Open input/output separately, and read/write separately

This is a refactor of the library API so that we will be able to have
multiple outputs.

cmd/remux_example/remux_example.c

@@ -21,10 +21,18 @@ int main(const int argc, const char * const * const argv)
 	const char * const output_url = "file:/tmp/out.mp4";
 	const bool verbose = true;
 
-	struct Videostreamer * const vs = vs_open(input_format, input_url,
-			output_format, output_url, verbose);
-	if (!vs) {
-		printf("unable to open videostreamer\n");
+	struct VSInput * const input = vs_open_input(input_format, input_url,
+			verbose);
+	if (!input) {
+		printf("unable to open input\n");
+		return 1;
+	}
+
+	struct VSOutput * const output = vs_open_output(output_format, output_url,
+			input, verbose);
+	if (!output) {
+		printf("unable to open output\n");
+		vs_destroy_input(input);
 		return 1;
 	}
 
@@ -32,14 +40,31 @@ int main(const int argc, const char * const * const argv)
 	int i = 0;
 
 	while (1) {
-		const int frame_size = vs_read_write(vs, verbose);
-		if (frame_size == -1) {
-			printf("read/write failed\n");
-			vs_destroy(vs);
+		AVPacket pkt;
+		memset(&pkt, 0, sizeof(AVPacket));
+
+		const int read_res = vs_read_packet(input, &pkt, verbose);
+		if (read_res == -1) {
+			printf("read failed\n");
+			vs_destroy_input(input);
+			vs_destroy_output(output);
 			return 1;
 		}
 
-		printf("frame size %d\n", frame_size);
+		if (read_res == 0) {
+			continue;
+		}
+
+		const int write_res = vs_write_packet(input, output, &pkt, verbose);
+		if (write_res == -1) {
+			printf("write failed\n");
+			vs_destroy_input(input);
+			vs_destroy_output(output);
+			av_packet_unref(&pkt);
+			return 1;
+		}
+
+		av_packet_unref(&pkt);
 
 		i++;
 		if (i == max_frames) {
@@ -47,7 +72,8 @@ int main(const int argc, const char * const * const argv)
 		}
 	}
 
-	vs_destroy(vs);
+	vs_destroy_input(input);
+	vs_destroy_output(output);
 
 	return 0;
 }

videostreamer.c

@@ -34,77 +34,55 @@ vs_setup(void)
 	avformat_network_init();
 }
 
-struct Videostreamer *
-vs_open(const char * const input_format_name, const char * const input_url,
-		const char * const output_format_name, const char * const output_url,
-		const bool verbose)
+struct VSInput *
+vs_open_input(const char * const input_format_name,
+		const char * const input_url, const bool verbose)
 {
 	if (!input_format_name || strlen(input_format_name) == 0 ||
-			!input_url || strlen(input_url) == 0 ||
-			!output_format_name || strlen(output_format_name) == 0 ||
-			!output_url || strlen(output_url) == 0) {
+			!input_url || strlen(input_url) == 0) {
 		printf("%s\n", strerror(EINVAL));
 		return NULL;
 	}
 
-	struct Videostreamer * const vs = calloc(1, sizeof(struct Videostreamer));
-	if (!vs) {
+	struct VSInput * const input = calloc(1, sizeof(struct VSInput));
+	if (!input) {
 		printf("%s\n", strerror(errno));
 		return NULL;
 	}
 
 
-	// Open input.
-
 	AVInputFormat * const input_format = av_find_input_format(input_format_name);
 	if (!input_format) {
 		printf("input format not found\n");
-		vs_destroy(vs);
+		vs_destroy_input(input);
 		return NULL;
 	}
 
-	if (avformat_open_input(&vs->input_format_ctx, input_url, input_format,
+	if (avformat_open_input(&input->format_ctx, input_url, input_format,
 				NULL) != 0) {
 		printf("unable to open input\n");
-		vs_destroy(vs);
+		vs_destroy_input(input);
 		return NULL;
 	}
 
-	if (avformat_find_stream_info(vs->input_format_ctx, NULL) < 0) {
+	if (avformat_find_stream_info(input->format_ctx, NULL) < 0) {
 		printf("failed to find stream info\n");
-		vs_destroy(vs);
+		vs_destroy_input(input);
 		return NULL;
 	}
 
+
 	if (verbose) {
-		av_dump_format(vs->input_format_ctx, 0, input_url, 0);
+		av_dump_format(input->format_ctx, 0, input_url, 0);
 	}
 
 
-	// Open output.
+	// Find the first video stream.
 
