From d9edf2a74e4d453b583ef255d1b0559c83c296c9 Mon Sep 17 00:00:00 2001
From: georgeG <georgkam@gmail.com>
Date: Mon, 11 Mar 2013 15:42:17 +0300
Subject: [PATCH] write the parser

---
 lib/bio-cd-hit-report.rb               | 12 ++++++-
 lib/bio-cd-hit-report/cd-hit-report.rb | 49 +++++---------------------
 lib/bio-cd-hit-report/cluster.rb       | 30 ++++++++--------
 lib/bio-cd-hit-report/parser.rb        | 19 ++++++++++
 4 files changed, 54 insertions(+), 56 deletions(-)
 create mode 100644 lib/bio-cd-hit-report/parser.rb

diff --git a/lib/bio-cd-hit-report.rb b/lib/bio-cd-hit-report.rb
index 8f7c311..d0f5c5e 100644
--- a/lib/bio-cd-hit-report.rb
+++ b/lib/bio-cd-hit-report.rb
@@ -1 +1,11 @@
-require_relative "bio-cd-hit-report/cd-hit-report"
+require_relative 'bio-cd-hit-report/cd-hit-report'
+
+#report = Bio::CdHitReport.new('bin/test.clstr')
+##report.report_file = 'bin/test.clstr'
+#report.parse.each do |cluster|
+  #puts cluster.id
+  ##puts cluster.members
+  #puts cluster.rep_seq
+#end
+
+#puts report.parse.length
diff --git a/lib/bio-cd-hit-report/cd-hit-report.rb b/lib/bio-cd-hit-report/cd-hit-report.rb
index 12d6b05..6f10ca1 100644
--- a/lib/bio-cd-hit-report/cd-hit-report.rb
+++ b/lib/bio-cd-hit-report/cd-hit-report.rb
@@ -1,49 +1,16 @@
 module Bio
-
-  require_relative 'cluster.rb'
+  require_relative 'cluster'
+  require_relative 'parser'
 
   class CdHitReport
-
     def initialize(file)
       @file = file
     end
 
-    def each_cluster(&block)
-      cluster_objs.each(&block)
-    end
-
-    def total_clusters
-      cluster_objs.size
-    end
-
-    def get_cluster(name)
-      cluster_objs.select{|cluster| cluster.name == name.to_s}.pop.members
-    end
-
-    def max_members
-      cluster_objs.map{|c|c.size}.max
+    def parse
+      report = CdHitParser.new
+      report.report_file = @file
+      report
     end
-
-    def min_members
-      cluster_objs.map{|c| c.size}.min
-    end
-
-    private
-    def cluster_objs
-      d = raw_data.map do |line|
-        cluster = line.split("\n").delete_if{|x| x == ">Cluster "}
-        id = cluster.first
-        cluster.shift
-        #puts id.inspect
-        Cluster.new(id,cluster)
-      end
-      d.delete_if {|obj| obj.id.nil?}
-    end
-
-
-    def raw_data
-      File.open(@file).readlines
-    end
-
-  end #class
-end #module
+  end
+end
diff --git a/lib/bio-cd-hit-report/cluster.rb b/lib/bio-cd-hit-report/cluster.rb
index 14c67fd..c883a55 100644
--- a/lib/bio-cd-hit-report/cluster.rb
+++ b/lib/bio-cd-hit-report/cluster.rb
@@ -1,28 +1,30 @@
+class Cluster
+  attr_accessor :name, :data
 
-class Cluster < Struct.new(:name,:data)
-  $/ = ">Cluster "
-
-  def id
-    name
+  def initialize(arg={})
+    @name = arg[:name]
+    @data = arg[:data] 
   end
 
-  def size
-    entries.size
+  def id
+    name.scan(/Cluster\s(.)/).join
   end
 
   def members
     entries.join(',')
   end
 
-  def get_seqs(file)
-    seqs = Bio::FlatFile.auto(file).map{ |f| f}
-    puts entries.map{|entry| seqs.select {|seq| seq.definition == entry }}
+  def representative
+    @data.split("\n").map{|line|line.scan(/>(.+)\.{3}\s\*/)}.flatten
   end
+  alias :rep_seq :representative
 
-  private
-  def entries
-    data.map {|entry| entry.scan(/>(.+)\.{3}/)}.flatten
+  def size
+    entries.size
   end
+  alias :length :size
 
+  def entries
+    @data.split("\n").map{|line|line.scan(/>(.+)\.{3}/)}
+  end
 end
-
diff --git a/lib/bio-cd-hit-report/parser.rb b/lib/bio-cd-hit-report/parser.rb
new file mode 100644
index 0000000..294f8f7
--- /dev/null
+++ b/lib/bio-cd-hit-report/parser.rb
@@ -0,0 +1,19 @@
+class CdHitParser
+  attr_accessor :report_file
+
+  include Enumerable
+
+  def each
+    data,header = nil, nil
+    File.open(report_file).each do |line|
+      if line[0].chr == '>'
+        yield Cluster.new(:name => header,:data => data) if data
+        data = ''
+        header = line[1..-1].strip
+      else
+        data << line
+      end
+    end
+    yield Cluster.new(:name => header, :data => data)
+  end
+end