The current method of scanning a Real file for metadata
iterates through all the objects/chunks of the file.
Despite the fact that the method parses only PROP,
MDPR, and CONT objects, all objects are read from the
file, including the irrelevant and huge DATA objects
containing video and audio. This is inefficient,
especially when Real files are scanned over an NFS or
Samba mount.
This patch fixes the situation by reading (file.read)
only PROP, MDPR and CONT objects and skipping
(file.seek) all other objects which aren't parsed
anyway. For 30 MB .rm video files mounted over 100Mbps
LAN on my machine the method is around 100 times
faster. The patch is against the 0.4.9 version of mmpython.
The patch for video/realinfo.py version 0.4.9