diff mbox series

[FFmpeg-devel] lavfi/drawtext: Add localtime_ms for millisecond precision

Message ID bcdf1a41-1101-9a2c-3928-344a9a0038c8@mail.de
State New
Headers show
Series [FFmpeg-devel] lavfi/drawtext: Add localtime_ms for millisecond precision | expand

Checks

Context Check Description
andriy/x86_make success Make finished
andriy/x86_make_fate success Make fate finished
andriy/PPC64_make success Make finished
andriy/PPC64_make_fate success Make fate finished

Commit Message

Thilo Borgmann June 8, 2021, 4:42 p.m. UTC
Hi,

add %{localtime_ms} function to the drawtext filter. Same as %{localtime} but with additional millisecond part. 

-Thilo
From f3b32f3e2b01f50b4fa41f8432ccd5a5eeea1ec1 Mon Sep 17 00:00:00 2001
From: Kirill Pugin <ikir@fb.com>
Date: Tue, 8 Jun 2021 18:32:45 +0200
Subject: [PATCH] lavfi/drawtext: Add localtime_ms for millisecond precision

Suggested-By: ffmpeg@fb.com
---
 doc/filters.texi          |  4 ++++
 libavfilter/vf_drawtext.c | 20 +++++++++++++++++++-
 2 files changed, 23 insertions(+), 1 deletion(-)

Comments

Valerii Zapodovnikov June 8, 2021, 4:47 p.m. UTC | #1
Plese resend it in new thread but rename it to .txt. thanks.
Andreas Rheinhardt June 8, 2021, 4:50 p.m. UTC | #2
Valerii Zapodovnikov:
> Plese resend it in new thread but rename it to .txt. thanks.

Why should he do so?

- Andreas
Valerii Zapodovnikov June 8, 2021, 4:52 p.m. UTC | #3
Oh, so it worked on patchwork? Strnge, okay then. No need to.
Thilo Borgmann June 8, 2021, 4:57 p.m. UTC | #4
Am 08.06.21 um 18:52 schrieb Valerii Zapodovnikov:
> Oh, so it worked on patchwork? Strnge, okay then. No need to.

Please read [1] and [2]. Please keep all old text and do inline-commenting.

-Thilo

[1] http://ffmpeg.org/developer.html
[2] http://ffmpeg.org/mailing-list-faq.html
Valerii Zapodovnikov June 8, 2021, 5 p.m. UTC | #5
вт, 8 июн. 2021 г., 19:57 Thilo Borgmann <thilo.borgmann@mail.de>:

> Am 08.06.21 um 18:52 schrieb Valerii Zapodovnikov:
> > Oh, so it worked on patchwork? Strnge, okay then. No need to.
>
> Please read [1] and [2]. Please keep all old text and do inline-commenting.
>
> -Thilo
>

No.

>
> [1] http://ffmpeg.org/developer.html
> [2] http://ffmpeg.org/mailing-list-faq.html
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
>
Thilo Borgmann June 8, 2021, 5:24 p.m. UTC | #6
Am 08.06.21 um 19:00 schrieb Valerii Zapodovnikov:
> вт, 8 июн. 2021 г., 19:57 Thilo Borgmann <thilo.borgmann@mail.de>:
> 
>> Am 08.06.21 um 18:52 schrieb Valerii Zapodovnikov:
>>> Oh, so it worked on patchwork? Strnge, okay then. No need to.
>>
>> Please read [1] and [2]. Please keep all old text and do inline-commenting.
>>
>> -Thilo
>>
> 
> No.

In ignoring our rules for the mailing list as stated in [3] you just got yourself banned.

-Thilo

[3] http://ffmpeg.org/contact.html#MailingLists

>>
>> [1] http://ffmpeg.org/developer.html
>> [2] http://ffmpeg.org/mailing-list-faq.html
>> _______________________________________________
>> ffmpeg-devel mailing list
>> ffmpeg-devel@ffmpeg.org
>> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>>
>> To unsubscribe, visit link above, or email
>> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
>>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
>
Thilo Borgmann June 14, 2021, 7:14 p.m. UTC | #7
Am 08.06.21 um 18:42 schrieb Thilo Borgmann:
> Hi,
> 
> add %{localtime_ms} function to the drawtext filter. Same as %{localtime} but with additional millisecond part. 