-	AVOutputFormat * const output_format = av_guess_format(output_format_name,
-			NULL, NULL);
-	if (!output_format) {
-		printf("output format not found\n");
-		vs_destroy(vs);
-		return NULL;
-	}
+	input->video_stream_index = -1;
 
-	if (avformat_alloc_output_context2(&vs->output_format_ctx, output_format,
-				NULL, NULL) < 0) {
-		printf("unable to create output context\n");
-		vs_destroy(vs);
-		return NULL;
-	}
-
-
-	// Copy the first video stream.
-
-	vs->video_stream_index = -1;
-
-	for (unsigned int i = 0; i < vs->input_format_ctx->nb_streams; i++) {
-		AVStream * const in_stream = vs->input_format_ctx->streams[i];
+	for (unsigned int i = 0; i < input->format_ctx->nb_streams; i++) {
+		AVStream * const in_stream = input->format_ctx->streams[i];
 
 		if (in_stream->codecpar->codec_type != AVMEDIA_TYPE_VIDEO) {
 			if (verbose) {
@@ -113,41 +91,100 @@ vs_open(const char * const input_format_name, const char * const input_url,
 			continue;
 		}
 
-		AVStream * const out_stream = avformat_new_stream(vs->output_format_ctx,
-				NULL);
-		if (!out_stream) {
-			printf("unable to add stream\n");
-			vs_destroy(vs);
+		// I take the first video stream only.
+		input->video_stream_index = (int) i;
+		break;
+	}
+
+	if (input->video_stream_index == -1) {
+		printf("no video stream found\n");
+		vs_destroy_input(input);
 		return NULL;
 	}
 
+
+	return input;
+}
+
+void
+vs_destroy_input(struct VSInput * const input)
+{
+	if (!input) {
+		return;
+	}
+
+	if (input->format_ctx) {
+		avformat_close_input(&input->format_ctx);
+		avformat_free_context(input->format_ctx);
+	}
+
+	free(input);
+}
+
+struct VSOutput *
+vs_open_output(const char * const output_format_name,
+		const char * const output_url, const struct VSInput * const input,
+		const bool verbose)
+{
+	if (!output_format_name || strlen(output_format_name) == 0 ||
+			!output_url || strlen(output_url) == 0 ||
+			!input) {
+		printf("%s\n", strerror(EINVAL));
+		return NULL;
+	}
+
+	struct VSOutput * const output = calloc(1, sizeof(struct VSOutput));
+	if (!output) {
+		printf("%s\n", strerror(errno));
+		return NULL;
+	}
+
+
+	AVOutputFormat * const output_format = av_guess_format(output_format_name,
+			NULL, NULL);
+	if (!output_format) {
+		printf("output format not found\n");
+		vs_destroy_output(output);
+		return NULL;
+	}
+
+	if (avformat_alloc_output_context2(&output->format_ctx, output_format,
+				NULL, NULL) < 0) {
+		printf("unable to create output context\n");
+		vs_destroy_output(output);
+		return NULL;
+	}
+
+
+	// Copy the video stream.
+
+	AVStream * const out_stream = avformat_new_stream(output->format_ctx, NULL);
+	if (!out_stream) {
+		printf("unable to add stream\n");
+		vs_destroy_output(output);
+		return NULL;
+	}
+
+	AVStream * const in_stream = input->format_ctx->streams[
+		input->video_stream_index];
+
 	if (avcodec_parameters_copy(out_stream->codecpar,
 				in_stream->codecpar) < 0) {
 		printf("unable to copy codec parameters\n");
-			vs_destroy(vs);
+		vs_destroy_output(output);
 		return NULL;
 	}
 
