[mythtv] [RFC] AvFormatDecoder Improvements.

[Isaac Richards]

On Wednesday 19 November 2003 08:39 am, Doug Larrick wrote:
> All right, the problems with ff/rew in HDTV are really starting to bug
> me, so I'm gonna work on this part, using frame number as the index in
> positionMap.  I'll add a new marktype (in recordedmarkup) for the by-
> frame positionmap.  I'll use a (sorted) QValueVector in avformatdecoder
> so that we can binary-search by either frame number or position.  This
> approach (rather than the hashing of a QMap) also allows finding the
> nearest entry on a miss, and easily stepping forward/backward by one
> keyframe.
>
> I'm not planning to touch the current features that allow seeking w/o a
> positionMap, so my changes should not preclude future work in that
> area.

I wouldn't be adverse to having a database update that simply changes the 
existing markup type to be frame numbers instead of keyframe numbers.  Should 
be pretty easy to handle the upgrade (the code in dbcheck would multiply the 
existing numbers in the table by 15 or 12 depending on NTSC or PAL), then 
change things to use your vector for seeking.  Wouldn't be two sets of 
seeking support code munging things up that way, either.

Isaac