[FFmpeg-devel] [patch] allow wordexp globs in image2 file sequence import

[Brian Olson]

On Jan 1, 2011, at 10:28 PM, Michael Niedermayer wrote:

> On Fri, Dec 31, 2010 at 12:11:26PM -0500, Brian Olson wrote:
>> It seemed cumbersome to me to have to symlink/rename my image sequence files to fit
>> ffmpeg -i prefix%05d.jpg
>> so I hacked libavformat/img2.c to use <wordexp.h> file globbing, and now I can
>> ffmpeg -i '*.jpg'
>> (and other shell-expnsion patterns like ? * [] {})
>> 
>> This is my first attempt at submitting to ffmpeg. I hope you like it. Feedback welcome. Hopefully I didn't break the conventions too much.
>> 
>> Brian Olson
>> http://bolson.org/
>> 
> 
>> img2.c |   63 +++++++++++++++++++++++++++++++++++++++++++++++++++++++--------
>> 1 file changed, 55 insertions(+), 8 deletions(-)
>> 27ac4fb6e59175875e0fc3ef0c46f21e3bdfe196  img2_wordexp.diff
> 
> is there an advantage in this over using the concat protocol?


My reading of the concat: protocol is that it would require all the files to be listed sequentially on the command line, and given enough frames that will simply be impossible. (I think I heard linux exec() has a limit of 64kb of effective command line.) Even if possible it would be inconvenient to type out all those filenames, or write a script to build the command. Stepping through the files internally seems to be the way to go. ffmpeg already can step through a very limited kind of file pattern internally; I think this way is better. It certainly makes my common usage case a lot easier.

Brian Olson
http://bolson.org/