From ec150889b5b73b4911f5be92aac044aaaa69d23b 2014-04-06 01:34:25
From: MinRK <benjaminrk@gmail.com>
Date: 2014-04-06 01:34:25
Subject: [PATCH] fix n^2 performance issue in coalesce_streams preprocessor

for n consecutive stream outputs, `\r` fix would be compiled n times,
and applied to each output (n-i) times.

- move pattern to module level
- apply replacement after coalescing outputs

---

diff --git a/IPython/nbconvert/preprocessors/coalescestreams.py b/IPython/nbconvert/preprocessors/coalescestreams.py
index 37fddd1..7ccc5c6 100644
--- a/IPython/nbconvert/preprocessors/coalescestreams.py
+++ b/IPython/nbconvert/preprocessors/coalescestreams.py
@@ -40,6 +40,7 @@ def cell_preprocessor(function):
         return nb, resources
     return wrappedfunc
 
+cr_pat = re.compile(r'.*\r(?=[^\n])')
 
 @cell_preprocessor
 def coalesce_streams(cell, resources, index):
@@ -64,7 +65,6 @@ def coalesce_streams(cell, resources, index):
     
     last = outputs[0]
     new_outputs = [last]
-    
     for output in outputs[1:]:
         if (output.output_type == 'stream' and
             last.output_type == 'stream' and
@@ -72,12 +72,14 @@ def coalesce_streams(cell, resources, index):
         ):
             last.text += output.text
 
-            # Respect \r characters.
-            cr_pat = re.compile(r'.*\r(?=[^\n])')
-            last.text = cr_pat.sub('', last.text)
         else:
             new_outputs.append(output)
             last = output
+    
+    # process \r characters
+    for output in new_outputs:
+        if output.output_type == 'stream':
+            output.text = cr_pat.sub('', output.text)
 
     cell.outputs = new_outputs
     return cell, resources