@isolin That sounds painful! I’m currently(in first 4 weeks) not working on importer, so cannot comment on what the new one would look like, but definitely not this bad. (see next comment for the hopeful preliminary results.)
@testure Yes, I had seen the thread when I was going over the project list, some months back. I’m taking care not to make those mistakes .
Thank you @_name for telling me about it!
Reminds me to be careful about memory allocations & function overhead.
Received a lot of input from the mentors regarding coding style, comments, documentation, and file structures. Spent quite some time on cleaning the hastily committed rB485cc433. See current version at Diffusion
Currently the exporter exports vertices, face normals, and face indices; in world coordinates.
It’s lacking in (1) multiple objects, (2) texture coordinates (3) Axes transformation. If you compare obj txt files written with python & C++ , you’ll see differences in vertex order, and some coordinates with opposite sign. But the shape is correct.
Did a comparison between the old & new exporters. Release configuration, lite build + opensubdiv, default cube with 6+3 subsurf modifiers. See images: F8586987F8586984
wm_handler_fileselect_do takes 82 seconds (python) vs 9 seconds (c++) (which is expected to be less than 14-15 s with texture coordinates).
Previous years we had 2 threads, one for reporting so it’s easy to see what was going on, and one with lively back and forth between users and student where a weekly report easily would have gotten lost in the noise.
A build to test should be available (thanks to Lazydodo!) after these changes are merged in soc-2020-io-performance.
The code upto rB86845ea907f8ae24b3bed4ab591f8e3687ac5f22 crashes. So please don’t try that.
Wow, those export times looks amazing!
FINALLY! Quite a few times I was embarrassed to first show people how great Blender can be and then having to show them how to import assets into Blender just to sit there in awkward silence for 5 min.
Please test the build!
Thank you @lazydodo
f01a7d508d0d is the latest hash as of now.
The exporter is in the expected place, File > export > OBJ new. It only exports the single selected object as of now. And that should be mesh only. Not curves etc.
Also, if you run blender by console, you can see the export timings too.
Let me know about the bugs! Best is to share the steps to make a problematic mesh. If sharing the file is a must, please minimise its size.
The OBJ exporter now exports Texture Coordinates. rBf01a7d50. Not the MTL file yet. It’s a goal of Week 3.
Also see the table of comparison in #15
Since fstream is not adding any benefits and is slower too, it was removed.
It also exports multiple objects in the same file. But so far only OB_MESH object types are supported. rB40736795.
Multiple frames can also be exported now, in individual files. rB6d088bdd.
Progress is now shown on per-file basis in Terminal and timing in the end. It should be improved to act on per-object basis when there is a single frame.
This is awesome! Thanks for your work on this Ankit, it’s really appreciated. Obj export got 10 times faster. Still using OBJs because of zbrush and it’s not gonna change any time soon, so this speedup is going to be a huge time saver.
Hey all! More features and less memory usage this week!
Week 3 Report: June 15 to 19
Animation: Allow all frames to be exported, instead of 0-1000. Also filenames are also edited with frame number only if there are multiple frames. rB45461012
Add support for curves to be exported after converting them to meshes.
Refactor the code upto now to object-oriented style. D8089
This allowed us to have minimal memory overhead with no timing cost. UV vertices, however, have to be allocated on heap and thus show a peak while exporting.
Unfortunately, I couldn’t finish some goals at time: Material library, NURBS and grouping settings. A quick thing to add is export selected objects only.
So the goal for next week is to finish them as soon as possible and start with importer.
Something that’s mildly annoying is obj’s inheriting that 90 degree rotation as they import (due to Z pointing up in blender I guess) Would it be possible to apply those transforms in the background as importing/exporting? Not sure if it would mess with peoples pipelines, for rigging it can be painful at times.
Also, something that’s probably out of the scope of this project but I personally do constantly in many packages (with the use of addons/plugins mostly) is to optionally export selected meshes as separate objs based on their mesh names, or import multiple objs at the same time with the naming being based on their file names.
I know it’s not on the roadmap and probably falls under a different category as I’ve seen plans for drag and drop and automatically detect the format when importing, but I figured I’d throw it in here.
I know the difference in the axes. The python exporter rotates the object about x-axis on its default settings. The cpp one doesn’t. It exports the object the way it looks in the viewport.
I don’t know why the rotation is there. I will see what other addons do or ask someone why they do it.
It’s a very minor change though.
I will expose the “select objects only” soon. More features on top of the goal of making the cpp one drop-in replacement of the older one should be handled when I’m not at a risk of failing. (:
Drag and drop is bound to the File Browser and the way operators work. (or something to that effect as far as I remember ) I think UI (and data, assets & IO) team should add the feature and the importers will make changes in their own code to make it work.
Cleanup and review update in the OOP refactor: rB0f383a3d.
Support NURBS being exported in parameter form, optionally. So far, they were exported as meshes. rB31d48017.
Support NURBS surfaces to be exported as a regular mesh. rB31d48017.
Broke down files to create new ones for OBJMesh and OBJNurbs . rB6c98925d Later on, MTLWriter will also have its own implementation file.
Support material library export. We decided against using the existing python writer, since there were no benefits of using it except saving some time now. To extend support, one would have to edit Node Shader Utils file first. Also debugging it would be a trouble. Since nodes are already in C/C++, we have more control there, instead of in the API.
Export of single material per object: rB827869a4. Support for multiple materials in one object: rBfaa11ec0
Show red colored warning if overriding an existing file while exporting (it was a wrong flag) rBc39128ca
Fix socket vector being read as socket RGBA rB7685d9e4
Remaining tasks and Week 5 todo
Smooth groups (and bitflags) & Vertex groups
Start with importer. I’m already lagging here User:Ankitm/GSoC_2020_Proposal_IO_Perf#Project_Schedule. But since the refactor I had planned for week 7 (along with profiling) won’t take long, I feel that I have a good buffer to work in.
Setup clang tidy with Ninja. On macOS, had to build llvm + clang 11 + tidy to get the executable. Also setup a build with clang 11 + leak sanitiser since Xcode’s asan doesn’t provide the functionality.
Import a mesh successfully in the viewport. And add its object in a collection. It took a long time, trying to avoid Mesh, BMesh asserts, seg-faults and memory leaks due to un-freed meshes.
– The mesh has vertex data only, so far. Texture coordinates are to be added.
– The parser takes a long time. ~70% of the import process. Making a Blender object from the parsed data is the rest ~30%. So need to optimise it.