Adding an option to dump images of trailers and headers of each part for debugging purpose.

removeads.py

@@ -359,7 +359,9 @@ def extractPictures(inputFile, begin, nbFrames, width=640, height=480):
     images = bytes()
     with Popen(['ffmpeg', '-loglevel', 'quiet' ,'-y', '-ss', '%s'%begin, '-i', '/proc/self/fd/%d' % infd, '-s', '%dx%d'%(width, height), '-vframes', '%d'%nbFrames, '-c:v', 'ppm', '-f', 'image2pipe', '/proc/self/fd/%d' % fdw ], stdout=PIPE, close_fds=False) as ffmpeg:
         while ffmpeg.poll() == None:
-            fds, _, _ = select([fdr, ffmpeg.stdout], [], [], .1)
+            # TODO: understand why this line ends up in reading on an already closed file descriptor
+            # fds, _, _ = select([fdr, ffmpeg.stdout], [], [], .1)
+            fds, _, _ = select([fdr], [], [], .1)
             if fdr in fds:
                 buf = os.read(fdr, 1000000)
                 # print("Read %d bytes of image. ffmpeg finished: %s" % (len(buf), ffmpeg.poll()))
@@ -415,10 +417,46 @@ def extractSound(inputFile, begin, outputFile, subChannel=0, nbPackets=0, sample
     if status != 0:
             logger.error('Sound extraction returns error code: %d' % status)
 
-def dumpPPM(pictures, prefix):
+def dumpPPM(pictures, prefix, temporaries):
+    logger = logging.getLogger(__name__)
+    
     # "P6\nWIDTH HEIGHT\n255\n"    
-    # headerLen=2+1+ceil(log(width, 10))+1+ceil(log(height, 10))+1+3+1
-    pass
+    pos = 0
+    picture = 0
+    while pos<len(pictures):
+        filename = '%s-%03d.ppm' % (prefix, picture)
+        header = BytesIO(pictures[pos:])
+        magic = header.readline().decode('utf8')
+        dimensions = header.readline().decode('utf8')
+        maxvalue = header.readline().decode('utf8')
+        if magic == 'P6\n':
+            pattern = re.compile('^(?P<width>[0-9]+) (?P<height>[0-9]+)\n$') 
+            m = pattern.match(dimensions)
+            if m != None:
+                width = int(m['width'])
+                height = int(m['height'])
+            else:
+                logger.error('Impossible to parse dimensions of picture')
+                return
+        else:
+            logger.error('Not a PPM picture')
+            return 
+        
+        headerLen=2+1+ceil(log(width, 10))+1+ceil(log(height, 10))+1+3+1
+        try:
+            out = open(filename, 'w')
+            outfd = out.fileno()
+        except IOError:
+            logger.error('Impossible to create file: %s' % filename)
+        temporaries.append(out)
+        
+        length=headerLen+3*width*height
+        nbBytes = 0
+        while nbBytes < length:
+            nbBytes+=os.write(outfd, pictures[pos+nbBytes:pos+length])
+        pos+=length
+        picture+=1
+        
         
 def extractAllStreams(inputFile, begin, end, streams, filesPrefix, nbFrames, width, height, temporaries, dumpPictures=False):
     logger = logging.getLogger(__name__)
@@ -460,7 +498,7 @@ def extractAllStreams(inputFile, begin, end, streams, filesPrefix, nbFrames, wid
                 codec = stream['codec_name']
                 imagesBytes = extractPictures(inputFile=inputFile, begin=begin, nbFrames=nbFrames, width=width, height=height)
                 if dumpPictures:
-                    dumpPPM(imagesBytes, '%s-%d' % (filesPrefix,videoID))
+                    dumpPPM(imagesBytes, '%s-%d' % (filesPrefix,videoID), temporaries)
                 
                 # imagesBytes contains now a buffer of bytes that represents the pictures that have been dumped by ffmpeg.
                 fdr, fdw = os.pipe()
@@ -634,7 +672,7 @@ def main():
     parser.add_argument("-o", "--output", dest='outputFile', type=str, required=True, help="Output MKV file to produce.")
     parser.add_argument("-p", "--part", dest='parts', nargs='+', required=False, action='append', metavar="hh:mm:ss[.mmm]-hh:mm:ss[.mmm]", help="Extract this exact part of the original file.")
     parser.add_argument("-k", "--keep", action='store_true', help="Do not cleanup temporary files after processing.")
-    parser.add_argument("--dump-pictures", action='store_true', help="For debug purpose, dump pictures of headers (and trailers) before (after) each part. They are kept in memory only otherwise.")  
+    parser.add_argument("--dump-pictures", action='store_true', dest='dump', help="For debug purpose, dump pictures of headers (and trailers) before (after) each part. They are kept in memory only otherwise.")  
   
     args = parser.parse_args()
     logger.debug("Arguments: %s" % args)
@@ -782,7 +820,7 @@ def main():
         
         if nbHeadFrames > 0:
             # We extract all frames between the beginning upto the frame that immediately preceeds the I-frame.
-            head = extractAllStreams(inputFile=mkv, begin=ts1, end=headIFrameTS, nbFrames=nbHeadFrames-1, filesPrefix='part-%d-head' % (partnum), streams=streams, width=width, height=height, temporaries=temporaries)
+            head = extractAllStreams(inputFile=mkv, begin=ts1, end=headIFrameTS, nbFrames=nbHeadFrames-1, filesPrefix='part-%d-head' % (partnum), streams=streams, width=width, height=height, temporaries=temporaries, dumpPictures=args.dump)
             subparts.append(head)
         
         # Creating MKV file that corresponds to current part between I-frames
@@ -797,7 +835,7 @@ def main():
         
         if nbTailFrames > 0:
             # We extract all frames between the I-frame (including it) upto the end.
-            tail = extractAllStreams(inputFile=mkv, begin=tailIFrameTS, end=ts2, nbFrames=nbTailFrames, filesPrefix='part-%d-tail' % (partnum), streams=streams, width=width, height=height, temporaries=temporaries)
+            tail = extractAllStreams(inputFile=mkv, begin=tailIFrameTS, end=ts2, nbFrames=nbTailFrames, filesPrefix='part-%d-tail' % (partnum), streams=streams, width=width, height=height, temporaries=temporaries, dumpPictures=args.dump)
             subparts.append(tail)
         
         logger.info('Merging: %s' % subparts)