to x, it would be bu_vls_strcpy(&x, a);
I'll work with this soon and keep you posted!
Shubham Rathore (Jul 24 2018 at 14:52):Okay, so you need to use sudo before gdb mged
er, no you don't :)
https://stackoverflow.com/questions/11504377/gdb-fails-with-unable-to-find-mach-task-port-for-process-id-error
@Sean Have a look. Not a new feature. All Os'es after snow leopard have this issue.
Shubham Rathore (Jul 25 2018 at 03:28):Screen-Shot-2018-07-24-at-5.51.12-PM.png It's not complete but I've hacked my way around printing the contents of a file using l describe. Currently I modified /librt/primitives/ell/ell.c to experiment.
well, seems to be a good attempt !
Sean (Jul 25 2018 at 03:30):https://stackoverflow.com/questions/11504377/gdb-fails-with-unable-to-find-mach-task-port-for-process-id-error
@Sean Have a look. Not a new feature. All Os'es after snow leopard have this issue.
I read up on it a bit later. I guess I hadn't noticed because GDB was removed from XCode quite some time ago, after they switched to LLVM. I don't think lldb has that issue. Or, at least, I have been using it regularly without going sudo... unless my pending update from last week changes that! :)
Sean (Jul 25 2018 at 04:12):well, seems to be a good attempt !
I agree, that looks good! Not far off of what I would expect "l" to show for a script object. You say you did that on the ell primitive? Or that you used ell as a guide for script?
Jaipal Singh (Jul 25 2018 at 05:15):Yea, I modified the ell primitive's l describe function. I'm using as a sandbox before I move the code to the script primitive. (Could've chosen any primitive to test, picked ell randomly)
Jaipal Singh (Jul 25 2018 at 06:10):I'm having a tough time copying a char array into a bu_vls. This the code I'm using to do the same : https://hastebin.com/anapelazez.php.
Jaipal Singh (Jul 25 2018 at 06:12):no, it doesn't. several issues -- first, you should initialize both on declaration (e.g., struct bu_vls x = BU_VLS_INIT_ZERO; char a[256] = {0};)
Doing this, raises the error :
brlcad/src/libged/typein.c:1513:27: expected expression eip->reference_path = BU_VLS_INIT_ZERO;
Shubham Rathore (Jul 25 2018 at 10:06):Okay so here is something I was thinking upon. What describe() does now is.... just print the file. @Jaipal Singh If you have explored the describe() function of other primitives then you would have found out that we get the primitive params on calling describe(). What we should exactly get here is ... the details of every primitive present in the script and there corresponding params. For ex. if you've made a tank through procedural scripts, then describe should mention all the primitives used here, be it an rcc or an sph etc. This is something to think over !
Jaipal Singh (Jul 25 2018 at 13:03):Nice recommendation @Shubham Rathore! I think something like this could and should be done.. If I were to do it from scratch, it would involve write a parser for our script and then based on calls to create individual objects in the script, we could indicate/print what primitives were used. This though I feel is a bigger project on it's own. There are numerous ways to go about doing it and given the multiple ways in which the scripts can be written, it becomes a very time intensive task. IMO
Jaipal Singh (Jul 25 2018 at 13:06):You'll have to handle the clang and gcc stuff ! :(
I still haven't managed to get the bedugger to work. Spent a couple of hours breaking my head on it and tried a few solutions (https://stackoverflow.com/questions/39702871/gdb-kind-of-doesnt-work-on-macos-sierra) but kept failing :(
Shubham Rathore (Jul 25 2018 at 14:10):You'll have to handle the clang and gcc stuff ! :(
I still haven't managed to get the bedugger to work. Spent a couple of hours breaking my head on it and tried a few solutions (https://stackoverflow.com/questions/39702871/gdb-kind-of-doesnt-work-on-macos-sierra) but kept failing :(
gdb and macos are a pain !
Jaipal Singh (Jul 25 2018 at 15:01):@Sean Why do you think typein.c saves the string but it doesn't reflect in src/librt/primitives/ell.c
Jaipal Singh (Jul 25 2018 at 21:55):@Sean could you also help me figure out why the script primitive that was stubbed keeps crashing? I have wasted enough time on trying to figure it out.
Also, our immediate goal (by the end of this week) should be to be able to invoke the python compiler from C and save the procedurally created geometry in our working database.
Sean (Jul 26 2018 at 02:07):no, it doesn't. several issues -- first, you should initialize both on declaration (e.g., struct bu_vls x = BU_VLS_INIT_ZERO; char a[256] = {0};)
Doing this, raises the error :
brlcad/src/libged/typein.c:1513:27: expected expression
eip->reference_path = BU_VLS_INIT_ZERO;
What you have there is not initialization on declaration - it's after. BU_VLS_INIT_ZERO is where it's first declared. You'd need to use something like bu_vls_init() to initialize it later. In C/C++, there's different mechanisms for initializing memory.
Sean (Jul 26 2018 at 02:13):I'm having a tough time copying a char array into a bu_vls. This the code I'm using to do the same : https://hastebin.com/anapelazez.php.
this actually looks about right if your cmd_argvs indices are correct other than the buffer copy seeming unnecessary
Sean (Jul 26 2018 at 03:17):What we should exactly get here is ... the details of every primitive present in the script and there corresponding params. For ex. if you've made a tank through procedural scripts, then describe should mention all the primitives used here, be it an rcc or an sph etc. This is something to think over !
Actually, I think we can get away with not knowing what the script made or how objects were named or how they were structured. All that we need is to get a list of top objects and load them.
Sean (Jul 26 2018 at 03:18):I still haven't managed to get the bedugger to work. Spent a couple of hours breaking my head on it and tried a few solutions (https://stackoverflow.com/questions/39702871/gdb-kind-of-doesnt-work-on-macos-sierra) but kept failing :(
Is there a reason you're not using lldb?
Sean (Jul 26 2018 at 03:20):@Sean Why do you think typein.c saves the string but it doesn't reflect in src/librt/primitives/ell.c
This is likely because the string needs to be saved and restored in two of the other callbacks. The import5 and export5 callbacks read and write geometry objects respectively.
Sean (Jul 26 2018 at 03:25):@Sean could you also help me figure out why the script primitive that was stubbed keeps crashing? I have wasted enough time on trying to figure it out.
