From 7f6ea5e614268ff3fe605d88b6f6517ff4b41160 Mon Sep 17 00:00:00 2001
From: Ichthyostega <prg@ichthyostega.de>
Date: Sat, 30 Aug 2014 23:42:49 +0200
Subject: [PATCH] Verarbeitungsmechanismus mit Laden/Schreiben und
 Kontextbehandlung
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

implementiert und getestet
Ein Kontext-Puffer wird mitgeführt, so daß stets eine bestimmte
Anzahl an Kontext-Samples mit verfügbar ist. Puffer-Steuerung erfolgt
pro Kanal, so daß auch variierende Dateilängen korrekt gehandhabt werden;
die Verarbeitung bricht aber insgesamt innerhalb einer Pufferlänge ab,
wenn die kürzesten Samples fertig bearbeitet sind.

die eigentliche Korrektur ist noch nicht implementiert; im Test,
nach dem Beheben der Implementierungs-Fehler, waren die bearbeiteten
Samples schließlich Bit für Bit identisch mit dem Original.
---
 sauSynKlick/Makefile    |  1 +
 sauSynKlick/SynKlick.cc | 70 +++++++++++++++++++++++++++++++++--------
 2 files changed, 58 insertions(+), 13 deletions(-)

diff --git a/sauSynKlick/Makefile b/sauSynKlick/Makefile
index ef72ea9..586b376 100644
--- a/sauSynKlick/Makefile
+++ b/sauSynKlick/Makefile
@@ -30,6 +30,7 @@ SynKlick:	SynKlick.cc
 all: SynKlick
 
 clean:
+	-rm -f SynKlick
 	-rm -f `find . -name "*.o"`
 	-rm -f `find .. -name "*~"`
 	-rm -f *.bak core score.srt
diff --git a/sauSynKlick/SynKlick.cc b/sauSynKlick/SynKlick.cc
index 22ebbd1..6348ae5 100644
--- a/sauSynKlick/SynKlick.cc
+++ b/sauSynKlick/SynKlick.cc
@@ -43,7 +43,16 @@ name_without_path (string rawName)
 {
   auto pos = rawName.rfind('/');
   return (pos == string::npos)? rawName
-                              : rawName.substr(pos);
+                              : rawName.substr(pos+1);
+}
+
+void
+reduceBufferPos (size_t & pos, size_t elementsToRemove)
+{
+  if (pos < elementsToRemove)
+    pos = 0;
+  else
+    pos -= elementsToRemove;
 }
 
 
@@ -66,11 +75,13 @@ class Channel
             info_out_;
     
     float buff_[CHUNK_SIZE];
-    size_t workPos=0,
-           loadPos=0;
+    size_t numSamples_ = 0;
+    size_t workPos_    = 0;
     
     
   public:
+    static bool hitEnd;
+    
     void
     open (string name, string target)
       {
@@ -81,6 +92,9 @@ class Channel
         in_ = sf_open (name.c_str(), SFM_READ, &info_in_);
         if (!in_)
           error ("could not open "+name+" -- " + sf_strerror (NULL));
+        if (1 != info_in_.channels)
+          error ("only Mono streams supported -- but '"+name_
+                +"' was found to have "+std::to_string(info_in_.channels)+" channels.");
         
         info_out_.format = info_in_.format;
         info_out_.channels = info_in_.channels;
@@ -93,10 +107,11 @@ class Channel
     
    ~Channel()
       {
-        sf_close (in_);
-        if (!sf_close (out_))
-          cerr << "WARNING: problems while closing target '" << name_without_path(name_)
-               << "' -- " << sf_strerror(out_) << "\n";
+        if (in_) sf_close (in_);
+        if (out_)
+          if (0 != sf_close (out_))
+            cerr << "WARNING: problems while closing target '" << name_without_path(name_)
+                 << "' -- " << sf_strerror(out_) << "\n";
         
         in_ = out_ = nullptr;
       }
@@ -105,18 +120,41 @@ class Channel
     void
     fillBuffer()
       {
-        error("TODO");
+        assert(numSamples_ < CHUNK_SIZE);
+        float* startPos = buff_ + numSamples_;
+        size_t samplesToRead = CHUNK_SIZE - numSamples_;
+        numSamples_ = sf_read_float(in_, startPos, samplesToRead);
+        if (numSamples_ < samplesToRead)
+          hitEnd = true; // causes the processing to end with this chunk.
+        
+        numSamples_ += (startPos - buff_);
+        assert(workPos_ <= numSamples_);
       }
     
     void
     writeBuffer()
       {
-        error("TODO");
+        size_t samplesToWrite = numSamples_ - (hitEnd? 0 : 2*CONTEXT);
+        size_t actuallyWritten = sf_write_float(out_, buff_, samplesToWrite);
+        if (actuallyWritten < samplesToWrite)
+          cerr << "Alarm: only "<<actuallyWritten<<" instead of "<<samplesToWrite
+               <<" samples written for " << name_ <<"\n";
+        
+        if (!hitEnd)
+          // move processing context samples to the beginning of the buffer;
+          // they will be written together with the next chunk of data
+          {
+            for (uint i=0; i < 2*CONTEXT; ++i)
+              buff_[i] = buff_[i + actuallyWritten];
+            reduceBufferPos(numSamples_, actuallyWritten);
+            reduceBufferPos(workPos_,    actuallyWritten);
+          }
+        assert(workPos_ <= numSamples_);
       }
     
     /**
-     * process a single sample position in the current buffer content.
-     * The sample is copied to the output and it is investigated for glitches.
+     * investigate a single sample position in the current buffer content.
+     * The delta to the preceding sample(s) is used to detect a signal glitch.
      * @param i position of current sample
      * @param event state of other tracks investigated thus far
      * @return the event detection state including the results of investigating this sample.
@@ -127,6 +165,12 @@ class Channel
     Event
     visit (uint i, Event event)
       {
+        if (i + CONTEXT < numSamples_
+            && CONTEXT  <= i
+            && workPos_ <= i)
+          {
+            workPos_ = i;
+          }
         return event;
       }
     
@@ -167,7 +211,7 @@ class Patch
 
 int num_sources = 0;
 vector<Channel> streams(0);
-bool hitEnd = false;
+bool Channel::hitEnd = false;
 
 
 void
@@ -195,7 +239,7 @@ closeStreams()
 void
 process()
 {
-  while (!hitEnd)
+  while (!Channel::hitEnd)
     {
       for (Channel& channel : streams)
         channel.fillBuffer();