Fix: longest_substr([]) -> ''

Previously the algorithm used to find the longest common starting substring
unintentionally raised an error when it was called on an empty list.

IPython/utils/tests/test_text.py

@@ -1,170 +1,173 b''
 # encoding: utf-8
 """Tests for IPython.utils.text"""
 
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Imports
 #-----------------------------------------------------------------------------
 
 import os
 import math
 import random
 
 import nose.tools as nt
 
 from nose import with_setup
 
 from IPython.testing import decorators as dec
 from IPython.utils import text
 
 #-----------------------------------------------------------------------------
 # Globals
 #-----------------------------------------------------------------------------
 
 def test_columnize():
     """Basic columnize tests."""
     size = 5
     items = [l*size for l in 'abc']
     out = text.columnize(items, displaywidth=80)
     nt.assert_equal(out, 'aaaaa  bbbbb  ccccc\n')
     out = text.columnize(items, displaywidth=12)
     nt.assert_equal(out, 'aaaaa  ccccc\nbbbbb\n')
     out = text.columnize(items, displaywidth=10)
     nt.assert_equal(out, 'aaaaa\nbbbbb\nccccc\n')
 
 def test_columnize_random():
     """Test with random input to hopfully catch edge case """
     for nitems in [random.randint(2,70) for i in range(2,20)]:
         displaywidth = random.randint(20,200)
         rand_len = [random.randint(2,displaywidth) for i in range(nitems)]
         items = ['x'*l for l in rand_len]
         out = text.columnize(items, displaywidth=displaywidth)
         longer_line = max([len(x) for x in out.split('\n')])
         longer_element = max(rand_len)
         if longer_line > displaywidth:
             print "Columnize displayed something lager than displaywidth : %s " % longer_line
             print "longer element : %s " % longer_element
             print "displaywidth : %s " % displaywidth
             print "number of element : %s " % nitems
             print "size of each element :\n %s" % rand_len
             assert False
 
 def test_columnize_medium():
     """Test with inputs than shouldn't be wider tahn 80 """
     size = 40
     items = [l*size for l in 'abc']
     out = text.columnize(items, displaywidth=80)
     nt.assert_equal(out, '\n'.join(items+['']))
 
 def test_columnize_long():
     """Test columnize with inputs longer than the display window"""
     size = 11
     items = [l*size for l in 'abc']
     out = text.columnize(items, displaywidth=size-1)
     nt.assert_equal(out, '\n'.join(items+['']))
 
 def eval_formatter_check(f):
     ns = dict(n=12, pi=math.pi, stuff='hello there', os=os, u=u"café", b="café")
     s = f.format("{n} {n//4} {stuff.split()[0]}", **ns)
     nt.assert_equal(s, "12 3 hello")
     s = f.format(' '.join(['{n//%i}'%i for i in range(1,8)]), **ns)
     nt.assert_equal(s, "12 6 4 3 2 2 1")
     s = f.format('{[n//i for i in range(1,8)]}', **ns)
     nt.assert_equal(s, "[12, 6, 4, 3, 2, 2, 1]")
     s = f.format("{stuff!s}", **ns)
     nt.assert_equal(s, ns['stuff'])
     s = f.format("{stuff!r}", **ns)
     nt.assert_equal(s, repr(ns['stuff']))
     
     # Check with unicode:
     s = f.format("{u}", **ns)
     nt.assert_equal(s, ns['u'])
     # This decodes in a platform dependent manner, but it shouldn't error out
     s = f.format("{b}", **ns)
         
     nt.assert_raises(NameError, f.format, '{dne}', **ns)
 
 def eval_formatter_slicing_check(f):
     ns = dict(n=12, pi=math.pi, stuff='hello there', os=os)
     s = f.format(" {stuff.split()[:]} ", **ns)
     nt.assert_equal(s, " ['hello', 'there'] ")
     s = f.format(" {stuff.split()[::-1]} ", **ns)
     nt.assert_equal(s, " ['there', 'hello'] ")
     s = f.format("{stuff[::2]}", **ns)
     nt.assert_equal(s, ns['stuff'][::2])
     
     nt.assert_raises(SyntaxError, f.format, "{n:x}", **ns)
 
