[FFmpeg-devel] [RFC] ffprobe -sep option

Stefano Sabatini stefano.sabatini-lala at poste.it
Sun Aug 21 13:31:36 CEST 2011


On date Saturday 2011-08-20 20:52:03 +0200, Clément Bœsch encoded:
> On Sat, Aug 20, 2011 at 08:40:24PM +0200, Stefano Sabatini wrote:
> > On date Friday 2011-08-19 16:19:40 +0200, Nicolas George encoded:
> > > Le decadi 30 thermidor, an CCXIX, Stefano Sabatini a écrit :
> > > > which is sometimes easier to parse of the corresponding
> > > > one-line-per-field output.
> > > 
> > > Very good news, thanks.
> > > 
> > > > [PACKET],codec_type=video,stream_index=0,pts=137250,pts_time=1.525000,dts=137250,dts_time=1.525000,duration=3750,duration_time=0.041667,size=16.000000,pos=1880,flags=_,[/PACKET]
> > > > which is not very nice (so should I add a separate option for that?).
> > > 
> > > This kind of output will probably be mostly piped to cut/awk/perl, so nice
> > > is not really necessary. No need to overthink it.
> > 
> > Looking forward for Clement patches which provide a more generic
> > solution and conflict with this one, I'll possibly base an updated
> > patch on his work.
> 
> The current solution I proposed is quite hackish. I'm looking for some
> something better; maybe some "writers" would be more appropriate (a
> default writer for the current output, a JSON writer, a "compact" writer,
> etc.). Unfortunately, it will certainly mean multiplying by 2 the size of
> the current ffprobe code.
> 
> Are you OK with that?

I'm not against if the code is not going to be too convoluted. Also we
may need output text drivers for other purposes (e.g. for printing
statistics regarding filtered frames, e.g. in showinfo/ashowinfo), so
maybe we can manage to push them to the lib soon or later.
-- 
FFmpeg = Foolish & Freak Mastering Ponderous Elastic Generator


More information about the ffmpeg-devel mailing list