Convert Collada (.DAE) animation files for Xcode Scenekit Animation use by removing unnecessary data.

ColladaAnimationForXcode.py

#!/usr/local/bin/python
# Jonathan Cardasis, 2018
#
# Cleans up a collada `dae` file removing all unnessasary data
# only leaving animations and bone structures behind.
# Combines multiple animation sequences into a single animation
# sequence for Xcode to use.
import sys
import os
import re
import subprocess

def print_usage(app_name):
  print 'Usage:'
  print '  {} [path(s) to collada file(s)...]'.format(app_name)
  print ''

def xml_is_collada(xml_string):
    return bool(re.search('(<COLLADA).*(>)', xml_string))

################
##    MAIN    ##
################
DAE_TAGS_TO_STRIP = ['library_geometries', 'library_materials', 'library_images']

if len(sys.argv) < 2:
    app_name = os.path.basename(sys.argv[0])
    print_usage(app_name)
    sys.exit(1)

print 'Stripping collada files of non-animation essential features...'
failed_file_conversions = 0

for file_path in sys.argv[1:]:
    try:
        print 'Stripping {} ...'.format(file_path)
        dae_filename = os.path.basename(file_path)
        renamed_dae_path = file_path + '.old'

        dae = open(file_path, 'r')
        xml_string = dae.read().strip()
        dae.close()

        # Ensure is a collada file
        if not xml_is_collada(xml_string):
            raise Exception('Not a proper Collada file.')

        # Strip tags
        for tag in DAE_TAGS_TO_STRIP:
            xml_string = re.sub('(?:<{tag}>)([\s\S]+?)(?:</{tag}>)'.format(tag=tag), '', xml_string)

        # Combine animation keys into single key:
        #  1. Remove all <animation> tags.
        #  2. Add leading and trailing <library_animation> tags with single <animation> tag between.
        xml_string = re.sub(r'\s*(<animation[^>]*>)\s*', '\n', xml_string)
        xml_string = re.sub(r'\s*(<\/animation\s*>.*)\s*', '', xml_string)

        xml_string = re.sub(r'\s*(<library_animations>)\s*', '<library_animations>\n<animation>\n', xml_string)
        xml_string = re.sub(r'\s*(<\/library_animations>)\s*', '\n</animation>\n</library_animations>', xml_string)

        # Rename original and dump xml to previous file location
        os.rename(file_path, renamed_dae_path)
        with open(file_path, 'w') as new_dae:
            new_dae.write(xml_string)
            print 'Finished processing {}. Old file can be found at {}.\n'.format(file_path, renamed_dae_path)
    except Exception as e:
        print '[!] Failed to correctly parse {}: {}'.format(file_path, e)
        failed_file_conversions += 1

if failed_file_conversions > 0:
    print '\nFailed {} conversion(s).'.format(failed_file_conversions)
    sys.exit(1)

I've noticed that THREE.js will have issues with animation nodes with with no frames (will load with NaNs which fucks up the renderer). If anyone else is having this issue, the js code below might help

//Remove this entire node if there's no animation data
//(as otherwise NaN's will occur). Node is an <animation> node
if(node.querySelector("float_array[count='0']")) {
	node.remove();
	return;
}

Is this to achieve smaller file size?
I arrived here from stack overflow - but I'm wanting to get clarity around if I have the (mixamo samba style) dances in FBX
can I some how import this animation (500kb) and have the usdz or dae model apply the fbx dance?
or do I need to spit out a new dae file with the skeleton /body / mesh ? ~ 20mb or can I import / export it? then apply it? (https://www.reddit.com/r/gamedev/comments/4jb68g/anyone_still_got_the_whole_mixamo_pack/ - talking about the giant animations 4.5gb here)