[FFmpeg-devel] [PATCH] lavc: add text subtitles decoder.

[Nicolas George]

Le tridi 23 vendémiaire, an CCXXI, Clément Bœsch a écrit :
> TODO: lavc bump
> ---
> This patch makes possible to extract & convert text subtitles from formats like
> Ogg, with a simple ffmpeg -i in.ogg out.ass.
> 
> At some point we might consider changing the AV_CODEC_ID_TEXT into
> AV_CODEC_ID_SUBRIP in the Ogg demuxer and some others (if some samples with
> SubRip markup appears, which is very likely). Still, I believe this decoder
> will be useful for some other cases such as:
> 
>  - potential formats making an explicit difference between raw text subtitles
>    and SubRip markup'ed text subtitles
>  - a bunch of old formats like vplayer, mpl2, will have this common decoder
>    (demuxers incoming when I'll have some time available to write them)
>  - keeping compat with the fork which might continue to add some demuxers with
>    text codec without the need for us to change it to subrip (until someone
>    request it because he just found a sample with markup in the given format)
> ---
>  libavcodec/Makefile    |  1 +
>  libavcodec/allcodecs.c |  1 +
>  libavcodec/textdec.c   | 97 ++++++++++++++++++++++++++++++++++++++++++++++++++
>  3 files changed, 99 insertions(+)
>  create mode 100644 libavcodec/textdec.c
> 
> diff --git a/libavcodec/Makefile b/libavcodec/Makefile
> index 47a01db..1811b7b 100644
> --- a/libavcodec/Makefile
> +++ b/libavcodec/Makefile
> @@ -402,6 +402,7 @@ OBJS-$(CONFIG_SVQ3_DECODER)            += svq3.o svq13.o h263.o h264.o        \
>                                            h264_loopfilter.o h264_direct.o     \
>                                            h264_sei.o h264_ps.o h264_refs.o    \
>                                            h264_cavlc.o h264_cabac.o cabac.o
> +OBJS-$(CONFIG_TEXT_DECODER)            += textdec.o ass.o
>  OBJS-$(CONFIG_TAK_DECODER)             += takdec.o tak.o
>  OBJS-$(CONFIG_TARGA_DECODER)           += targa.o
>  OBJS-$(CONFIG_TARGA_ENCODER)           += targaenc.o rle.o
> diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c
> index ae88fab..c023f35 100644
> --- a/libavcodec/allcodecs.c
> +++ b/libavcodec/allcodecs.c
> @@ -424,6 +424,7 @@ void avcodec_register_all(void)
>      REGISTER_ENCDEC  (SRT, srt);
>      REGISTER_ENCDEC  (SUBRIP, subrip);
>      REGISTER_DECODER (SUBVIEWER, subviewer);
> +    REGISTER_DECODER (TEXT, text);
>      REGISTER_DECODER (WEBVTT, webvtt);
>      REGISTER_ENCDEC  (XSUB, xsub);
>  
> diff --git a/libavcodec/textdec.c b/libavcodec/textdec.c
> new file mode 100644
> index 0000000..a154995
> --- /dev/null
> +++ b/libavcodec/textdec.c
> @@ -0,0 +1,97 @@
> +/*
> + * Copyright (c) 2012 Clément Bœsch
> + *
> + * This file is part of FFmpeg.
> + *
> + * FFmpeg is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU Lesser General Public
> + * License as published by the Free Software Foundation; either
> + * version 2.1 of the License, or (at your option) any later version.
> + *
> + * FFmpeg is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
> + * Lesser General Public License for more details.
> + *
> + * You should have received a copy of the GNU Lesser General Public
> + * License along with FFmpeg; if not, write to the Free Software
> + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
> + */
> +
> +/**
> + * @file
> + * Raw subtitles decoder
> + */
> +
> +#include "avcodec.h"
> +#include "ass.h"
> +#include "libavutil/bprint.h"
> +#include "libavutil/opt.h"
> +
> +typedef struct {
> +    AVClass *class;
> +    char *linebreaks;
> +    int escape_ass_tags;
> +} TextContext;
> +
> +#define OFFSET(x) offsetof(TextContext, x)
> +#define SD AV_OPT_FLAG_SUBTITLE_PARAM | AV_OPT_FLAG_DECODING_PARAM
> +static const AVOption options[] = {
> +    { "linebreaks",      "Extra line breaks characters",    OFFSET(linebreaks),      AV_OPT_TYPE_STRING, {.str=NULL}, .flags=SD },