-		// I take the first video stream only.
-		vs->video_stream_index = (int) i;
-		break;
-	}
-
-	if (vs->video_stream_index == -1) {
-		printf("no video stream found\n");
-		vs_destroy(vs);
-		return NULL;
-	}
 
 	if (verbose) {
-		av_dump_format(vs->output_format_ctx, 0, output_url, 1);
+		av_dump_format(output->format_ctx, 0, output_url, 1);
 	}
 
 
 	// Open output file.
-	if (avio_open(&vs->output_format_ctx->pb, output_url, AVIO_FLAG_WRITE) < 0) {
+	if (avio_open(&output->format_ctx->pb, output_url, AVIO_FLAG_WRITE) < 0) {
 		printf("unable to open output file\n");
-		vs_destroy(vs);
+		vs_destroy_output(output);
 		return NULL;
 	}
 
@@ -172,20 +209,20 @@ vs_open(const char * const input_format_name, const char * const input_url,
 	// empty_moov apparently writes some info at the start of the file.
 	if (av_dict_set(&opts, "movflags", "frag_keyframe+empty_moov", 0) < 0) {
 		printf("unable to set movflags opt\n");
-		vs_destroy(vs);
+		vs_destroy_output(output);
 		return NULL;
 	}
 
 	if (av_dict_set_int(&opts, "flush_packets", 1, 0) < 0) {
 		printf("unable to set flush_packets opt\n");
-		vs_destroy(vs);
+		vs_destroy_output(output);
 		av_dict_free(&opts);
 		return NULL;
 	}
 
-	if (avformat_write_header(vs->output_format_ctx, &opts) < 0) {
+	if (avformat_write_header(output->format_ctx, &opts) < 0) {
 		printf("unable to write header\n");
-		vs_destroy(vs);
+		vs_destroy_output(output);
 		av_dict_free(&opts);
 		return NULL;
 	}
@@ -195,7 +232,7 @@ vs_open(const char * const input_format_name, const char * const input_url,
 	// appropriate to set through the avformat_write_header().
 	if (av_dict_count(opts) != 0) {
 		printf("some options not set\n");
-		vs_destroy(vs);
+		vs_destroy_output(output);
 		av_dict_free(&opts);
 		return NULL;
 	}
@@ -203,103 +240,120 @@ vs_open(const char * const input_format_name, const char * const input_url,
 	av_dict_free(&opts);
 
 
-	return vs;
+	return output;
 }
 
 void
-vs_destroy(struct Videostreamer * const vs)
+vs_destroy_output(struct VSOutput * const output)
 {
-	if (!vs) {
+	if (!output) {
 		return;
 	}
 
-	if (vs->input_format_ctx) {
-		avformat_close_input(&vs->input_format_ctx);
-		avformat_free_context(vs->input_format_ctx);
-	}
-
-	if (vs->output_format_ctx) {
-		// Write file trailer.
-		if (av_write_trailer(vs->output_format_ctx) != 0) {
+	if (output->format_ctx) {
+		if (av_write_trailer(output->format_ctx) != 0) {
 			printf("unable to write trailer\n");
 		}
 
-		if (avio_closep(&vs->output_format_ctx->pb) != 0) {
+		if (avio_closep(&output->format_ctx->pb) != 0) {
 			printf("avio_closep failed\n");
 		}
 
-		avformat_free_context(vs->output_format_ctx);
+		avformat_free_context(output->format_ctx);
 	}
 
-	free(vs);
+	free(output);
 }
 
+// Read a compressed and encoded frame as a packet.
+//
+// Returns:
+// -1 if error
+// 0 if nothing useful read (e.g., non-video packet)
+// 1 if read a packet
 int
-vs_read_write(const struct Videostreamer * vs, const bool verbose)
+vs_read_packet(const struct VSInput * input, AVPacket * const pkt,
+		const bool verbose)
 {
-	if (!vs) {
+	if (!input || !pkt) {
 		printf("%s\n", strerror(errno));
 		return -1;
 	}
 