 def eval_formatter_no_slicing_check(f):
     ns = dict(n=12, pi=math.pi, stuff='hello there', os=os)
     
     s = f.format('{n:x} {pi**2:+f}', **ns)
     nt.assert_equal(s, "c +9.869604")
     
     s = f.format('{stuff[slice(1,4)]}', **ns)
     nt.assert_equal(s, 'ell')
     
     nt.assert_raises(SyntaxError, f.format, "{a[:]}")
 
 def test_eval_formatter():
     f = text.EvalFormatter()
     eval_formatter_check(f)
     eval_formatter_no_slicing_check(f)
 
 def test_full_eval_formatter():
     f = text.FullEvalFormatter()
     eval_formatter_check(f)
     eval_formatter_slicing_check(f)
 
 def test_dollar_formatter():
     f = text.DollarFormatter()
     eval_formatter_check(f)
     eval_formatter_slicing_check(f)
     
     ns = dict(n=12, pi=math.pi, stuff='hello there', os=os)
     s = f.format("$n", **ns)
     nt.assert_equal(s, "12")
     s = f.format("$n.real", **ns)
     nt.assert_equal(s, "12")
     s = f.format("$n/{stuff[:5]}", **ns)
     nt.assert_equal(s, "12/hello")
     s = f.format("$n $$HOME", **ns)
     nt.assert_equal(s, "12 $HOME")
     s = f.format("${foo}", foo="HOME")
     nt.assert_equal(s, "$HOME")
 
 
 def test_long_substr():
     data = ['hi']
     nt.assert_equal(text.long_substr(data), 'hi')
 
 
 def test_long_substr2():
     data = ['abc', 'abd', 'abf', 'ab']
     nt.assert_equal(text.long_substr(data), 'ab')
 
+def test_long_substr_empty():
+    data = []
+    nt.assert_equal(text.long_substr(data), '')
 
 def test_strip_email():
     src = """\
         >> >>> def f(x):
         >> ...   return x+1
         >> ... 
         >> >>> zz = f(2.5)"""
     cln = """\
 >>> def f(x):
 ...   return x+1
 ... 
 >>> zz = f(2.5)"""
     nt.assert_equal(text.strip_email_quotes(src), cln)
 
 
 def test_strip_email2():
     src = '> > > list()'
     cln = 'list()'
     nt.assert_equal(text.strip_email_quotes(src), cln)

IPython/utils/text.py

@@ -1,845 +1,845 b''
 # encoding: utf-8
 """
 Utilities for working with strings and text.
 """
 
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2008-2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Imports
 #-----------------------------------------------------------------------------
 
 import __main__
 
 import os
 import re
 import shutil
 import sys
 import textwrap
 from string import Formatter
 
 from IPython.external.path import path
 from IPython.testing.skipdoctest import skip_doctest_py3, skip_doctest
 from IPython.utils import py3compat
 from IPython.utils.io import nlprint
 from IPython.utils.data import flatten
 
 #-----------------------------------------------------------------------------
 # Code
 #-----------------------------------------------------------------------------
 
 def unquote_ends(istr):
     """Remove a single pair of quotes from the endpoints of a string."""
 
     if not istr:
         return istr
     if (istr[0]=="'" and istr[-1]=="'") or \
        (istr[0]=='"' and istr[-1]=='"'):
         return istr[1:-1]
     else:
         return istr
 
 
 class LSString(str):
     """String derivative with a special access attributes.
 
     These are normal strings, but with the special attributes:
 
         .l (or .list) : value as list (split on newlines).
         .n (or .nlstr): original value (the string itself).
         .s (or .spstr): value as whitespace-separated string.
         .p (or .paths): list of path objects
 
     Any values which require transformations are computed only once and
     cached.
 
     Such strings are very useful to efficiently interact with the shell, which
     typically only understands whitespace-separated options for commands."""
 
     def get_list(self):
         try:
             return self.__list
         except AttributeError:
             self.__list = self.split('\n')
             return self.__list
 
     l = list = property(get_list)
 
     def get_spstr(self):
         try:
             return self.__spstr
         except AttributeError:
             self.__spstr = self.replace('\n',' ')
             return self.__spstr
 
     s = spstr = property(get_spstr)
 
     def get_nlstr(self):
         return self
 
     n = nlstr = property(get_nlstr)
 
     def get_paths(self):
         try:
             return self.__paths
         except AttributeError:
             self.__paths = [path(p) for p in self.split('\n') if os.path.exists(p)]
             return self.__paths
 
     p = paths = property(get_paths)
 
 # FIXME: We need to reimplement type specific displayhook and then add this
 # back as a custom printer. This should also be moved outside utils into the
 # core.
 