Sure, post a diff of what you're trying and I'll take a look.
Also, our immediate goal (by the end of this week) should be to be able to invoke the python compiler from C and save the procedurally created geometry in our working database.
That sounds good. You'll want to use bu_temp_file() to create a file for you (see header declaration in include/bu dir for documentation on that function or search the code for some examples using it, which will get specified as an arg to the script. You can probably start with something as simple as popen() to invoke python on the script.
Shubham Rathore (Jul 26 2018 at 04:20):I still haven't managed to get the bedugger to work. Spent a couple of hours breaking my head on it and tried a few solutions (https://stackoverflow.com/questions/39702871/gdb-kind-of-doesnt-work-on-macos-sierra) but kept failing :(
Is there a reason you're not using lldb?
lldb gives - >
db_lookup(EZERO) failed: EZERO does not exist
db_string_to_path() of 'EZERO' failed on 'EZERO'
db_lookup(EZERO) failed: EZERO does not exist
Sean (Jul 26 2018 at 04:28):that looks like straight up memory corruption. EZERO is probably from a PAGEZERO literal somewhere in random memory.
Sean (Jul 26 2018 at 04:54):found the bug
Sean (Jul 26 2018 at 05:14):bugs
Sean (Jul 26 2018 at 05:17):@Jaipal Singh @Shubham Rathore on latest trunk (r71268), I fixed a couple bugs. the nvals value was off by one and the function signature wasn't right on script_in(). also the vls serialization in export5/import5 doesn't look right so I commented it out. left some !!! debugging code in there so you can see what's going on now, a simple substitute for debugging without a debugger.
Shubham Rathore (Jul 26 2018 at 05:19):yes, we figured out nvals a day before. Ah fine..
Sean (Jul 26 2018 at 05:19):script_in was the big one ... stack corruption
Shubham Rathore (Jul 26 2018 at 05:21):cmd_argvs[0] ? shouldn't it be 3 ?
Jaipal Singh (Jul 26 2018 at 11:13):Screen-Shot-2018-07-26-at-4.32.46-PM.png
Shubham Rathore (Jul 26 2018 at 11:18):Need to give this thing some time
Jaipal Singh (Jul 26 2018 at 11:20):The import_5 function body looks standard like it is for almost all the other primitives. I don't see why db_i is becoming a Zero_Magic_Number.
Jaipal Singh (Jul 26 2018 at 11:24):The structure pointer fails to get validated.
Shubham Rathore (Jul 26 2018 at 11:37):Don't have access to the code, will look into it later today
Shubham Rathore (Jul 26 2018 at 11:38):But yes, it is identical, I did this previously for the annot primitive. I didn't find much of a difference. Will have to figure this out.
Jaipal Singh (Jul 26 2018 at 11:44):cmd_argvs[0] ? shouldn't it be 3 ?
Yes!
Jaipal Singh (Jul 26 2018 at 13:16):I still haven't managed to get the bedugger to work. Spent a couple of hours breaking my head on it and tried a few solutions (https://stackoverflow.com/questions/39702871/gdb-kind-of-doesnt-work-on-macos-sierra) but kept failing :(
Is there a reason you're not using lldb?
Started using it now!
Sean (Jul 26 2018 at 13:23):cmd_argvs[0] ? shouldn't it be 3 ?
I didn't try to fix all of the problems, just got it working with some debugging so it's clear what's going on ;)
Sean (Jul 26 2018 at 13:26):@Jaipal Singh you should add more logging or step through the debugger and look at the value of variables for the next step. the script_in function creates the initial object, that should be going to export5 next, followed by import5... you'll want to make sure each one is doing the right thing
Sean (Jul 26 2018 at 13:26):saving out the vls string and then recreating it
Jaipal Singh (Jul 26 2018 at 13:38):Sure.
Sean (Jul 26 2018 at 13:44):the reports are perfect, by the way
Sean (Jul 26 2018 at 13:45):let me know if you can't figure out the vls serialization by tomorrow, let me know -- there's a good example in the dsp.c code
Jaipal Singh (Jul 26 2018 at 14:23):Thanks for the clues! I am genuinely distraught about not being able to copy and reference a string! Positively will finish by tomorrow!
Jaipal Singh (Jul 27 2018 at 02:59):In any rt_primitive_import5, when you set : BU_ASSERT(ep->ext_nbytes = X). How do you decide what is X? Is it the sum the sizes of all the variables present in that primitives struct rt_primitive_internal?
Shubham Rathore (Jul 27 2018 at 03:01):BTW the export function is fixed and is working fine. The import function is giving in some errors, that will be fixed soon.
Jaipal Singh (Jul 27 2018 at 03:01):Awesome! Thanks! What was the issue?
Shubham Rathore (Jul 27 2018 at 03:02):Ah.. some bu_vls related stuff ;)
Shubham Rathore (Jul 27 2018 at 03:02):I'll commit it soon, once import is fixed
Jaipal Singh (Jul 27 2018 at 03:03):Great. :D
Shubham Rathore (Jul 27 2018 at 03:06):In any rt_primitive_import5, when you set : BU_ASSERT(ep->ext_nbytes = X). How do you decide what is X? Is it the sum the sizes of all the variables present in that primitives struct rt_primitive_internal?
Look for the same assert line in ell.c and look for the params of ell in rt/geom.h . I think you'll find your answer. ;)
Jaipal Singh (Jul 27 2018 at 03:12):That's exactly what the hypothesis was based on :P
Sean (Jul 27 2018 at 03:16):Thanks for the clues! I am genuinely distraught about not being able to copy and reference a string! Positively will finish by tomorrow!
Not a big deal if you can't figure it out -- I have in mind some other things we might want to pivot onto for the time remaining. This is definitely code that will need to be written, but there's also still more work on the Python side too.
Jaipal Singh (Jul 27 2018 at 03:19):I'll revert back if I can't get it to work in an hour from now.
Yes, eager for more work on the python side!
Sean (Jul 27 2018 at 03:52):In any rt_primitive_import5, when you set : BU_ASSERT(ep->ext_nbytes = X). How do you decide what is X? Is it the sum the sizes of all the variables present in that primitives struct rt_primitive_internal?
also would note that "ep->ext_nbytes = X" would be an error... you probably mean ==
Jaipal Singh (Jul 27 2018 at 03:54):@Sean , in (src/librt/primitives/dsp/dsp.c) dsp_export5(), line 4510, we're copying dsp_ip->dsp_name to cp. Why do we do this? cp is a local var...
