FIX: try more reliable way to get string length

Fixed width byte/unicode dtypes seem to have the pattern
"[|<>][US][1-9][0-9]*" thus if we drop the first two charters we will
have the fixed width.

Author: tacaswell

lib/matplotlib/mlab.py

@@ -3188,7 +3188,8 @@ def get_justify(colname, column, precision):
         ntype = column.dtype
 
         if np.issubdtype(ntype, str) or np.issubdtype(ntype, bytes):
-            length = max(len(colname), len(column[0]))
+            fixed_width = int(ntype.str[2:])
+            length = max(len(colname), fixed_width)
             return 0, length+padding, "%s"  # left justify
 
         if np.issubdtype(ntype, np.int):