+	memset(pkt, 0, sizeof(AVPacket));
+
 
 	// Read encoded frame (as a packet).
-	AVPacket pkt;
-	memset(&pkt, 0, sizeof(AVPacket));
 
-	if (av_read_frame(vs->input_format_ctx, &pkt) != 0) {
+	if (av_read_frame(input->format_ctx, pkt) != 0) {
 		printf("unable to read frame\n");
 		return -1;
 	}
 
 
 	// Ignore it if it's not our video stream.
-	if (pkt.stream_index != vs->video_stream_index) {
+
+	if (pkt->stream_index != input->video_stream_index) {
 		if (verbose) {
 			printf("skipping non video packet\n");
 		}
+
+		av_packet_unref(pkt);
 		return 0;
 	}
 
 
-	AVStream * const in_stream  = vs->input_format_ctx->streams[pkt.stream_index];
-	AVStream * const out_stream =
-		vs->output_format_ctx->streams[pkt.stream_index];
-
 	if (verbose) {
-		__vs_log_packet(vs->input_format_ctx, &pkt, "in");
+		__vs_log_packet(input->format_ctx, pkt, "in");
 	}
 
+	return 1;
+}
 
-	// Copy packet
-	pkt.pts = av_rescale_q_rnd(pkt.pts, in_stream->time_base,
+// We change the packet's pts, dts, duration, pos.
+//
+// We do not unref it.
+//
+// Returns:
+// -1 if error
+// 1 if wrote packet
+int
+vs_write_packet(const struct VSInput * const input,
+		const struct VSOutput * const output, AVPacket * const pkt,
+		const bool verbose)
+{
+	if (!input || !output || !pkt) {
+		printf("%s\n", strerror(EINVAL));
+		return -1;
+	}
+
+	AVStream * const in_stream  = input->format_ctx->streams[pkt->stream_index];
+	AVStream * const out_stream = output->format_ctx->streams[pkt->stream_index];
+
+	pkt->pts = av_rescale_q_rnd(pkt->pts, in_stream->time_base,
 			out_stream->time_base, AV_ROUND_NEAR_INF|AV_ROUND_PASS_MINMAX);
-	pkt.dts = av_rescale_q_rnd(pkt.dts, in_stream->time_base,
+	pkt->dts = av_rescale_q_rnd(pkt->dts, in_stream->time_base,
 			out_stream->time_base, AV_ROUND_NEAR_INF|AV_ROUND_PASS_MINMAX);
-	pkt.duration = av_rescale_q(pkt.duration, in_stream->time_base,
+	pkt->duration = av_rescale_q(pkt->duration, in_stream->time_base,
 			out_stream->time_base);
-	pkt.pos = -1;
+	pkt->pos = -1;
+
 
 	if (verbose) {
-		__vs_log_packet(vs->output_format_ctx, &pkt, "out");
+		__vs_log_packet(output->format_ctx, pkt, "out");
 	}
 
 
-	const int pkt_size = pkt.size;
-
 	// Write encoded frame (as a packet).
 
 	// av_interleaved_write_frame() works too, but I don't think it is needed.
 	// Using av_write_frame() skips buffering.
-	if (av_write_frame(vs->output_format_ctx, &pkt) != 0) {
+	if (av_write_frame(output->format_ctx, pkt) != 0) {
 		printf("unable to write frame\n");
-		av_packet_unref(&pkt);
 		return -1;
 	}
 
-	av_packet_unref(&pkt);
-
-	return pkt_size;
+	return 1;
 }
 
 static void

videostreamer.go

@@ -250,27 +250,34 @@ func videoEncoder(outPipe *os.File, inputFormat, inputURL string,
 	stopChan <-chan struct{}, doneChan chan<- struct{}, frameChan chan<- int) {
 	inputFormatC := C.CString(inputFormat)
 	inputURLC := C.CString(inputURL)
-	outputFormatC := C.CString("mp4")
-	outputURLC := C.CString(fmt.Sprintf("pipe:%d", outPipe.Fd()))
 	verbose := C.bool(true)
 