Sean (Jul 27 2018 at 03:56):@Jaipal Singh of all the primitives in your spreadsheet, which ones aren't yet bound on the python side? wasn't that was supposed to be a column too? :)
Sean (Jul 27 2018 at 03:56):/me looks
Sean (Jul 27 2018 at 04:06):@Sean , in (src/librt/primitives/dsp/dsp.c) dsp_export5(), line 4510, we're copying dsp_ip->dsp_name to cp. Why do we do this? cp is a local var...
see line 4451 -- cp is a pointer to a byte buffer .. so printing into cp is printing into that buffer. we're writing the name of the dsp file/object into the buffer which is then written out to disk. note that it's written to cp after a bunch of other things are written into cp and the pointer is advanced forward by the number of bytes written. this is a low-level pointer.
Jaipal Singh (Jul 27 2018 at 04:09):Correct, I get how cp works. But then cp is lost after the function call is completed right?
Jaipal Singh (Jul 27 2018 at 04:10):@Jaipal Singh of all the primitives in your spreadsheet, which ones aren't yet bound on the python side? wasn't that was supposed to be a column too? :)
I made another sheet for the same IIRC. Let me find it, else will update the sheet on github in 5 min!
Jaipal Singh (Jul 27 2018 at 04:17):Yes! I had made a sheet (https://docs.google.com/spreadsheets/d/1XeeI1mpLlPFZtc7i5GeOmjLFE-s8JMsYpwVszIPsrmQ/edit?usp=sharing) but this was for when we were working with the TCL approach. I'll add a new column and update asap.
Sean (Jul 27 2018 at 12:34):Correct, I get how cp works. But then cp is lost after the function call is completed right?
The cp pointer does go out of scope after the function call is completed, but nothing is "lost". The buffer memory it was pointing to is still wherever it was at and is accessible via the ext_buf in the bu_external object. I could just as well declare another pointer and set it to that address.
Jaipal Singh (Jul 28 2018 at 04:06):Uh-huh!
Jaipal Singh (Jul 28 2018 at 04:08):@Sean , what's the work on the python end you were talking about?
Sean (Jul 28 2018 at 13:50):@Sean , what's the work on the python end you were talking about?
alright, so with the time left, I think we should focus on tidying up the python side more
Sean (Jul 28 2018 at 13:51):we've gotten things started on the C side and can finish that up later, you got us a great to-do list (the spreadsheets) for things that need to be completed there
Sean (Jul 28 2018 at 13:56):@Jaipal Singh on the python side, there are a handful of things needed but I think one of the most useful will be a python script that creates one of everything currently supported by the interface
Sean (Jul 28 2018 at 13:57):maybe create a 'half' for ground, then organize a scene with all the primitives on top/above it
Sean (Jul 28 2018 at 13:59):use can your creativity on how they're organized, maybe group them by general type or shape
Sean (Jul 28 2018 at 14:00):alternatively, you could create one "scene" per primitive type, showing the variety of shapes that primitive can make -- like how an rcc cylinder can be short and fat like a disc, tall and skinny like a pencil, or deep like a bucket, etc
Sean (Jul 28 2018 at 14:01):then all primitives could be tiled into a bigger scene for all of them
Sean (Jul 28 2018 at 14:03):That should take you a couple days, from there I think we need to focus on documentation -- so maybe keep that in mind when creating the primitives code -- so that the code can serve as documentation of the primitives
Sean (Jul 28 2018 at 14:04):last but not least, how about writing up a brief 1-page installation+run tutorial
Shubham Rathore (Jul 29 2018 at 05:14):The export function was fixed. I committed the code. AFA the import function is considered, it absolutely looks fine but gives ABORT trap 6 on running.
Shubham Rathore (Jul 29 2018 at 05:15):Tried a bu_log just at the beginning of the import func() and nothing is displayed on running. The control somehow is not entering the function.
Jaipal Singh (Jul 30 2018 at 03:59):@Sean Affirmative! I've got something to do for the entire week now. Will start off with the script to create all the primitives now and then move onto implementing half! :)
Shubham Rathore (Jul 30 2018 at 04:03):Make sure, you leave time for the documentation. That is something devs tend to ignore, but is actually something of great importance. We still struggle for documentation. Cheers !
Jaipal Singh (Jul 30 2018 at 05:11):Indeed :)
Sean (Jul 31 2018 at 01:43):Tried a bu_log just at the beginning of the import func() and nothing is displayed on running. The control somehow is not entering the function.
thank you @Shubham Rathore ! excellent work. I'll take a look to see if I can spot what's wrong. likely a mismatch between import and export (they must match byte for byte)
Sean (Jul 31 2018 at 01:43):and in the right byte ordering
Shubham Rathore (Jul 31 2018 at 04:27):@Sean FIXED !!
Jaipal Singh (Jul 31 2018 at 05:00):@Sean , @Shubham Rathore I'm almost done with the tutorial for making all the primitives. I've shared the link in the lastest report. Will fix a couple of broken things today. I made changes in some files and I haven't updated on of the tutorials (read report for more) .
After that I'll start work on the half primitive and the big scene! :)
Jaipal Singh (Jul 31 2018 at 05:00):@Shubham Rathore Awesome!
Shubham Rathore (Jul 31 2018 at 05:07):So, check for the lastest revision. Script object is created without any issue. @Jaipal Singh Once you are done with the python work that Sean told you to do, do try to incorporate the describe function that you tried by tweaking ell's func on the script. With only a few days to go, try to make things work faster. Cheers !
Jaipal Singh (Jul 31 2018 at 05:19):Yea sure! The major body of the describe function is ready, I'll just need to modify it to sit in the script primitive. Let me know if the examples look alright btw?
Shubham Rathore (Jul 31 2018 at 10:29):Had a glimpse of the examples, couldn't look at them in detail (stuck somewhere) . There seem to be some formatting issues on the first look. I'll look at them in detail soon and let you know !
Sean (Jul 31 2018 at 13:31):@Jaipal Singh what work is there to do on the half primitive?
Sean (Jul 31 2018 at 13:34):@Shubham Rathore excellent work, what was wrong?
