1

Over on this question I found out about io_blend_utils/blend/blendfile.py, which is a one-file-module to read a blender file without actually loading it into blender. I need to read the elements of a CollectionProperty, which I am currently only able to do for the first element.

I fear this is a very specific and advanced question. I am hoping @ideasman42 may be able to help?

My use case

I need to crawl through a bunch of .blend files to find all the texture atlases (created with the uv_texture_atlas plugin) contained in them. Since there are many and large .blend files, I do not want to have the overhead of loading all the files into blender.

My code so far

    def group_iterator(block):
        """
        Iterate over a IDPropertyGroup BlendFileBlock.
        :param block: block to iterate over
        :return: All IDProperty blocks in the group or list
        """
        i = block.get_pointer((b'data', b'group', b'first'))
        while i is not None:
            yield i
            i = i.get_pointer(b'next')

    atlases = collections.OrderedDict()
    for blend_file_path in blends:
        with blendfile.open_blend(blend_file_path) as blend:
            scenes = [block for block in blend.blocks if block.code == b'SC']

            for scene in scenes:
                # get custom properties
                properties = scene.get_pointer((b'id', b'properties'))

                if properties is None:
                    continue

                # iterate through all the property groups
                for itor in group_iterator(properties):
                    if itor.get(b'name') == "ms_lightmap_groups":
                        group = itor.get_pointer((b'data', b'pointer'))

                        if group is not None:
                            # iterate over all the properties
                            for prop in group_iterator(group):
                                name = prop.get(b'name')

                                if name == 'name':
                                    offset = prop.get((b'data', b'pointer'))
                                    length = prop.get(b'len')
                                    blk = blend.block_from_offset.get(offset)
                                    blend.handle.seek(blk.file_offset, os.SEEK_SET)
                                    atlas_name = blendfile.DNA_IO.read_string0(blend.handle, length)
                                elif name == 'resolutionX':
                                    resX = prop.get((b'data', b'val'))
                                elif name == 'resolutionY':
                                    resY = prop.get((b'data', b'val'))

                            atlases[atlas_name] = (2**(resX+8), 2**(resY+8))
                            print("Found atlas %s" % atlas_name)

Thank you so much in advance (I have been trying to figure this out for over a day now)!

Regards, Jonathan.

Community
  • 1
Squareys
  • 41
  • 6
  • So, I hear @ideasman42 left blender. In this case I will figure this out myself and post an aswer, once I found out. Also, there is #blenderpython IRC, I will post my question there, too. – Squareys Sep 05 '16 at 11:23

1 Answers1

1

blendfile.py actually supports getting elements of a collection property with some super neat and simple API!

    for blend_file_path in blends:
        with blendfile.open_blend(blend_file_path) as blend:
            scenes = [block for block in blend.blocks if block.code == b'SC']

            for scene in scenes:
                # get custom properties
                properties = scene.get_pointer((b'id', b'properties'))

                if properties is None:
                    continue

                # iterate through all the property groups
                itor = properties.get_pointer((b'data', b'group', b'first'))
                while itor is not None:
                    if itor.get(b'name') == "ms_lightmap_groups":
                        length = itor.get(b'len')

                        for i in range(0, length):
                            group = itor.get_pointer((b'data', b'pointer'))

                            if group is not None:
                                # iterate over all the properties
                                # (!) i specifies the index in the collection
                                prop = group.get_pointer((b'data', b'group', b'first'), base_index=i)  
                                while prop is not None:
                                    name = prop.get(b'name')

                                    if name == 'name':
                                        offset = prop.get((b'data', b'pointer'))
                                        length = prop.get(b'len')
                                        blk = blend.block_from_offset.get(offset)
                                        blend.handle.seek(blk.file_offset, os.SEEK_SET)
                                        atlas_name = blendfile.DNA_IO.read_string0(blend.handle, length)
                                    elif name == 'resolutionX':
                                        resX = prop.get((b'data', b'val'))
                                    elif name == 'resolutionY':
                                        resY = prop.get((b'data', b'val'))

                                    prop = prop.get(b'next')

                                atlases[atlas_name] = (2**(resX+8), 2**(resY+8))
                                print("Found atlas %s" % atlas_name)

                    properties_itor = properties_itor.get_pointer(b'next')

BlendFileBlock.get(...) and BlendFileBlock.get_pointer(...) both have an optional base_index parameter which is gets you exactly this.

May this be of use to someone else out there :)

Happy blending! :D

Squareys
  • 41
  • 6