ffmpeg 源代码简单分析 av read frame

通过av_read_packet(***)，读取一个包，需要说明的是此函数必须是包含整数帧的，不存在半帧的情况，以ts流为例，是读取一个完整的PES包（一个完整pes包包含若干视频或音频es包），读取完毕后，通过av_parser_parse2(***)分析出视频一帧（或音频若干帧），返回，下次进入循环的时候，如果上次的数据没有完全取完，则st = s->cur_st;不会是NULL，即再此进入av_parser_parse2(***)流程，而不是下面的av_read_packet（**）流程，这样就保证了，如果读取一次包含了N帧视频数据（以视频为例），则调用av_read_frame（***）N次都不会去读数据，而是返回第一次读取的数据，直到全部解析完毕。

av_read_frame()的声明位于libavformat\avformat.h，如下所示。

/** * Return the next frame of a stream. * This function returns what is stored in the file, and does not validate * that what is there are valid frames for the decoder. It will split what is * stored in the file into frames and return one for each call. It will not * omit invalid data between valid frames so as to give the decoder the maximum * information possible for decoding. * * If pkt->buf is NULL, then the packet is valid until the next * av_read_frame() or until avformat_close_input(). Otherwise the packet * is valid indefinitely. In both cases the packet must be freed with * av_free_packet when it is no longer needed. For video, the packet contains * exactly one frame. For audio, it contains an integer number of frames if each * frame has a known fixed size (e.g. PCM or ADPCM data). If the audio frames * have a variable size (e.g. MPEG audio), then it contains one frame. * * pkt->pts, pkt->dts and pkt->duration are always set to correct * values in AVStream.time_base units (and guessed if the format cannot * provide them). pkt->pts can be AV_NOPTS_VALUE if the video format * has B-frames, so it is better to rely on pkt->dts if you do not * decompress the payload. * * @return 0 if OK, < 0 on error or end of file */int av_read_frame(AVFormatContext *s, AVPacket *pkt);

av_read_frame()使用方法在注释中写得很详细，用中文简单描述一下它的两个参数：

s：输入的AVFormatContext

pkt：输出的AVPacket

如果返回0则说明读取正常。

函数调用结构图

函数调用结构图如下所示。

av_read_frame()

av_read_frame()的定义位于libavformat\utils.c，如下所示：

//获取一个AVPacket/* * av_read_frame - 新版本的ffmpeg用的是av_read_frame，而老版本的是av_read_packet * 。区别是av_read_packet读出的是包，它可能是半帧或多帧，不保证帧的完整性。av_read_frame对 * av_read_packet进行了封装，使读出的数据总是完整的帧 */int av_read_frame(AVFormatContext *s, AVPacket *pkt){    const int genpts = s->flags & AVFMT_FLAG_GENPTS;    int          eof = 0;    if (!genpts)     /**      * This buffer is only needed when packets were already buffered but      * not decoded, for example to get the codec parameters in MPEG      * streams.      * 一般情况下会调用read_frame_internal(s, pkt)      * 直接返回      */        return s->packet_buffer ? read_from_packet_buffer(s, pkt) :                                  read_frame_internal(s, pkt);    for (;;) {        int ret;        AVPacketList *pktl = s->packet_buffer;        if (pktl) {            AVPacket *next_pkt = &pktl->pkt;            if (next_pkt->dts != AV_NOPTS_VALUE) {                int wrap_bits = s->streams[next_pkt->stream_index]->pts_wrap_bits;                while (pktl && next_pkt->pts == AV_NOPTS_VALUE) {                    if (pktl->pkt.stream_index == next_pkt->stream_index &&                        (av_compare_mod(next_pkt->dts, pktl->pkt.dts, 2LL << (wrap_bits - 1)) < 0) &&                         av_compare_mod(pktl->pkt.pts, pktl->pkt.dts, 2LL << (wrap_bits - 1))) { //not b frame                        next_pkt->pts = pktl->pkt.dts;                    }                    pktl = pktl->next;                }                pktl = s->packet_buffer;            }            /* read packet from packet buffer, if there is data */            if (!(next_pkt->pts == AV_NOPTS_VALUE &&                  next_pkt->dts != AV_NOPTS_VALUE &&