Shubham Rathore (Jul 31 2018 at 13:37):The allocation for ep->ext_nbytes was a bit messy. :P
Jaipal Singh (Jul 31 2018 at 14:21):@Jaipal Singh what work is there to do on the half primitive?
Nothing major, will share the commit soon.
Jaipal Singh (Jul 31 2018 at 14:24):maybe create a 'half' for ground, then organize a scene with all the primitives on top/above it
Just not able to create the half for ground :P Screen-Shot-2018-07-31-at-7.53.20-PM.png
Jaipal Singh (Jul 31 2018 at 14:24):Does this scene look good?
Additionally I'll create scenes for individual primitives too.
Jaipal Singh (Jul 31 2018 at 15:02):alternatively, you could create one "scene" per primitive type, showing the variety of shapes that primitive can make -- like how an rcc cylinder can be short and fat like a disc, tall and skinny like a pencil, or deep like a bucket, etc
@Sean , I'm going to work on this now! Should be done in a day or two. The will move onto documentation.
Jaipal Singh (Jul 31 2018 at 18:04):@Shubham Rathore I just came to know about the mentor summit! Are you headed for it?
Jaipal Singh (Jul 31 2018 at 18:04):Also, do you get to take along a student with you? :P
Shubham Rathore (Jul 31 2018 at 18:30):maybe create a 'half' for ground, then organize a scene with all the primitives on top/above it
Just not able to create the half for ground :P Screen-Shot-2018-07-31-at-7.53.20-PM.png
So you managed to get the most of them... wish this was aesthetically more pleasing to see.
Shubham Rathore (Jul 31 2018 at 18:31):alternatively, you could create one "scene" per primitive type, showing the variety of shapes that primitive can make -- like how an rcc cylinder can be short and fat like a disc, tall and skinny like a pencil, or deep like a bucket, etc
@Sean , I'm going to work on this now! Should be done in a day or two. The will move onto documentation.
This shouldn't take time. Make this faster .
Jaipal Singh (Jul 31 2018 at 18:32):Ah! sorry it wasn't. I'll work on making it look better.
Jaipal Singh (Jul 31 2018 at 18:32):Hmm, ok.
Shubham Rathore (Jul 31 2018 at 18:32):@Shubham Rathore I just came to know about the mentor summit! Are you headed for it?
Limited slots! Lets see
Shubham Rathore (Jul 31 2018 at 18:33):Also, do you get to take along a student with you? :P
Sean would answer this better :P
Jaipal Singh (Jul 31 2018 at 18:34):Ah ok! [Fingers crossed]
Sean (Aug 01 2018 at 04:29):Does this scene look good?
Additionally I'll create scenes for individual primitives too.
It looks like a good start .. they need to be organized in some meaningful manner ... they're not balanced in size, not on a horizontal plane (looks like they go vertically?), could use some color or lights to make it a showcase scene. Also, aren't there some missing? There are nearly 10 more on the spreadsheet listed as done, no?
Jaipal Singh (Aug 01 2018 at 06:48):Does this sort of grouping look alright? Screen-Shot-2018-08-01-at-12.18.21-PM.png
Jaipal Singh (Aug 01 2018 at 06:49):Yea, they're on a vertical place.
Shubham Rathore (Aug 01 2018 at 06:52):Have you included all of them ??
Shubham Rathore (Aug 01 2018 at 06:52):Looks like we are missing some objects
Jaipal Singh (Aug 01 2018 at 06:54):No, no, this is still in progress.
Jaipal Singh (Aug 01 2018 at 06:54):I'm asking about if this type of clustering is alright.
Jaipal Singh (Aug 01 2018 at 06:54):alternatively, you could create one "scene" per primitive type, showing the variety of shapes that primitive can make -- like how an rcc cylinder can be short and fat like a disc, tall and skinny like a pencil, or deep like a bucket, etc
Something like this.
Shubham Rathore (Aug 01 2018 at 06:55):Better than the last one !
Shubham Rathore (Aug 01 2018 at 06:56):@Sean will have more to say !
Jaipal Singh (Aug 01 2018 at 06:56):Cool.
Jaipal Singh (Aug 01 2018 at 08:22):I'm drawing a grip with the following attributes :
"grip.s",
center = (0, 5, 1),
normal = (1, 0, 0),
magnitude = 3,
Jaipal Singh (Aug 01 2018 at 08:22):It gets drawn alright, but when I try to render it, nothing renders. It doesn't say rt failed, it just shows a blank screen, any idea why this might be happening?
Jaipal Singh (Aug 01 2018 at 15:40):So, there's a few primitives like extrude, sketch, and a few more (will share the list) which I don't have an understanding of. Those are the ones missing from the scene that I had shared. It was also missing a few basic primitive like grip, superell and metaball which I've introduced now.
Jaipal Singh (Aug 01 2018 at 15:41):I will spend a little time to understand and implement the primitives that are new to me.
Jaipal Singh (Aug 01 2018 at 15:41):If you have any leads or good examples, please share. :)
Jaipal Singh (Aug 01 2018 at 18:34):Specifically examples of 'vol'.
Shubham Rathore (Aug 02 2018 at 02:51):Grip doesn't show in rt :P
Shubham Rathore (Aug 02 2018 at 02:52):Its absolutely normal !
Shubham Rathore (Aug 02 2018 at 02:54):So, there's a few primitives like extrude, sketch, and a few more (will share the list) which I don't have an understanding of. Those are the ones missing from the scene that I had shared. It was also missing a few basic primitive like grip, superell and metaball which I've introduced now.
Sketch, extrude aren't big things. Sketch doesn't raytrace again. Extrude is built on top of sketch. Have a read.
Jaipal Singh (Aug 02 2018 at 03:38):Yea, I've read through them. I'm just experimenting with including them in a scene..
Jaipal Singh (Aug 02 2018 at 05:08):https://docs.google.com/spreadsheets/d/1XeeI1mpLlPFZtc7i5GeOmjLFE-s8JMsYpwVszIPsrmQ/edit?usp=sharing
Jaipal Singh (Aug 02 2018 at 05:09):^Here's the primitives that can be created. I've introduced a column for those that I have not yet included in the scene with comments.