Ping for a volunteer to actually have a look at the patch for review.

Thanks,
Thilo
Steven Liu June 15, 2021, 3:42 a.m. UTC | #8
Thilo Borgmann <thilo.borgmann@mail.de> 于2021年6月15日周二 上午3:14写道:
>
> Am 08.06.21 um 18:42 schrieb Thilo Borgmann:
> > Hi,
> >
> > add %{localtime_ms} function to the drawtext filter. Same as %{localtime} but with additional millisecond part.
>
> Ping for a volunteer to actually have a look at the patch for review.
lgtm

>
> Thanks,
> Thilo
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".

Thanks
Steven
Lingjiang Fang June 16, 2021, 1:56 a.m. UTC | #9
On Tue, 15 Jun 2021 11:42:44 +0800
Steven Liu <lingjiujianke@gmail.com> wrote:

> Thilo Borgmann <thilo.borgmann@mail.de> 于2021年6月15日周二
> 上午3:14写道:
> >
> > Am 08.06.21 um 18:42 schrieb Thilo Borgmann:  
> > > Hi,
> > >
> > > add %{localtime_ms} function to the drawtext filter. Same as
> > > %{localtime} but with additional millisecond part.  
> >
> > Ping for a volunteer to actually have a look at the patch for
> > review.  
> lgtm
it has similar function with a previous patch from Zhao Zhili, but
is more unified with current logic, thank you all for adding this

I think it's very useful to adding millisecond timestamp at source
frames, makeing detecting frame lost, reorder, repeat much easier.

Hoping this patch can be merged soon :)

BTW, if we has an OCR filter to extract this, it will be perfect :)

> 
> >
> > Thanks,
> > Thilo
> > _______________________________________________
> > ffmpeg-devel mailing list
> > ffmpeg-devel@ffmpeg.org
> > https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> >
> > To unsubscribe, visit link above, or email
> > ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".  
> 
> Thanks
> Steven
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".



Regards,
Lingjiang Fang
Liu Steven June 16, 2021, 2:52 a.m. UTC | #10
> 2021年6月16日 上午9:56,Lingjiang Fang <vacingfang@hotmail.com> 写道:
> 
> On Tue, 15 Jun 2021 11:42:44 +0800
> Steven Liu <lingjiujianke@gmail.com> wrote:
> 
>> Thilo Borgmann <thilo.borgmann@mail.de> 于2021年6月15日周二
>> 上午3:14写道:
>>> 
>>> Am 08.06.21 um 18:42 schrieb Thilo Borgmann:  
>>>> Hi,
>>>> 
>>>> add %{localtime_ms} function to the drawtext filter. Same as
>>>> %{localtime} but with additional millisecond part.  
>>> 
>>> Ping for a volunteer to actually have a look at the patch for
>>> review.  
>> lgtm
> it has similar function with a previous patch from Zhao Zhili, but
> is more unified with current logic, thank you all for adding this
> 
> I think it's very useful to adding millisecond timestamp at source
> frames, makeing detecting frame lost, reorder, repeat much easier.
> 
> Hoping this patch can be merged soon :)
> 
> BTW, if we has an OCR filter to extract this, it will be perfect :)
Do you mean : libavfilter/vf_ocr.c ?
> 
>> 
>>> 
>>> Thanks,
>>> Thilo
>>> _______________________________________________
>>> ffmpeg-devel mailing list
>>> ffmpeg-devel@ffmpeg.org
>>> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>>> 
>>> To unsubscribe, visit link above, or email
>>> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".  
>> 
>> Thanks
>> Steven
>> _______________________________________________
>> ffmpeg-devel mailing list
>> ffmpeg-devel@ffmpeg.org
>> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>> 
>> To unsubscribe, visit link above, or email
>> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
> 
> 
> 
> Regards,
> Lingjiang Fang
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".

