[mythtv-commits] Ticket #6889: New commercial detection method to flag Pre and post roll only

[MythTV]

#6889: New commercial detection method to flag Pre and post roll only
--------------------------------------+-------------------------------------
 Reporter:  Paul <Paul at formagio.net>  |        Owner:  cpinkham
     Type:  enhancement               |       Status:  assigned
 Priority:  minor                     |    Milestone:  0.22    
Component:  MythTV - General          |      Version:  head    
 Severity:  low                       |   Resolution:          
  Mlocked:  0                         |  
--------------------------------------+-------------------------------------

Comment(by Paul <Paul at formagio.net>):

 Replying to [comment:2 cpinkham]:
 > I don't see where you are waiting until after the pre-roll to being
 flagging.  In ::go(), requiredHeadStart is set to requiredBuffer which is
 set to 120, you never add in preRoll.
 >


 I think it is taking the preroll into account by using the using the
 startedAt time instead of the recordingStartedAt time.

 > Can you wrap your long lines at 80 columns by splitting strings and
 putting function or QString args on the next line, etc.?
 >
 > Myth's coding standard says that opening brackets for blocks of code go
 on the next line, not on the if/while/for/function line, please correct
 these as well.
 >
 > Duplicated code here:  (COMM_FRAME_SCENE_CHANGE | COMM_FRAME_BLANK |
 COMM_FRAME_SCENE_CHANGE)
 >
 > Please include a patch to libs/libmyth/programinfo.cpp to add this
 detection method to the SkipTypeToString() and
 GetPreferredSkipTypeCombinations() functions there so that the user would
 not have to edit the database to change the default or channel-specific
 flagging methods.
 >

 These should all be addressed by the new patch.

 > How accurate is this and what kind of pre/post roll did you use when
 testing.  It appears that you are only looking for the closest blank frame
 or scene change frame to the pre/post roll times.
 >
 > If you can correct these issues in an updated patch, I should be able to
 get this in before the feature freeze on Monday.

 On my system i generally use a 5 minute preroll and a 1 minute post roll.
 In my experience the pre-roll accuracy is pretty good. However, the
 postroll accuracy leaves a lot to be desired. My theory at this point is
 that shows generally end before the scheduled times to allow time for
 commercials or previews from the station. You are correct that i am simply
 looking for whichever blank frame/scene change is closest to the estimated
 preroll/postroll frames. I haven't put much effort into improving it at
 this point as the preroll was the most important for me, and i just coded
 it shortly before the feature freeze was announced.

-- 
Ticket URL: <http://svn.mythtv.org/trac/ticket/6889#comment:3>
MythTV <http://www.mythtv.org/>
MythTV