Sean (Aug 02 2018 at 05:15):I'm asking about if this type of clustering is alright.
better! please keep them on a horizontal plane, though, so that the base of every primitive ends up at Z=0 (not centered at Z=0, the _base_ or lowest point at Z=0 ... )
Jaipal Singh (Aug 02 2018 at 05:15):Ok, sure, could do that
Sean (Aug 02 2018 at 05:17):fyi, there are 3 or 4 other attempts at this you can check out with a search, e.g., https://www.google.com/search?q=brl-cad+primitives&safe=off&client=safari&rls=en&source=lnms&tbm=isch&sa=X&ved=2ahUKEwjona2rzs3cAhWLwFkKHXKrDSIQ_AUoAnoECAsQBA&biw=1494&bih=912
Sean (Aug 02 2018 at 05:18):It gets drawn alright, but when I try to render it, nothing renders. It doesn't say rt failed, it just shows a blank screen, any idea why this might be happening?
grips don't have a visual representation -- they are controls for animating other geometry
Jaipal Singh (Aug 02 2018 at 05:18):Ah! The very image, was my sort of inspiration. I'll give the arrangement a little more attention now :)
Jaipal Singh (Aug 02 2018 at 05:18):Uh-huh
Sean (Aug 02 2018 at 05:19):these may help with a few you're missing: http://brlcad.org/wiki/Sketch
Sean (Aug 02 2018 at 05:19):http://brlcad.org/wiki/BRL-CAD_Primitives#ars
Jaipal Singh (Aug 02 2018 at 05:19):Yea, I've been using these only. Will finish off all the 2d derived primitives soon
Sean (Aug 02 2018 at 05:20):http://brlcad.org/wiki/Extrude
Jaipal Singh (Aug 02 2018 at 05:20):Just wanted to know, how do you copy, paste in mged?
Sean (Aug 02 2018 at 05:21):revolve is a lot like extrude, but simpler in a lot of ways as you don't need or want to define a closed loop, you just draw a 2D line or curve
Jaipal Singh (Aug 02 2018 at 05:21):Uh-huh
Sean (Aug 02 2018 at 05:21):it depends on your platform, it's typically a system or X11 key binding
Jaipal Singh (Aug 02 2018 at 05:23):Got it! (ctrl+v and not cmd+v) :P
Sean (Aug 02 2018 at 05:23):there's a basic 2D sketch editor in mged that you can use to create sketch objects for testing with extrude and revolve ... when you get them working, you can then figure out how to describe something similar in the python api
Sean (Aug 02 2018 at 05:24):yeah, on mac, that's technically an X11 window -- so X11 bindings are ctrl by default
Sean (Aug 02 2018 at 05:24):there is an native aqua version that uses cmd, but that's not readily available to you
Jaipal Singh (Aug 02 2018 at 05:25):Alright
Jaipal Singh (Aug 02 2018 at 05:43):@Sean , just curious, why not organize everything along the z-axis? Is it because it's not aesthetically pleasing?
Jaipal Singh (Aug 02 2018 at 05:45):(Something like this)[https://brlcad.org/gallery/var/resizes/diagrams/primitives.png]
Sean (Aug 02 2018 at 06:01):you're welcome to, but it's a LOT harder to make it look aesthetically pleasing or have any organization in my experience -- that diagram you show wasn't modeled up the Z -- it would be quite hard to compose it that way
Sean (Aug 02 2018 at 06:01):if it doesn't look good, it won't be looked at, and if it's not looked at, it won't be useful
Sean (Aug 02 2018 at 06:02):on a horizontal plane, it becomes a landscape that is easily related with and will have optical effects that can be added to the scene like shadows, perspective, and environment mapping to make it look good
Jaipal Singh (Aug 02 2018 at 06:04):Aright! GG! :D
Sean (Aug 02 2018 at 06:04):not the best example, but http://write.flossmanuals.net/contributors-guide-to-brl-cad/feature-overview/static/sflake2_cc.png
Sean (Aug 02 2018 at 06:06):imagine all the primitives organized out in that checkerfield, casting nice shadows with ambient occlusion, maybe other effects
Sean (Aug 02 2018 at 06:07):boring af: https://img1.cgtrader.com/items/31910/2e26f3ba8d/deluxe-primitives-3d-model-low-poly-obj.png
Sean (Aug 02 2018 at 06:08):less boring: https://cdn.instructables.com/FZV/PQHS/IRXTUV78/FZVPQHSIRXTUV78.LARGE.jpg
Sean (Aug 02 2018 at 06:09):pretty cool: http://www.renderography.com/wp-content/uploads/unit_primitives_examples.jpg
Sean (Aug 02 2018 at 06:09):
Sean (Aug 02 2018 at 06:10):no good: http://brlcad.org/OLD/reports/arl-mr-103/images/figure_2.gif
Sean (Aug 02 2018 at 06:13):can work tiled if all shapes (or sets of shapes) roughly fill the same area, e.g., https://cmkt-image-prd.global.ssl.fastly.net/0.1.0/ps/13124/580/386/m1/fpnw/wm0/3d_geometric_shapes_1-.png?1365485103&s=83e651804c9eae187fd3246775155dec
Sean (Aug 02 2018 at 06:20):getting interesting: http://2.bp.blogspot.com/-rM0Yueq5t44/VXleslyHN4I/AAAAAAAABzc/56_gHGo2KT0/s1600/standard_primitives.jpg
Shubham Rathore (Aug 02 2018 at 06:32):@Jaipal Singh I hope all these examples would help you.... but don't forget to work on describe once you are done with the arrangements . Cheers !
Sean (Aug 03 2018 at 00:23):Specifically examples of 'vol'.
You mean like http://brlcad.org/wiki/VOL ?
Jaipal Singh (Aug 03 2018 at 03:58):So, the sketch primitive doesn't work and is sort of hardcoded in the current python API. I'm working on fixing it, shouldn't take long. I guess this could be a priority before we create the bigger scene.
Jaipal Singh (Aug 03 2018 at 03:59):Specifically examples of 'vol'.
You mean like http://brlcad.org/wiki/VOL ?
Returns a 404
Sean (Aug 03 2018 at 03:59):what's "doesn't work" mean?
Sean (Aug 03 2018 at 04:00):Returns a 404
woah, wiki disappeared
Jaipal Singh (Aug 03 2018 at 04:03):Alright, from what I've gathered whoever wrote this code earlier, didn't complete the implementation of the script primitive.