Thanks

Steven Liu
Zhao Zhili June 16, 2021, 6:23 a.m. UTC | #11
> On Jun 15, 2021, at 3:14 AM, Thilo Borgmann <thilo.borgmann@mail.de> wrote:
> 
> Am 08.06.21 um 18:42 schrieb Thilo Borgmann:
>> Hi,
>> 
>> add %{localtime_ms} function to the drawtext filter. Same as %{localtime} but with additional millisecond part. 
> 
> Ping for a volunteer to actually have a look at the patch for review.
> 
> Thanks,
> Thilo
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe”.
> 

> diff --git a/libavfilter/vf_drawtext.c b/libavfilter/vf_drawtext.c
> index 382d589e26..684f954236 100644
> --- a/libavfilter/vf_drawtext.c
> +++ b/libavfilter/vf_drawtext.c
> @@ -53,6 +53,7 @@
>  #include "libavutil/parseutils.h"
>  #include "libavutil/timecode.h"
>  #include "libavutil/time_internal.h"
> +#include "libavutil/time.h"
>  #include "libavutil/tree.h"
>  #include "libavutil/lfg.h"
>  #include "libavutil/detection_bbox.h"
> @@ -1049,11 +1050,27 @@ static int func_strftime(AVFilterContext *ctx, AVBPrint *bp,
>      struct tm tm;
>  
>      time(&now);
> -    if (tag == 'L')
> +    if (tag == 'L' || tag == 'M')
>          localtime_r(&now, &tm);
>      else
>          tm = *gmtime_r(&now, &tm);
>      av_bprint_strftime(bp, fmt, &tm);
> +
> +    if (tag == 'M') {
> +        char ms[5] = {0};
> +        // get time returns time since epoch in micro seconds, localtime_r would
> +        // already take care of representing seconds part, so we need to get
> +        // milliseconds, to do that we round up microseconds to seconds
> +        // then convert seconds back to microseconds - this will drop microsecond
> +        // precision, then compute difference between av_gettime result and
> +        // rounded up sec -> usec representation and convert that to milliseconds
> +        int64_t usecs      = av_gettime();

There is no guarantee that when av_gettime() is called, it’s in the same second as ’time(&now)’
was called. For example:

time(&now) was called at 10.999999 second, and return 10.
av_gettime() was called at 11.001 second

It will show 10.001, which should be 11.001.

> +        int64_t secs       = usecs / 1000000;
> +        int64_t msec_delta = (usecs - secs * 1000000) / 1000;
> +        snprintf(ms, 5, ".%03d", (int)msec_delta);
> +        av_bprint_append_data(bp, ms, 4);
> +    }
> +
>      return 0;
>  }
>  
> @@ -1153,6 +1170,7 @@ static const struct drawtext_function {
>      { "pts",       0, 3, 0,   func_pts      },
>      { "gmtime",    0, 1, 'G', func_strftime },
>      { "localtime", 0, 1, 'L', func_strftime },
> +    { "localtime_ms", 0, 1, 'M', func_strftime },

Then why not add a gmtime_ms?

>      { "frame_num", 0, 0, 0,   func_frame_num },
>      { "n",         0, 0, 0,   func_frame_num },
>      { "metadata",  1, 2, 0,   func_metadata },
> -- 
> 2.20.1 (Apple Git-117)
Lingjiang Fang June 17, 2021, 5:11 a.m. UTC | #12
On Wed, 16 Jun 2021 10:52:31 +0800
Steven Liu <lq@chinaffmpeg.org> wrote:

> > 2021年6月16日 上午9:56,Lingjiang Fang <vacingfang@hotmail.com>
> > 写道:
> > 
> > On Tue, 15 Jun 2021 11:42:44 +0800
> > Steven Liu <lingjiujianke@gmail.com> wrote:
> >   
> >> Thilo Borgmann <thilo.borgmann@mail.de> 于2021年6月15日周二
> >> 上午3:14写道:  
> >>> 
> >>> Am 08.06.21 um 18:42 schrieb Thilo Borgmann:    
> >>>> Hi,
> >>>> 
> >>>> add %{localtime_ms} function to the drawtext filter. Same as
> >>>> %{localtime} but with additional millisecond part.    
> >>> 
> >>> Ping for a volunteer to actually have a look at the patch for
> >>> review.    
> >> lgtm  
> > it has similar function with a previous patch from Zhao Zhili, but
> > is more unified with current logic, thank you all for adding this
> > 
> > I think it's very useful to adding millisecond timestamp at source
> > frames, makeing detecting frame lost, reorder, repeat much easier.
> > 
> > Hoping this patch can be merged soon :)
> > 
> > BTW, if we has an OCR filter to extract this, it will be perfect :)
> >  
> Do you mean : libavfilter/vf_ocr.c ?
yes, vf_ocr is almost what I want, but need a little bit improve,
supporting specify a sub region of the image, because I just want to
know the timestamp on it, not the content of the image : )

