sweeping node graph changes for new rendering pipeline

The new rendering pipeline strives to simplify the nodes themselves as much
as possible and move much of the logic to an external rendering engine. This
change removes all of the responsibilities that no longer belong to the
nodes themselves and will soon be folded into the renderer.

app/node/output.h

@@ -21,6 +21,7 @@
 #ifndef NODEOUTPUT_H
 #define NODEOUTPUT_H
 
+#include "common/timerange.h"
 #include "param.h"
 
 /**
@@ -39,30 +40,15 @@ public:
    */
   virtual Type type() override;
 
-  /**
-   * @brief Get the value of this output at a given tie
-   *
-   * This function is intended to primarily be called by any connected NodeInputs.
-   *
-   * The first thing this function does is request the parent Node object to Process() at this time. The Node should
-   * then perform whatever actions necessary (usually taking data from inputs and creating output data) to set the
-   * correct value that this output should have at this time (the Node should use set_value() for this). This function
-   * will then return the value that was set after Process() returned.
-   *
-   * In many cases for efficiency, the Node can also ignore this request if it knows the output data will not change
-   * (i.e. if the time has not changed from the last Process()).
-   */
-  virtual QVariant get_value(const rational &in, const rational &out);
+  QVariant get_realtime_value();
 
-  void push_value(const QVariant& v, const rational& in, const rational &out);
-
-  NodeInput* linked_input();
-  void set_linked_input(NodeInput* link);
+  bool has_cached_value(const TimeRange& time);
+  QVariant get_cached_value(const TimeRange& time);
+  void cache_value(const TimeRange& time, const QVariant& value);
+  void drop_cached_values();
 
 private:
-  QMutex mutex_;
-
-  NodeInput* linked_input_;
+  QMap<TimeRange, QVariant> cached_values_;
 
 };