-	vs := C.vs_open(inputFormatC, inputURLC, outputFormatC, outputURLC, verbose)
-	if vs == nil {
+	input := C.vs_open_input(inputFormatC, inputURLC, verbose)
+	if input == nil {
 		log.Printf("Unable to open input")
 		C.free(unsafe.Pointer(inputFormatC))
 		C.free(unsafe.Pointer(inputURLC))
-		C.free(unsafe.Pointer(outputFormatC))
-		C.free(unsafe.Pointer(outputURLC))
 		doneChan <- struct{}{}
 		return
 	}
 	C.free(unsafe.Pointer(inputFormatC))
 	C.free(unsafe.Pointer(inputURLC))
+	defer C.vs_destroy_input(input)
+
+	outputFormatC := C.CString("mp4")
+	outputURLC := C.CString(fmt.Sprintf("pipe:%d", outPipe.Fd()))
+
+	output := C.vs_open_output(outputFormatC, outputURLC, input, verbose)
+	if output == nil {
+		log.Printf("Unable to open output")
 		C.free(unsafe.Pointer(outputFormatC))
 		C.free(unsafe.Pointer(outputURLC))
-	defer C.vs_destroy(vs)
-
-	frameChan <- 792
+		doneChan <- struct{}{}
+		return
+	}
+	C.free(unsafe.Pointer(outputFormatC))
+	C.free(unsafe.Pointer(outputURLC))
+	defer C.vs_destroy_output(output)
 
 	for {
 		select {
@@ -282,21 +289,33 @@ func videoEncoder(outPipe *os.File, inputFormat, inputURL string,
 		default:
 		}
 
-		frameSize := C.int(0)
-		frameSize = C.vs_read_write(vs, verbose)
-		if frameSize == -1 {
-			log.Printf("Failure decoding/encoding")
+		var pkt C.AVPacket
+		readRes := C.int(0)
+		readRes = C.vs_read_packet(input, &pkt, verbose)
+		if readRes == -1 {
+			log.Printf("Failure reading")
 			doneChan <- struct{}{}
 			return
 		}
 
-		// We did some work.
-
-		log.Printf("frame size %d", frameSize)
-
-		if frameSize > 0 {
-			frameChan <- int(frameSize)
+		if readRes == 0 {
+			continue
 		}
+
+		writeRes := C.int(0)
+		writeRes = C.vs_write_packet(input, output, &pkt, verbose)
+		if writeRes == -1 {
+			log.Printf("Failure writing")
+			C.av_packet_unref(&pkt)
+			doneChan <- struct{}{}
+			return
+		}
+
+		pktSize := int(pkt.size)
+
+		C.av_packet_unref(&pkt)
+
+		frameChan <- int(pktSize)
 	}
 }
 

videostreamer.h

@@ -4,24 +4,40 @@
 #include <libavformat/avformat.h>
 #include <stdbool.h>
 
-struct Videostreamer {
-	AVFormatContext * input_format_ctx;
-	AVFormatContext * output_format_ctx;
+struct VSInput {
+	AVFormatContext * format_ctx;
 	int video_stream_index;
 };
 
+struct VSOutput {
+	AVFormatContext * format_ctx;
+};
+
 void
 vs_setup(void);
 
-struct Videostreamer *
-vs_open(const char * const, const char * const,
-		const char * const, const char * const,
+struct VSInput *
+vs_open_input(const char * const,
+		const char * const, const bool);
+
+void
+vs_destroy_input(struct VSInput * const);
+
+struct VSOutput *
+vs_open_output(const char * const,
+		const char * const, const struct VSInput * const,
 		const bool);
 
 void
-vs_destroy(struct Videostreamer * const);
+vs_destroy_output(struct VSOutput * const);
 
 int
-vs_read_write(const struct Videostreamer *, const bool);
+vs_read_packet(const struct VSInput *, AVPacket * const,
+		const bool);
+
+int
+vs_write_packet(const struct VSInput * const,
+		const struct VSOutput * const, AVPacket * const,
+		const bool);
 
 #endif