Patch to allow drawtext frame_num/n var to start at a number other than 0

[mjmvisser]

In reference to https://ffmpeg.org/trac/ffmpeg/ticket/1949

The suggested workaround to use expr\\\:n+1 doesn't solve the problem, as the frame number is drawn as a double (1.000000000) instead of as an integer.

I followed the example set by the image2 format and added a start_number parameter. For example, the following draws the frame number at the bottom right, starting at 30:

-vf drawtext="fontsize=15:fontfile=/path/to/font.ttf:text=%{n}:start_number=30:fontsize=25:fontcolor='white':x=w-text_w-line_h:\y=h-text_h-line_h"

[ubitux]

Note that you can do something like:

ffplay -f lavfi color=black -vf "drawtext=text='%{e\:n+123}':fontcolor=white"

May I ask what is the purpose behind such feature? Is it just to solve the problem of n starting at 0 or 1?

Edit1:
Oh mmh, I should have read your initial post, my bad.

But another note; now that the counter is stored in a common place (link->frame_count), I wonder if we couldn't have a more global starting number instead. That would make more sense when using for instance in combination with showinfo filter (which prints the n), or actually any kind of filters where the counter is used (anything with timeline for instance, or various expressions).

Edit2:
To be more specific, maybe have a look to filters_common_options in avfilter.c, and try to add it here. Unfortunately, if you do that, it will be limited to each filter scope. Maybe we need some global libavfilter options, so when a -start_n is specified by a tool, it is populated everywhere. Such thing should be discussed on ffmpeg-devel mailing list or IRC.

I know I'm being pedantic for such a trivial feature, but I believe it can be done way better and have a few benefits in the long term. Adding such option like this now will mean a depreciation burden next time it's implemented generically.

[mjmvisser]

Good idea (filters_common_options), I hadn't realized n was used outside of the drawtext filter. I'll update this patch when I have a chance and post a link to it on the mailing list.

cheers,
-Mark

[mjmvisser]

Do to some n00bness on my part, this pull request was empty. I've resubmitted it here: #24