I wonder if we could make these arrays more readable using macros for the
whole lines:

OPT_STR("linebreaks", "Extra line breaks characters", linebreaks, NULL),

The macros could be in a common private header and used everywhere.

But of course this is unrelated.

> +    { "escape_ass_tags", "Set if ASS tags must be escaped", OFFSET(escape_ass_tags), AV_OPT_TYPE_INT,    {.i64=1},    .flags=SD },

I would be slightly happier of the option was the other way around:
-keep_ass_markup 1. But of course this is up to you.

> +    { NULL }
> +};
> +
> +static const AVClass text_decoder_class = {
> +    .class_name = "text decoder",
> +    .item_name  = av_default_item_name,
> +    .option     = options,
> +    .version    = LIBAVUTIL_VERSION_INT,
> +};
> +
> +static int text_event_to_ass(const AVCodecContext *avctx, AVBPrint *buf, const char *p)
> +{
> +    const TextContext *text = avctx->priv_data;
> +
> +    while (*p) {
> +        if ((text->linebreaks && strchr(text->linebreaks, *p)) ||
> +            (p[0] == '\n' && p[1]))
> +            av_bprintf(buf, "\\N");
> +        else if (text->escape_ass_tags && (*p == '{' || *p == '}'))
> +            av_bprintf(buf, "\\%c", *p);

Does the closing brace need to be escaped?

OTOH, the backslash needs to be escaped I believe.

> +        else if (*p != '\r')
> +            av_bprint_chars(buf, *p, 1);

???

> +        p++;
> +    }
> +    av_bprintf(buf, "\r\n");

I thought the decoded subtitles were not supposed to have a final line
break.

Also, keeping the CRLF in the internal encoding seems wrong.

> +    return 0;
> +}
> +
> +static int text_decode_frame(AVCodecContext *avctx, void *data,
> +                             int *got_sub_ptr, AVPacket *avpkt)
> +{
> +    AVBPrint buf;
> +    AVSubtitle *sub = data;
> +    const char *ptr = avpkt->data;
> +    const int ts_start     = av_rescale_q(avpkt->pts,      avctx->time_base, (AVRational){1,100});
> +    const int ts_duration  = avpkt->duration != -1 ?
> +                             av_rescale_q(avpkt->duration, avctx->time_base, (AVRational){1,100}) : -1;
> +
> +    av_bprint_init(&buf, 0, AV_BPRINT_SIZE_UNLIMITED);
> +    if (ptr && *ptr && avpkt->size > 0 &&
> +        !text_event_to_ass(avctx, &buf, ptr))
> +        ff_ass_add_rect(sub, buf.str, ts_start, ts_duration, 0);

You do not check if buf has been truncated due to out of memory.

> +    *got_sub_ptr = sub->num_rects > 0;
> +    av_bprint_finalize(&buf, NULL);
> +    return avpkt->size;
> +}
> +
> +AVCodec ff_text_decoder = {
> +    .name      = "text",
> +    .priv_data_size = sizeof(TextContext),
> +    .long_name = NULL_IF_CONFIG_SMALL("Raw text subtitle"),
> +    .type      = AVMEDIA_TYPE_SUBTITLE,
> +    .id        = AV_CODEC_ID_TEXT,
> +    .decode    = text_decode_frame,
> +    .init      = ff_ass_subtitle_header_default,
> +};

Also, are we sure that everything handled by this decoder (and the SUBRIP
decoder too, in fact) will already be in UTF-8? This is probably not true,
and it will require unification.

On the whole, looks good and useful, thanks.

Regards,

-- 
  Nicolas George
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20121014/ff7e8a5a/attachment.asc>