From 940bb8d73805da09c8f90af767dbeef1ad89cbfa Mon Sep 17 00:00:00 2001
From: Daniel Blankenberg <dan@bx.psu.edu>
Date: Sun, 25 Sep 2011 13:27:14 -0400
Subject: [PATCH] Some random lines tool tweaks.

---
 tools/filters/random_lines_two_pass.py | 40 +++++++++++++++++++-------
 tools/filters/randomlines.xml          |  2 +-
 2 files changed, 31 insertions(+), 11 deletions(-)

diff --git a/tools/filters/random_lines_two_pass.py b/tools/filters/random_lines_two_pass.py
index 7d0b2bb7758..25ba02d1c1a 100644
--- a/tools/filters/random_lines_two_pass.py
+++ b/tools/filters/random_lines_two_pass.py
@@ -2,9 +2,25 @@
 #Dan Blankenberg
 #Selects N random lines from a file and outputs to another file, maintaining original line order
 #allows specifying a seed
-#does two passes to determine line counts and offsets, and then to output contents
+#does two passes to determine line offsets/count, and then to output contents
 
-import optparse, random 
+import optparse, random
+
+def get_random_by_subtraction( line_offsets, num_lines ):
+    while len( line_offsets ) > num_lines:
+        del line_offsets[ random.randint( 0, len( line_offsets ) - 1 ) ]
+    return line_offsets
+
+def get_random_by_sample( line_offsets, num_lines ):
+    line_offsets = random.sample( line_offsets, num_lines )
+    line_offsets.sort()
+    return line_offsets
+
+def get_random( line_offsets, num_lines ):
+    if num_lines > ( len( line_offsets ) / 2 ):
+        return get_random_by_subtraction( line_offsets, num_lines )
+    else:
+        return get_random_by_sample( line_offsets, num_lines )
 
 def __main__():
     #Parse Command Line
@@ -14,7 +30,7 @@ def __main__():
     
     assert len( args ) == 3, "Invalid command line specified."
     
-    input = open( args[0] )
+    input = open( args[0], 'rb' )
     output = open( args[1], 'wb' )
     num_lines = int( args[2] )
     assert num_lines > 0, "You must select at least one line."
@@ -24,10 +40,13 @@ def __main__():
     
     #get line offsets
     line_offsets = []
+    teller = input.tell
+    readliner = input.readline
+    appender = line_offsets.append
     while True:
-        offset = input.tell()
-        if input.readline():
-            line_offsets.append( offset )
+        offset = teller()
+        if readliner():
+            appender( offset )
         else:
             break
     
@@ -35,13 +54,14 @@ def __main__():
     assert num_lines <= total_lines, "Error: asked to select more lines (%i) than there were in the file (%i)." % ( num_lines, total_lines )
     
     #get random line offsets
-    while len( line_offsets ) > num_lines:
-        line_offsets.pop( random.randint( 0, len( line_offsets ) - 1 ) )
+    line_offsets = get_random( line_offsets, num_lines )
     
     #write out random lines
+    seeker = input.seek
+    writer = output.write
     for line_offset in line_offsets:
-        input.seek( line_offset )
-        output.write( input.readline() )
+        seeker( line_offset )
+        writer( readliner() )
     input.close()
     output.close()
     print "Kept %i of %i total lines." % ( num_lines, total_lines )
diff --git a/tools/filters/randomlines.xml b/tools/filters/randomlines.xml
index 1089d73a613..526c66d4b59 100644
--- a/tools/filters/randomlines.xml
+++ b/tools/filters/randomlines.xml
@@ -1,4 +1,4 @@
-<tool id="random_lines1" name="Select random lines" version="2.0.0">
+<tool id="random_lines1" name="Select random lines" version="2.0.1">
   <description>from a file</description>
   <command interpreter="python">random_lines_two_pass.py "${input}" "${out_file1}" "${num_lines}"
   #if str( $seed_source.seed_source_selector ) == "set_seed":