[FFmpeg-devel] [PATCH 1/2] avformat: add avformat_query_seekable

[Gyan Doshi]

On 2025-04-13 09:39 pm, softworkz . wrote:
>
>> -----Original Message-----
>> From: ffmpeg-devel<ffmpeg-devel-bounces at ffmpeg.org>  On Behalf Of Gyan
>> Doshi
>> Sent: Sonntag, 13. April 2025 13:16
>> To:ffmpeg-devel at ffmpeg.org
>> Subject: [FFmpeg-devel] [PATCH 1/2] avformat: add
>> avformat_query_seekable
>>
>> Utility function to report seekability features for a given input.
>>
>> Useful for ffprobe and to extend seek possibilities in fftools.
>> ---
>>   doc/APIchanges         |  3 +++
>>   libavformat/avformat.h | 22 ++++++++++++++++++
>>   libavformat/seek.c     | 53 ++++++++++++++++++++++++++++++++++++++++++
>>   libavformat/version.h  |  2 +-
>>   4 files changed, 79 insertions(+), 1 deletion(-)
>>
>> diff --git a/doc/APIchanges b/doc/APIchanges
>> index 65bf5a9419..879f56b572 100644
>> --- a/doc/APIchanges
>> +++ b/doc/APIchanges
>> @@ -2,6 +2,9 @@ The last version increases of all libraries were on
>> 2025-03-28
>>
>>   API changes, most recent first:
>>
>> +2025-04-xx - xxxxxxxxxx - lavf 62.1.100 - avformat.h
>> +  Add avformat_query_seekable().
>> +
>>   2025-04-07 - 19e9a203b7 - lavu 60.01.100 - dict.h
>>     Add AV_DICT_DEDUP.
>>
>> diff --git a/libavformat/avformat.h b/libavformat/avformat.h
>> index 498c3020a5..e00f1ed0d9 100644
>> --- a/libavformat/avformat.h
>> +++ b/libavformat/avformat.h
>> @@ -2338,6 +2338,28 @@ int av_seek_frame(AVFormatContext *s, int
>> stream_index, int64_t timestamp,
>>    */
>>   int avformat_seek_file(AVFormatContext *s, int stream_index, int64_t
>> min_ts, int64_t ts, int64_t max_ts, int flags);
> Hi Gyan,
>
> I like the idea of showing seekability in FFprobe output!
>
> As far as I understand, these flags are expressing three different things:
>
>
> 1. Seek method
>
>> +#define AVSEEKABLE_NORMAL     0x01 ///< I/O is seekable like a local
>> file
>> +#define AVSEEKABLE_PROTOCOL   0x02 ///< I/O seek is through protocol
>> request via avio_seek_time
>> +#define AVSEEKABLE_DEMUXER    0x04 ///< demuxer has a seek function
>> +#define AVSEEKABLE_PKTSCAN    0x08 ///< seek is performed by consuming

The first is not a method. It indicates whether the input is streamed.

> 2. Seek parameter type
>
>> and scanning packet timestamps
>> +#define AVSEEKABLE_TIME       0x10 ///< seek target can be a timestamp
>> +#define AVSEEKABLE_BYTE       0x20 ///< seek target can be in bytes
>> +#define AVSEEKABLE_FRAME      0x40 ///< seek target can be a frame
>> index
> 3. Additional seek behavior
>
>> +#define AVSEEKABLE_PTS       0x100 ///< seek target timestamp is
>> expected to be PTS
>> +#define AVSEEKABLE_FAST      0x200 ///< demuxer allows fast but
>> inaccurate seeking
>> +#define AVSEEKABLE_FWDONLY   0x400 ///< set seek will be equal or
>> forward of specified seek point
>
> Wouldn't it be better for clarity when this differentiation would be
> reflected in the naming?
>
> Maybe something like this:
>
> AVSEEKABLE_VIA_DEMUXER
>
> AVSEEKABLE_BY_TIME
>
> AVSEEKABLE_BEHAVIOR_FAST

Will do.

> Another note:
>
>> +#define AVSEEKABLE_NORMAL     0x01 ///< I/O is seekable like a local
>> +#define AVSEEKABLE_PROTOCOL   0x02 ///< I/O seek is through protocol
>> +#define AVSEEKABLE_DEMUXER    0x04 ///< demuxer has a seek function
>> +#define AVSEEKABLE_PKTSCAN    0x08 ///< seek is performed by consuming
> There's no (numeric) room for extending the first block

At the moment, there are 21 unused bits so there's no need as such.

>
>> +#define AVSEEKABLE_TIME       0x10 ///< seek target can be a timestamp
>> +#define AVSEEKABLE_BYTE       0x20 ///< seek target can be in bytes
>> +#define AVSEEKABLE_FRAME      0x40 ///< seek target can be a frame
>
>> +    // TODO: no flag for frame seeking. Add flag and update this check
>> +    if (s->iformat->flags & 0)
>> +        ret |= AVSEEKABLE_FRAME;
>> +
>> +    if (s->iformat->flags & AVFMT_SEEK_TO_PTS)
>> +        ret |= AVSEEKABLE_PTS;
> Maybe add a check for AVSEEKABLE_TIME - for sanity?

Only a handful of demuxers have this flag, and all have TS.
This is a job for reviewers when adding the flag.

>
>> +    // TODO: flag exists but not added to the demuxers which support it
>> +    if (s->iformat->flags & AVFMT_FLAG_FAST_SEEK)
>> +        ret |= AVSEEKABLE_FAST;
>
>
> Another idea would be to return a struct (like AVSeekability) with a clear
> separation of those three types of information, maybe that would provide
> better clarity?

As alluded above, there's only a handful of attributes being indicated. 
I'd rather not spawn a new type when an inbuilt simple storage type will do.

Regards,
Gyan