summaryrefslogtreecommitdiffstats
path: root/src/ModPlugin.h
diff options
context:
space:
mode:
Diffstat (limited to 'src/ModPlugin.h')
-rw-r--r--src/ModPlugin.h37
1 files changed, 32 insertions, 5 deletions
diff --git a/src/ModPlugin.h b/src/ModPlugin.h
index d3aa780..c0f1c1a 100644
--- a/src/ModPlugin.h
+++ b/src/ModPlugin.h
@@ -30,16 +30,34 @@
# include <config.h>
#endif
-
#include "Buffer.h"
#include "ThreadsafeQueue.h"
-
-#include <sys/types.h>
+#include <cstddef>
#include <vector>
#include <memory>
#include <thread>
#include <atomic>
+// All flowgraph elements derive from ModPlugin, or a variant of it.
+// Some ModPlugins also support handling metadata.
+
+struct frame_timestamp;
+struct flowgraph_metadata {
+ std::shared_ptr<struct frame_timestamp> ts;
+};
+
+using meta_vec_t = std::vector<flowgraph_metadata>;
+
+/* ModPlugins that support metadata derive from ModMetadata */
+class ModMetadata {
+ public:
+ // Receives metadata from all inputs, and process them, and output
+ // a sequence of metadata.
+ virtual meta_vec_t process_metadata(const meta_vec_t& metadataIn) = 0;
+};
+
+
+/* Abstract base class for all flowgraph elements */
class ModPlugin
{
public:
@@ -69,7 +87,11 @@ public:
virtual int process(Buffer* const dataIn, Buffer* dataOut) = 0;
};
-class PipelinedModCodec : public ModCodec
+/* Pipelined ModCodecs run their processing in a separate thread, and
+ * have a one-call-to-process() latency. Because of this latency, they
+ * must also handle the metadata
+ */
+class PipelinedModCodec : public ModCodec, public ModMetadata
{
public:
PipelinedModCodec();
@@ -82,6 +104,8 @@ public:
virtual int process(Buffer* const dataIn, Buffer* dataOut) final;
virtual const char* name() = 0;
+ virtual meta_vec_t process_metadata(const meta_vec_t& metadataIn) final;
+
protected:
// Once the instance implementing PipelinedModCodec has been constructed,
// it must call start_pipeline_thread()
@@ -89,11 +113,13 @@ protected:
virtual int internal_process(Buffer* const dataIn, Buffer* dataOut) = 0;
private:
- size_t m_number_of_runs;
+ bool m_ready_to_output_data = false;
ThreadsafeQueue<std::shared_ptr<Buffer> > m_input_queue;
ThreadsafeQueue<std::shared_ptr<Buffer> > m_output_queue;
+ std::deque<meta_vec_t> m_metadata_fifo;
+
std::atomic<bool> m_running;
std::thread m_thread;
void process_thread(void);
@@ -119,3 +145,4 @@ public:
std::vector<Buffer*> dataOut);
virtual int process(Buffer* dataIn) = 0;
};
+