If you look at https://github.com/Killthebug/python-brlcad/blob/efe542615971cd7fd88b41e83fc4d8b646364925/brlcad/geometry.py#L297. This is what is invoked when the user wants to create a sketch, but this doesn't take any input from the use, rather just accepts a sketch object. If you look at other functions in the file, they all take the params required to create their respective primitives.
Jaipal Singh (Aug 03 2018 at 04:03):Instead, to create a primitive, this function uses the default values that have been hard coded in the Sketch class. https://github.com/Killthebug/python-brlcad/blob/efe542615971cd7fd88b41e83fc4d8b646364925/brlcad/primitives/sketch.py#L366
Jaipal Singh (Aug 03 2018 at 04:04):I'm now modifying geometry.py to be able to accept user input and create the sketch.
Sean (Aug 03 2018 at 04:08):wiki fixed
Jaipal Singh (Aug 03 2018 at 04:09):Thanks!
Jaipal Singh (Aug 03 2018 at 04:11):Alright, from what I've gathered whoever wrote this code earlier, didn't complete the implementation of the script primitive.
If you look at https://github.com/Killthebug/python-brlcad/blob/efe542615971cd7fd88b41e83fc4d8b646364925/brlcad/geometry.py#L297. This is what is invoked when the user wants to create a sketch, but this doesn't take any input from the use, rather just accepts a sketch object. If you look at other functions in the file, they all take the params required to create their respective primitives.
Either I'm not understand how to pass input to create a sketch object or this is genuinely wrongly implemented..
Sean (Aug 03 2018 at 04:12):Alright, from what I've gathered whoever wrote this code earlier, didn't complete the implementation of the script primitive.
If you look at https://github.com/Killthebug/python-brlcad/blob/efe542615971cd7fd88b41e83fc4d8b646364925/brlcad/geometry.py#L297. This is what is invoked when the user wants to create a sketch, but this doesn't take any input from the use, rather just accepts a sketch object. If you look at other functions in the file, they all take the params required to create their respective primitives.
that's not how I read it -- sketch is nothing like the other primitives, so it's perfectly normal that it's construction would be different
Sean (Aug 03 2018 at 04:12):it takes an arbitrary amount of data
Sean (Aug 03 2018 at 04:13):would need to see how primitives.Sketch is implemented, but it otherwise looks like it could be complete at a glance
Sean (Aug 03 2018 at 04:13):Instead, to create a primitive, this function uses the default values that have been hard coded in the Sketch class.
it only does that when the user didn't pass one
Sean (Aug 03 2018 at 04:17):take a look at an example sketch: http://brlcad.org/wiki/Sketch
Sean (Aug 03 2018 at 04:17):vertex list and a bunch of variadic sketch line types
Jaipal Singh (Aug 04 2018 at 06:32):In mged is it possible to save a certain POV? In the view option you have toggles for left, right, front, rear, etc. but is it possible to save a custom angle?
Sean (Aug 04 2018 at 07:38):yes, depending on what you mean by saving a view
Sean (Aug 04 2018 at 07:39):the 'ae' command will set an arbitrary azimuth/elevation (e.g., ae 35 25 or ae 45 65)
Sean (Aug 04 2018 at 07:40):the saveview command will save the current view information out to a file, which can be read back with the loadview command later
Sean (Aug 04 2018 at 07:41):there are several others too like setting based on a view direction or quaternion
Jaipal Singh (Aug 04 2018 at 14:07):the 'ae' command will set an arbitrary azimuth/elevation (e.g., ae 35 25 or ae 45 65)
Exactly what I was looking for!
Shubham Rathore (Aug 05 2018 at 15:34):@Jaipal Singh Hey what are you up to ? any leads with the sketch primitive ?
Jaipal Singh (Aug 06 2018 at 02:28):Hey @Shubham Rathore , I've been working on creating the scene with all the primitives. I've made multiple attempts in terms of arrangements, orientation, viewing angles and sizes. Finally I've been able to finalize on a particular design. I'm working on perfecting it and it should be completed in a few hours from now. Work on the sketch primitive has also progressed but it's not completed yet. I will shift focus onto it, once I'm done perfecting the scene.
Jaipal Singh (Aug 06 2018 at 04:07):I'm just stuck on getting the pipe made correctly, I have troubles figuring out the bend_radii.
Jaipal Singh (Aug 06 2018 at 04:49):Got it! xD
Shubham Rathore (Aug 06 2018 at 10:32):Just went through your report. The arrangement of the primitives is better than the previous ones. Regarding the sketch. Usually we use the gui but we can also create a sketch using put. This is what runs in the backend of the gui editor. All this shouldn't take much time ! Please try to wrap it up, we need to work on the describe as well plus the errors that we may encounter and the documentation part as well .
Shubham Rathore (Aug 07 2018 at 05:40):@Jaipal Singh are you aware of the final report submission with all the code links and the details of your work ? You need to submit this along with your final evaluation
Sean (Aug 07 2018 at 05:43):and you'll need to separate your code from the code that was already there -- e.g., you can identify all your commits and extract them as a big patch file or pull them into a branch and tag them
Sean (Aug 07 2018 at 05:44):you're basically required to submit all the code that you wrote, modified, removed, etc.
Jaipal Singh (Aug 07 2018 at 05:51):Sure! I could do that!
Jaipal Singh (Aug 07 2018 at 05:52):@Jaipal Singh are you aware of the final report submission with all the code links and the details of your work ? You need to submit this along with your final evaluation
Yea, I was notified via email yesterday.
Jaipal Singh (Aug 07 2018 at 05:55):Screen-Shot-2018-08-07-at-12.34.17-AM.png
Jaipal Singh (Aug 07 2018 at 05:55):This entire scene was created using :
https://github.com/Killthebug/python-brlcad/blob/master/Tutorial/scene.py
Jaipal Singh (Aug 07 2018 at 05:55):@Sean , @Shubham Rathore What do you think?
Sean (Aug 07 2018 at 06:08):That's looking much better!
Sean (Aug 07 2018 at 06:09):looks like a few issues, some primitives missing, some with undistinguished parameters, but looks like that'll set the stage (literally!)
Sean (Aug 07 2018 at 06:10):mind if I share that image?
Jaipal Singh (Aug 07 2018 at 06:22):mind if I share that image?
