BUG: Use stdlib csv module to write csv, so commas in data are escaped

correctly.

Author: takluyver

pandas/core/frame.py

@@ -13,6 +13,7 @@
 # pylint: disable=W0212,W0231,W0703,W0622
 
 from StringIO import StringIO
+import csv
 import operator
 import warnings
 
@@ -485,13 +486,13 @@ def to_csv(self, path, nanRep='', cols=None, header=True,
         mode : Python write mode, default 'wb'
         """
         f = open(path, mode)
+        csvout = csv.writer(f)
 
         if cols is None:
             cols = self.columns
 
         series = self._series
         if header:
-            joined_cols = ','.join([str(c) for c in cols])
             if index:
                 # should write something for index label
                 if index_label is None:
@@ -509,31 +510,26 @@ def to_csv(self, path, nanRep='', cols=None, header=True,
                 elif not isinstance(index_label, (list, tuple, np.ndarray)):
                     # given a string for a DF with Index
                     index_label = [index_label]
-                f.write('%s,%s' % (",".join(index_label), joined_cols))
+                csvout.writerow(list(index_label) + list(cols))
             else:
-                f.write(joined_cols)
-            f.write('\n')
+                csvout.writerow(cols)
 
         nlevels = getattr(self.index, 'nlevels', 1)
         for idx in self.index:
+            row_fields = []
             if index:
                 if nlevels == 1:
-                    f.write(str(idx))
+                    row_fields = [idx]
                 else: # handle MultiIndex
-                    f.write(",".join([str(i) for i in idx]))
+                    row_fields = list(idx)
             for i, col in enumerate(cols):
                 val = series[col].get(idx)
                 if isnull(val):
                     val = nanRep
-                else:
-                    val = str(val)
 
-                if i > 0 or index:
-                    f.write(',%s' % val)
-                else:
-                    f.write('%s' % val)
+                row_fields.append(val)
 
-            f.write('\n')
+            csvout.writerow(row_fields)
 
         f.close()
 

pandas/core/series.py

@@ -5,6 +5,7 @@
 # pylint: disable=E1101,E1103
 # pylint: disable=W0703,W0622,W0613,W0201
 
+import csv
 import itertools
 import operator
 import sys
@@ -1589,8 +1590,8 @@ def to_csv(self, path):
             Output filepath. If None, write to stdout
         """
         f = open(path, 'wb')
-        for idx, value in self.iteritems():
-            f.write(str(idx) + ',' + str(value) + '\n')
+        csvout = csv.writer(f)
+        csvout.writerows(self.iteritems())
         f.close()
 
     def dropna(self):

pandas/tests/test_frame.py

@@ -1582,6 +1582,16 @@ def test_to_csv_float32_nanrep(self):
         lines = open(pth).readlines()
         self.assert_(lines[1].split(',')[2] == '999')
         os.remove(pth)
+    
+    def test_to_csv_withcommas(self):
+        "Commas inside fields should be correctly escaped when saving as CSV."
+        path = '__tmp__'
+        df = DataFrame({'A':[1,2,3], 'B':['5,6','7,8','9,0']})
+        df.to_csv(path)
+        df2 = DataFrame.from_csv(path)
+        assert_frame_equal(df2, df)
+        
+        os.remove(path)
 
     def test_info(self):
         io = StringIO()