ons 2024-08-21 klockan 23:19 +0200 skrev Marton Balint:
> 
> 
> On Wed, 21 Aug 2024, Marc-Antoine ARNAUD wrote:
> 
> > Le sam. 17 août 2024 à 21:15, Marton Balint <c...@passwd.hu> a écrit
> > :
> > 
> > > 
> > > 
> > > On Wed, 14 Aug 2024, Marc-Antoine Arnaud wrote:
> > > 
> > > > ---
> > > > libavformat/mxfdec.c | 14 +++++++++++++-
> > > > 1 file changed, 13 insertions(+), 1 deletion(-)
> > > 
> > > Why would you want to tune this?
> > > 
> > 
> > Hi Marton
> > 
> > In some MXF files, index tables are present in the first body
> > partition or
> > fully in the footer partition.
> > This patch will skip parsing all body partitions if the index table
> > is
> > coherent.
> 
> That still does not answer the "why". You want to avoid reading every
> partition pack when the file is opened? If you ignore reading the RIP
> (and consequentally force the system to do a forward parse and not
> read all 
> partition packs in advance) then I am not sure how seeking will work,
> because even if there is a full index, you won't have a list of
> partitions 
> and the essence body offsets for the partitions.
> 
> The quick and dirty way of forcing opening a file without reading all
> partition packs is using -seekable 0 when opening that file... Isn't 
> that enough for your use case?
> 
> Something more advanced is not trivial, because you'd have to parse
> the 
> partitions packs on demand.

A feature like this would be really neat. You'd need logic that
guesstimates which partition pack to parse, sees if the desired edit
unit is contained within it and optionally seeks forward/backward as
necessary to find the appropriate partition. That's quite complicated
to implement I suspect (you'd need a sparse range-aware data
structure), but it'd be the best way. That would also take care of some
nasty properties of the current demuxer, which is terribly slow to open
MXF files produced by mxfenc if working over HTTP. It gets even worse
with melt that opens files multiple times

/Tomas
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".

Reply via email to