preserve trailing newlines in ipynb

Principally important for distinguishing a sequence of `print n,` messages
from a sequence of `print n` messages.

IPython/nbformat/v3/rwbase.py

@@ -86,17 +86,17 b' def split_lines(nb):'
         for cell in ws.cells:
             if cell.cell_type == 'code':
                 if 'input' in cell and isinstance(cell.input, basestring):
-                    cell.input = cell.input.splitlines()
+                    cell.input = (cell.input + '\n').splitlines()
                 for output in cell.outputs:
                     for key in _multiline_outputs:
                         item = output.get(key, None)
                         if isinstance(item, basestring):
-                            output[key] = item.splitlines()
+                            output[key] = (item + '\n').splitlines()
             else: # text, heading cell
                 for key in ['source', 'rendered']:
                     item = cell.get(key, None)
                     if isinstance(item, basestring):
-                        cell[key] = item.splitlines()
+                        cell[key] = (item + '\n').splitlines()
     return nb
 
 # b64 encode/decode are never actually used, because all bytes objects in

IPython/nbformat/v3/tests/nbexamples.py

@@ -49,6 +49,14 b' ws.cells.append(new_code_cell('
     prompt_number=2,
     collapsed=True
 ))
+ws.cells.append(new_code_cell(
+    input='a = 10\nb = 5\n',
+    prompt_number=3,
+))
+ws.cells.append(new_code_cell(
+    input='a = 10\nb = 5',
+    prompt_number=4,
+))
 
 ws.cells.append(new_code_cell(
     input=u'print "ünîcødé"',
@@ -122,6 +130,16 b' a = numpy.random.rand(100)'
 
 # <codecell>
 
+a = 10
+b = 5
+
+# <codecell>
+
+a = 10
+b = 5
+
+# <codecell>
+
 print "ünîcødé"
 
 """ % nbformat

IPython/nbformat/v3/tests/test_nbpy.py

@@ -28,6 +28,11 b' class TestPy(formattest.NBFormatTestCase):'
             for a,b in zip(da, db):
                 self.assertSubset(a,b)
         else:
+            if isinstance(da, basestring) and isinstance(db, basestring):
+                # pyfile is not sensitive to preserving leading/trailing
+                # newlines in blocks through roundtrip
+                da = da.strip('\n')
+                db = db.strip('\n')
             self.assertEquals(da, db)
         return True