[FFmpeg-devel] AVWriter again (was: v2 1/2] avformat/url: check double dot is not to parent directory)

[Alexander Strasser]

On 2020-07-28 20:43 +0800, zhilizhao wrote:
> > On Jul 28, 2020, at 8:02 PM, Nicolas George <george at nsup.org> wrote:
> >
> > zhilizhao (12020-07-28):
> >> I think jb is referring to
> >>
> >>     FILE *open_memstream(char **bufp, size_t *sizep);
> >> https://linux.die.net/man/3/open_memstream <https://linux.die.net/man/3/open_memstream>
> >>
> >> VLC has a wrapper for it:
> >>
> >> https://code.videolan.org/videolan/vlc/-/blob/master/include/vlc_memstream.h <https://code.videolan.org/videolan/vlc/-/blob/master/include/vlc_memstream.h>
> >
> > Oh. Thanks. I had not realized such an useful function had been
> > standardized.
> >
> > That means the argument about it being in lavf is wotrhless. But these
> > two arguments:
> >
> >>> - avio_open_dyn_buf() always uses dynamic allocation. AVWriter starts
> >>> entirely on the stack (or whatever we want), and only resorts to
> >>> dynamic allocation if the buffer becomes too large. As a corollary,
> >>> avio_open_dyn_buf() requires an error check at the beginning, which
> >>> AVWriter does not.
>
> fmemopen() can work with memory on stack or on heap, although it doesn’t
> support switch between the two dynamically. It’s a good to have feature like
> the small string optimization in std::string, but i’m not too keen on it.
>
> >>>
> >>> - AVWriter is polymorphic. I have shown only av_dynbuf_writer(), but
> >>> there are others, and applications can implement their own. That is
> >>> the key to the "wonderful features" I have evoked. I would be happy to
> >>> tell more about them.
> >
> > apply to open_memstream() too.
> >
> > Please ask me about the wonderful features ;)
>
> Please, what wonderful features are in your plan?

I would say FILE * is polymorphic as well.

Maybe we are not in control to provide more implementations, the
question still would be which implementations we would want.


  Alexander