Not at all! :D
Jaipal Singh (Aug 07 2018 at 06:24):looks like a few issues, some primitives missing, some with undistinguished parameters, but looks like that'll set the stage (literally!)
Yes, I'm missing sketch and other 2-d derived primitives. Also, not included in this scene are superell, ebm, submodel and grip. Though examples on how to implement them have been written in https://github.com/Killthebug/python-brlcad/blob/master/Tutorial/core_tutorial.py
Jaipal Singh (Aug 07 2018 at 06:25):core_tutorial.py is meant to act as a reference file for creating primitives using python brlcad.
Jaipal Singh (Aug 07 2018 at 06:26):What about the undistinguished parameters, are you referring to the arb7? (orange in top row)
Jaipal Singh (Aug 07 2018 at 06:26):Even the bot (bottom row pink) could use a better orientation, IMO
Jaipal Singh (Aug 07 2018 at 06:30):Oh yea, an rpp is also missing from the scene! :(
Shubham Rathore (Aug 07 2018 at 06:43):@Sean , @Shubham Rathore What do you think?
Looks great indeed !
Shubham Rathore (Aug 07 2018 at 06:44):Yeah, need to account for the missing primitives :P
Jaipal Singh (Aug 07 2018 at 06:44):On it!
Sean (Aug 07 2018 at 07:20):What about the undistinguished parameters, are you referring to the arb7? (orange in top row)
arb7 is one of them, looks like an arb6. all of the arbs could use some rotation, like 20-50 degrees about Z so that we can see at least 2 sides of the arb4. green ehy(?) on second-from-top row looks like an arb6. green and red pointy cylinders -- (what are those?) look like rcc, too pointy. bot should be something more detailed (maybe teapot from the 'tea' command). and is the last red one in the corner an ars? something like the wiki example is needed, something more interesting than a wedge-looking shape.
Sean (Aug 07 2018 at 07:22):missing I think are dsp, superell, sketch, extrude, revolve, ebm, vol? any others?
Sean (Aug 07 2018 at 07:25):can just include the sample terrain from the 'terrain' tool for dsp, and wiki examples for sketch, extrude, ebm, and vol (maybe, it's nearly identical to ebm), just leaving superell and revolve and whatever else I missed from memory
Shubham Rathore (Aug 07 2018 at 12:29):missing I think are dsp, superell, sketch, extrude, revolve, ebm, vol? any others?
hrt !
Jaipal Singh (Aug 07 2018 at 12:30):Hey @Shubham Rathore , https://docs.google.com/spreadsheets/d/1XeeI1mpLlPFZtc7i5GeOmjLFE-s8JMsYpwVszIPsrmQ/edit?usp=sharing here's a list of primitives that are and are not a part of python brl-cad. IIRC, I had also shared this sheet ~10 days ago.
Jaipal Singh (Aug 07 2018 at 12:32):Please ignore the comments column in this sheet, it has not been updated
Jaipal Singh (Aug 07 2018 at 12:33):missing I think are dsp, superell, sketch, extrude, revolve, ebm, vol? any others?
hrt !
hrt is one of the primitives that I've not included. Few more are present in the sheet! :)
Shubham Rathore (Aug 08 2018 at 06:01):@Jaipal Singh Please update your daily logs
Jaipal Singh (Aug 08 2018 at 06:25):Yes!
Shubham Rathore (Aug 08 2018 at 09:41):And submit a patch with the describe function that you wrote... I'll try appending it to script
Jaipal Singh (Aug 08 2018 at 10:24):@Shubham Rathore https://hastebin.com/xiwohagonu.cpp . The file reading and writing out part works well, I was earlier having troubles saving and loading the filename in the import and export functions. Though all my attempts were before you had fixed it.
Jaipal Singh (Aug 08 2018 at 10:26):Also, for the last 2 days, I've been stuck on getting the sketch primitive to work, I'm able to create a script object with the points and vertices but am unable to get the curves created..
Jaipal Singh (Aug 08 2018 at 10:30):There's multiple ways in which the curves can be created as defined by the script primitive https://github.com/Killthebug/python-brlcad/blob/cd3ed94705933600f2a94f9e7ccde2e2ddd69497/brlcad/primitives/sketch.py#L408 . I
have figured out what the argument structure if for each of them, but I keep running into this error : https://hastebin.com/unapohiduq.coffeescript
Shubham Rathore (Aug 08 2018 at 10:35):I'll check in the evening and let you know !
Jaipal Singh (Aug 08 2018 at 10:36):Sure! Thanks :)
Jaipal Singh (Aug 08 2018 at 11:14):What about the undistinguished parameters, are you referring to the arb7? (orange in top row)
arb7 is one of them, looks like an arb6. all of the arbs could use some rotation, like 20-50 degrees about Z so that we can see at least 2 sides of the arb4. green ehy(?) on second-from-top row looks like an arb6. green and red pointy cylinders -- (what are those?) look like rcc, too pointy. bot should be something more detailed (maybe teapot from the 'tea' command). and is the last red one in the corner an ars? something like the wiki example is needed, something more interesting than a wedge-looking shape.
Rotating all the arb* to maximize viewing angle.
The green and red pointy cylinders are : rhc & rpc respectively. Inspired from the example at : (https://brlcad.org/gallery/diagrams/primitives)
Jaipal Singh (Aug 08 2018 at 21:39):Guys, I'm still stuck on the error that I had shared earlier ( https://hastebin.com/unapohiduq.coffeescript), could you please have a look and see if you can give any leads..
Cezar (Aug 08 2018 at 22:24):genptr_t is "generic pointer type", which appears to have been deprecated in 7.16, and replaced with void *
Cezar (Aug 08 2018 at 22:25):https://github.com/kanzure/python-brlcad/blob/master/brlcad/primitives/sketch.py#L473
Cezar (Aug 08 2018 at 22:26):c_void_p is what you're looking for, like on that line
Sean (Aug 09 2018 at 06:23):have figured out what the argument structure if for each of them, but I keep running into this error : https://hastebin.com/unapohiduq.coffeescript
@Jaipal Singh it looks like the ctypesgen interface is out of sync with the C code. there are no longer genptr_t types anywhere. they were all converted to void* ... not sure what that means for you on the python side, but it's something wrong in the python binding
Sean (Aug 09 2018 at 06:32):The green and red pointy cylinders are : rhc & rpc respectively. Inspired from the example at : (https://brlcad.org/gallery/diagrams/primitives)
That's not possible, rhc/rpc are the second row, green and fleshtone on the left (and the rhc is too pointy). I'd believe the gray and green next to it are epa and ehy, but again one of them is way too pointy -- looks like a trc. and if they are, that leaves a mystery for the red one in the 5th bottom row on the left.
