Implement an optional name in the motif parser

ticket:116
Using the spirit parser the motif file now takes a sequence, optional name,
and the RGB float colors.

The down side to this is I don't currently know how to get spirit to report
parse errors.

diff --git a/alg/mussa.cpp b/alg/mussa.cpp
index cc5825ef68999dfacbe6370b3f38fbe0e00df20d..ccba221f61ca45a8007906046ccd2a07209eb510 100644 (file)
--- a/alg/mussa.cpp
+++ b/alg/mussa.cpp
@@ -16,6 +16,12 @@
 #include <boost/filesystem/fstream.hpp>
 namespace fs = boost::filesystem;
 
+#include <boost/spirit/core.hpp>
+#include <boost/spirit/actor/push_back_actor.hpp>
+#include <boost/spirit/iterator/file_iterator.hpp>
+#include <boost/spirit/utility/chset.hpp>
+namespace spirit = boost::spirit;
+
 #include <iostream>
 #include <sstream>
 
@@ -737,54 +743,86 @@ void Mussa::set_motifs(const vector<Sequence>& motifs,
   update_sequences_motifs();
 }
 
+
+// Helper functor to append created motifs to our Mussa analysis
+struct push_back_motif {
+  std::set<Sequence>& motif_set;
+  boost::shared_ptr<AnnotationColors> color_mapper;
+  std::string& seq_string;
+  std::string& name;
+  float& red;
+  float& green;
+  float& blue;
+
+  push_back_motif(std::set<Sequence>& motif_set_,
+                  boost::shared_ptr<AnnotationColors> color_mapper_,
+                  std::string& seq_, 
+                  std::string& name_,
+                  float red_, float green_, float blue_)
+    : motif_set(motif_set_),
+      color_mapper(color_mapper_),
+      seq_string(seq_),
+      name(name_),
+      red(red_),
+      green(green_),
+      blue(blue_)
+  {
+  }
+
+  void operator()(std::string::const_iterator, 
+                  std::string::const_iterator) const 
+  {
+    //std::cout << "motif: " << seq_string << "/" << name << endl;
+
+    Sequence seq(seq_string);
+    // shouldn't we have a better field than "fasta header" and speices?
+    seq.set_fasta_header(name);
+    // we need to clear the name in case the next motif doesn't have one.
+    name.clear();
+    // be nice if glsequence was a subclass of sequence so we could
+    // just attach colors directly to the motif.
+    Color c(red, green, blue);
+    color_mapper->appendInstanceColor("motif", seq.c_str(), c);
+    motif_set.insert(seq);
+  };
+};
+
 // I mostly split the ifstream out so I can use a stringstream to test it.
 void Mussa::load_motifs(std::istream &in)
 {
+  std::string data;
+  const char *alphabet = Sequence::nucleic_iupac_alphabet.c_str();
   string seq;
+  string name;
   float red;
   float green;
   float blue;
 
-  while(in.good())
-  {
-    in >> seq >> red >> green >> blue;
-    // if we couldn't read this line 'cause we're like at the end of the file
-    // try to exit the loop
-    if (!in.good())
-      break;
-    try {
-      seq = Sequence::motif_normalize(seq);
-    } catch(motif_normalize_error e) {
-      clog << "unable to parse " << seq << " skipping" << endl;
-      clog << e.what() << endl;
-      continue;
-    }
-    if (red < 0.0 or red > 1.0) {
-      clog << "invalid red value " << red << ". must be in range [0..1]" 
-           << endl;
-      continue;
-    }
-    if (green < 0.0 or green > 1.0) {
-      clog << "invalid green value " << green << ". must be in range [0..1]" 
-           << endl;
-      continue;
-    }
-    if (blue < 0.0 or blue > 1.0) {
-      clog << "invalid blue value " << blue << ". must be in range [0..1]" 
-           << endl;
-      continue;
-    }
-    if (motif_sequences.find(seq) == motif_sequences.end()) {
-      // sequence wasn't found
-      motif_sequences.insert(seq);
-      Color c(red, green, blue);
-      color_mapper->appendInstanceColor("motif", seq, c);
-    } else {
-      clog << "sequence " << seq << " was already defined skipping" 
-           << endl;
-      continue;
-    }
+  // slurp our data into a string
+  std::streamsize bytes_read = 1;
+  while (in.good() and bytes_read) {
+    const std::streamsize bufsiz=512;
+    char buf[bufsiz];
+    bytes_read = in.readsome(buf, bufsiz);
+    data.append(buf, buf+bytes_read);
   }
+  // parse our string
+  bool status = spirit::parse(data.begin(), data.end(),
+     *( 
+       ( 
+        (
+         (+spirit::chset<>(alphabet))[spirit::assign_a(seq)] >> 
+         +spirit::space_p
+        ) >>
+        !(
+          (spirit::alpha_p >> *spirit::alnum_p)[spirit::assign_a(name)]
+          >> +spirit::space_p
+        ) >>
+        spirit::real_p[spirit::assign_a(red)] >> +spirit::space_p >>
+        spirit::real_p[spirit::assign_a(green)] >> +spirit::space_p >>
+        spirit::real_p[spirit::assign_a(blue)] >> +spirit::space_p
+       )[push_back_motif(motif_sequences, color_mapper, seq, name, red, green, blue)]
+     )).full;
   update_sequences_motifs();
 }
 

diff --git a/alg/test/test_mussa.cpp b/alg/test/test_mussa.cpp
index 4b88c242e5ff4b5dfb8a6342cfeea1e3009c79fb..bebb38d2548e2cec34b43d642ac73c33627ec413 100644 (file)
--- a/alg/test/test_mussa.cpp
+++ b/alg/test/test_mussa.cpp
@@ -173,7 +173,7 @@ BOOST_AUTO_TEST_CASE( mussa_load_motif )
 
   Mussa m1;
   m1.append_sequence("AAAAGGGGTTTT");
-  m1.append_sequence("GGGCCCCTTGGTT");
+  m1.append_sequence("GGGCCCCTTCCAATT");
   m1.load_motifs(test_istream);
 
   for (Mussa::vector_sequence_type::const_iterator seq_i = m1.sequences().begin();
@@ -184,6 +184,21 @@ BOOST_AUTO_TEST_CASE( mussa_load_motif )
   }
 }
 
+BOOST_AUTO_TEST_CASE( mussa_named_motif )
+{
+  string data = "CCAATT cat 0.1 0.2 0.3\n";
+  istringstream test_istream(data);
+
+  Mussa m1;
+  m1.append_sequence("AAAAGGGGTTTT");
+  m1.append_sequence("GGGCCCCTTCCAATT");
+  m1.load_motifs(test_istream);
+
+  std::set<Sequence> motifs = m1.motifs();
+  BOOST_REQUIRE_EQUAL(motifs.size(), 1);
+  BOOST_CHECK_EQUAL(motifs.begin()->get_name(), "cat");
+}
+
 BOOST_AUTO_TEST_CASE( mussa_add_motif )
 {
   vector<Sequence> motifs;