jdiez17

class GIFProcessor(Processor):
     def sync(self, file):
         mp4(file) 
         webm(file)
         ogv(file)
         
     time = 300

jdiez17

class MP4Processor(Processor):
    sync = [validate_mp4, webm, ogv]
    async = [png_frame]
    time = 300

    def resolution(self, file):
        # Some computation 
        file.resolution = 1234

    def audiostreams(self, file):

jdiez17

In your snippet, sync and async are functions. In mine, they are arrays of functions which can raise exceptions. The idea behind this is to remove code duplication. If sync() was a function, it would look something like this (https://gist.github.com/jdiez17/7760198).

Even though the idea of the functions in `sync` is to operate on files (in the disk), it doesn't mean that they're designed "around the idea that input file is converted to output files". A function that would validate GIFs/MP4s/whatever would a call to a command whose output is based on a file in the disk.

A major disadvantage of passing the Redis file object to all the functions is that each function would have to figure out the path in the file system, which leads to more clutter and code duplication than in the gist I linked.

For operations that require storing information on the database, passing the file path to the function is not enough: that's the use for the `extra` function (probably not the best name). Hell, for all I care, I could

jdiez17

from mediacrush.processing.invocation import Invocation

def detect():
    a = Invocation("ls -la")
    a.run()
    print a.stdout # blah
    print a.returncode # 0
    print a.exited # False (True if it times out)

jdiez17

P1: A => B ^ C

P2: B => ¬D

P3: C => D

P4: (added by Proof by Contradiction) A

Conclusion: false (proof by contradiction)

jdiez17

{
  "blob_type": "video", 
  "compression": 9.22, 
  "extras": [
    {
      "file": "/TayC7vkb31F5.png", 
      "type": "image/png"
    }
  ], 
  "files": [

jdiez17

from mediacrush.objects import File
from mediacrush.database import r, _k
from mediacrush.processing.detect import detect
from mediacrush.mimetypes import get_mimetype, extension
from mediacrush.files import file_storage

if __name__ == '__main__':
    for h in r.smembers(_k("file")):
        f = File.from_hash(h)
        mimetype = get_mimetype(f.original)

jdiez17

from mediacrush.objects import File
from mediacrush.database import r, _k
from mediacrush.processing.detect import detect
from mediacrush.mimetypes import get_mimetype, extension
from mediacrush.files import file_storage

if __name__ == '__main__':
    for h in r.smembers(_k("file")):
        f = File.from_hash(h)

jdiez17

We're currently working on getting support for VTT, SRT, and ASS subtitles on MediaCrush, but we're not
done. If we finish it up, we'll revisit this post and describe how it's done.
You'll need to discover and extract subtitles and fonts, as well as provide a client-side subtitle
rendering solution like [libjass](https://github.com/Arnavion/libjass) or [WebVTT](https://wiki.mozilla.org/WebVTT).

jdiez17

flags = {
  'video': BitVector('a', 'b', 'c')
  'image': BitVector('d', 'e')
}

----

def process_file:
  processor, flags = detect()