Sean (Aug 09 2018 at 06:33):thanks @Cezar ! looks spot on
Shubham Rathore (Aug 09 2018 at 07:21):Guys, I'm still stuck on the error that I had shared earlier ( https://hastebin.com/unapohiduq.coffeescript), could you please have a look and see if you can give any leads..
Cezar stands correct.
Shubham Rathore (Aug 09 2018 at 07:22):@Shubham Rathore https://hastebin.com/xiwohagonu.cpp . The file reading and writing out part works well, I was earlier having troubles saving and loading the filename in the import and export functions. Though all my attempts were before you had fixed it.
This should work now. I'll give it a try this evening
Jaipal Singh (Aug 09 2018 at 16:16):
c_void_pis what you're looking for, like on that line
This seemed to have solved that error, but I've run into many more after it.
Jaipal Singh (Aug 09 2018 at 16:19):@Sean , @Shubham Rathore There are a lot of incomplete functions in the sketch primitive api and these are not pythonic errors, rather it's not been tested in the first place when it was created. There are no tests or examples for the sketch primitive and all the code is without any sort of comments for direction.
Jaipal Singh (Aug 09 2018 at 16:20):The issue arises in creating the curves in a sketch. There are a lot of helper functions involved that seem to be acting together. If I fix one, I break the other and the cycle seems to continue.
Jaipal Singh (Aug 09 2018 at 16:21):I'm going to devote one more day to else I will re-write the entire binding on my own. The current binding might also be working but I just can't wrap my head around it.
Jaipal Singh (Aug 09 2018 at 16:22):I will soon share the primitive scene with the requested changes and we can work on perfecting that. Along with it, I will try and get the describe function for the script primitive working because I had spent a fair amount of time on it as well earlier (before @Shubham Rathore had fixed it).
Jaipal Singh (Aug 09 2018 at 16:23):
genptr_tis "generic pointer type", which appears to have been deprecated in 7.16, and replaced withvoid *
@Cezar Thanks a ton for this. I knew something was deprecated but I didn't have a clue what it was. :)
Sean (Aug 09 2018 at 18:24):If you can make straight lines, how about just making an example using that, like a tessellated ‘S’ sketch with like 20 line segments? Then use that with extrude. Can leave a fixme in the code for curves.
Sean (Aug 09 2018 at 18:26):Otherwise sounds look or a plan.. just please leave yourself a day to write up a summary report with images to post somewhere, showcase everything you did blogpost style. :)
Shubham Rathore (Aug 10 2018 at 06:46):@Sean , @Shubham Rathore There are a lot of incomplete functions in the sketch primitive api and these are not pythonic errors, rather it's not been tested in the first place when it was created. There are no tests or examples for the sketch primitive and all the code is without any sort of comments for direction.
I know how it feels. Don't worry we'll fix it. Don't get stuck on this, move on with the reports and the documentation.
Jaipal Singh (Aug 10 2018 at 09:25):Screen-Shot-2018-08-10-at-2.54.28-PM.png
@Shubham Rathore The describe function works :)
Jaipal Singh (Aug 10 2018 at 09:31):BTW, some of the c-code that you're seeing that has been logged was present in the new.py file :P
Jaipal Singh (Aug 10 2018 at 09:32):There are a small hack and a minor change that I had to put in place. It doesn't break anything, but it has to be fixed in the export function. I'll share the diff soon and point out in that.
Shubham Rathore (Aug 10 2018 at 10:45):I dont exactly get it.... Will you please share the code, I'll go through it
Shubham Rathore (Aug 10 2018 at 11:04):What exactly is the issue that you are facing with the buffer (ep->extbytes) ?
Jaipal Singh (Aug 12 2018 at 18:40):@Shubham Rathore , sending patch.
Jaipal Singh (Aug 12 2018 at 18:40):@Sean , @Shubham Rathore Here's a basic blog post on how to work with python brlcad, do have a look and let me know what you think? :D
Jaipal Singh (Aug 12 2018 at 18:44):https://medium.com/@Mr_Jaypee/990e3c286a63
Jaipal Singh (Aug 13 2018 at 12:30): Jaipal Singh (Aug 13 2018 at 12:30):@Shubham Rathore ^ Here's the diff file.
Jaipal Singh (Aug 13 2018 at 13:07):Daily logs updated :)
Sean (Aug 13 2018 at 18:59):this is looking good @Jaipal Singh ! I like the overview of procedural geometry. would it make sense to say more about "Approach 2"?
you introduce it but then jump back to all the detail on Approach 1. reads a little confusing. I would suggest either explaining the downsides of Approach 2 and why Approach 1 was developed further, giving you the lead into the next section. also, suggest removing the three 'wdb' references (maybe just call them geometry) since we don't want to have to explain it. also, don't really need "brl_" prefix -- db works great and is easier to understand. it would be good to end the article with a mention of where there are more docs -- you mention the github repo, so maybe just say see comments in the code or point them to a specific file or something similar.
Sean (Aug 13 2018 at 18:59):minor point, but it would be good to update all the BRLCAD to BRL-CAD for consistency
Sean (Aug 13 2018 at 19:00):nice write-up regardless. did we give up on a straight-edge sketch/extrude/revolve?
Jaipal Singh (Aug 14 2018 at 03:24):Done!
Jaipal Singh (Aug 14 2018 at 03:25):nice write-up regardless. did we give up on a straight-edge sketch/extrude/revolve?
Yea, for now :/
Even the straight edge (line), depends on the curve class, which itself seems to be causing all the trouble
Shubham Rathore (Aug 14 2018 at 06:21):Awesome @Jaipal Singh !
Shubham Rathore (Aug 14 2018 at 06:22):Ill look into it by the evening !
Shubham Rathore (Aug 14 2018 at 17:40):Some issues with the buffer that need to be rectified..
Last updated: Nov 15 2024 at 00:49 UTC