I am working on it, demo can work now

> >   
> >>   
> >>> 
> >>> Thanks,
> >>> Thilo
> >>> _______________________________________________
> >>> ffmpeg-devel mailing list
> >>> ffmpeg-devel@ffmpeg.org
> >>> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> >>> 
> >>> To unsubscribe, visit link above, or email
> >>> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".    
> >> 
> >> Thanks
> >> Steven
> >> _______________________________________________
> >> ffmpeg-devel mailing list
> >> ffmpeg-devel@ffmpeg.org
> >> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> >> 
> >> To unsubscribe, visit link above, or email
> >> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".  
> > 
> > 
> > 
> > Regards,
> > Lingjiang Fang
> > _______________________________________________
> > ffmpeg-devel mailing list
> > ffmpeg-devel@ffmpeg.org
> > https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> > 
> > To unsubscribe, visit link above, or email
> > ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".  
> 
> Thanks
> 
> Steven Liu
> 
> 
> 
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".



Regards,
Lingjiang Fang
diff mbox series

Patch

diff --git a/doc/filters.texi b/doc/filters.texi
index 78faf767cf..bf073e0560 100644
--- a/doc/filters.texi
+++ b/doc/filters.texi
@@ -10953,6 +10953,10 @@  It can accept an argument: a strftime() format string.
 The time at which the filter is running, expressed in the local time zone.
 It can accept an argument: a strftime() format string.
 
+@item localtime_ms
+Same as @code{localtime} but with millisecond precision.
+It can accept an argument: a strftime() format string.
+
 @item metadata
 Frame metadata. Takes one or two arguments.
 
diff --git a/libavfilter/vf_drawtext.c b/libavfilter/vf_drawtext.c
index 382d589e26..684f954236 100644
--- a/libavfilter/vf_drawtext.c
+++ b/libavfilter/vf_drawtext.c
@@ -53,6 +53,7 @@ 
 #include "libavutil/parseutils.h"
 #include "libavutil/timecode.h"
 #include "libavutil/time_internal.h"
+#include "libavutil/time.h"
 #include "libavutil/tree.h"
 #include "libavutil/lfg.h"
 #include "libavutil/detection_bbox.h"
@@ -1049,11 +1050,27 @@  static int func_strftime(AVFilterContext *ctx, AVBPrint *bp,
     struct tm tm;
 
     time(&now);
-    if (tag == 'L')
+    if (tag == 'L' || tag == 'M')
         localtime_r(&now, &tm);
     else
         tm = *gmtime_r(&now, &tm);
     av_bprint_strftime(bp, fmt, &tm);
+
+    if (tag == 'M') {
+        char ms[5] = {0};
+        // get time returns time since epoch in micro seconds, localtime_r would
+        // already take care of representing seconds part, so we need to get
+        // milliseconds, to do that we round up microseconds to seconds
+        // then convert seconds back to microseconds - this will drop microsecond
+        // precision, then compute difference between av_gettime result and
+        // rounded up sec -> usec representation and convert that to milliseconds
+        int64_t usecs      = av_gettime();
+        int64_t secs       = usecs / 1000000;
+        int64_t msec_delta = (usecs - secs * 1000000) / 1000;
+        snprintf(ms, 5, ".%03d", (int)msec_delta);
+        av_bprint_append_data(bp, ms, 4);
+    }
+
     return 0;
 }
 
@@ -1153,6 +1170,7 @@  static const struct drawtext_function {
     { "pts",       0, 3, 0,   func_pts      },
     { "gmtime",    0, 1, 'G', func_strftime },
     { "localtime", 0, 1, 'L', func_strftime },
+    { "localtime_ms", 0, 1, 'M', func_strftime },
     { "frame_num", 0, 0, 0,   func_frame_num },
     { "n",         0, 0, 0,   func_frame_num },
     { "metadata",  1, 2, 0,   func_metadata },