gcv API · brlcad · Zulip Chat Archive

Stream: brlcad

Topic: gcv API

starseeker (Sep 10 2020 at 12:03):

Sean said:

starseeker so an issue you might be able to help sort out... ran into a problem creating an image-to-geometry plugin for libgcv that hadn't been fully considered when the mime stuff was put in place. gcv's interface technically only handles "bu_mime_model" types or "unknown"/"auto". that type/grouping limitation is technically wrong, and is compounded considering one might want to create geometry off of other types (e.g., excel spreadsheet or something).

That doesn't surprise me, actually - I think the mime stuff was me, and I'm pretty sure I didn't consider a non-model to model mapping when I set it up.

starseeker (Sep 10 2020 at 12:07):

If we want an anything to anything mapping, probably the thing to do is to not use the category typing at all, but just assemble a more generalized mapping setup... hmm...

Sean (Sep 10 2020 at 13:06):

@starseeker you mean like instead of using enums, doing something dictionary-based?

could see something string-based work, maybe where we just stash all the mime.types mappings into a std::map<std::string, std::list<std::string> internally and expose that through a search/lookup call using string matching. completely alternatively, could stick with the enums, but just have on honking enum list all of the same type. hm.

starseeker (Sep 10 2020 at 14:29):

@Sean I made a quick stab at generalizing in r77099

starseeker (Sep 10 2020 at 14:30):

Shouldn't be merged to RELEASE, but conveys the idea

Sean (Sep 10 2020 at 14:30):

yeah, this isn't needed for release, just came up yesterday helping set up an example plugin

starseeker (Sep 10 2020 at 14:31):

That's a real quick run at it - having to do paperwork fun today and probably tomorrow, so couldn't test more extensively.

starseeker (Sep 10 2020 at 14:32):

On the plus side, I doubt I'll be stacking up too many more commits for you to review ;-)

Sean (Sep 10 2020 at 14:32):

no worries, I'm waist-deep in training checking off the last remaining commits in between the reboots and slow dialog

starseeker (Sep 10 2020 at 14:35):

/me may finally have found the threshold for "committing too often"

Sean (Sep 10 2020 at 14:37):

nah, it's just our/my review methods need to adapt better to higher frequency and bursts

Sean (Sep 10 2020 at 14:38):

95% of commits are trivially reviewed and can be checked off. really need something (other than inbox) that will help keep track of just that 5% that beg closer inspection