 # def print_lsstring(arg):
 #     """ Prettier (non-repr-like) and more informative printer for LSString """
 #     print "LSString (.p, .n, .l, .s available). Value:"
 #     print arg
 #
 #
 # print_lsstring = result_display.when_type(LSString)(print_lsstring)
 
 
 class SList(list):
     """List derivative with a special access attributes.
 
     These are normal lists, but with the special attributes:
 
         .l (or .list) : value as list (the list itself).
         .n (or .nlstr): value as a string, joined on newlines.
         .s (or .spstr): value as a string, joined on spaces.
         .p (or .paths): list of path objects
 
     Any values which require transformations are computed only once and
     cached."""
 
     def get_list(self):
         return self
 
     l = list = property(get_list)
 
     def get_spstr(self):
         try:
             return self.__spstr
         except AttributeError:
             self.__spstr = ' '.join(self)
             return self.__spstr
 
     s = spstr = property(get_spstr)
 
     def get_nlstr(self):
         try:
             return self.__nlstr
         except AttributeError:
             self.__nlstr = '\n'.join(self)
             return self.__nlstr
 
     n = nlstr = property(get_nlstr)
 
     def get_paths(self):
         try:
             return self.__paths
         except AttributeError:
             self.__paths = [path(p) for p in self if os.path.exists(p)]
             return self.__paths
 
     p = paths = property(get_paths)
 
     def grep(self, pattern, prune = False, field = None):
         """ Return all strings matching 'pattern' (a regex or callable)
 
         This is case-insensitive. If prune is true, return all items
         NOT matching the pattern.
 
         If field is specified, the match must occur in the specified
         whitespace-separated field.
 
         Examples::
 
             a.grep( lambda x: x.startswith('C') )
             a.grep('Cha.*log', prune=1)
             a.grep('chm', field=-1)
         """
 
         def match_target(s):
             if field is None:
                 return s
             parts = s.split()
             try:
                 tgt = parts[field]
                 return tgt
             except IndexError:
                 return ""
 
         if isinstance(pattern, basestring):
             pred = lambda x : re.search(pattern, x, re.IGNORECASE)
         else:
             pred = pattern
         if not prune:
             return SList([el for el in self if pred(match_target(el))])
         else:
             return SList([el for el in self if not pred(match_target(el))])
 
     def fields(self, *fields):
         """ Collect whitespace-separated fields from string list
 
         Allows quick awk-like usage of string lists.
 
         Example data (in var a, created by 'a = !ls -l')::
             -rwxrwxrwx  1 ville None      18 Dec 14  2006 ChangeLog
             drwxrwxrwx+ 6 ville None       0 Oct 24 18:05 IPython
 
         a.fields(0) is ['-rwxrwxrwx', 'drwxrwxrwx+']
         a.fields(1,0) is ['1 -rwxrwxrwx', '6 drwxrwxrwx+']
         (note the joining by space).
         a.fields(-1) is ['ChangeLog', 'IPython']
 
         IndexErrors are ignored.
 
         Without args, fields() just split()'s the strings.
         """
         if len(fields) == 0:
             return [el.split() for el in self]
 
         res = SList()
         for el in [f.split() for f in self]:
             lineparts = []
 
             for fd in fields:
                 try:
                     lineparts.append(el[fd])
                 except IndexError:
                     pass
             if lineparts:
                 res.append(" ".join(lineparts))
 
         return res
 
     def sort(self,field= None,  nums = False):
         """ sort by specified fields (see fields())
 
         Example::
             a.sort(1, nums = True)
 
         Sorts a by second field, in numerical order (so that 21 > 3)
 
         """
 
         #decorate, sort, undecorate
         if field is not None:
             dsu = [[SList([line]).fields(field),  line] for line in self]
         else:
             dsu = [[line,  line] for line in self]
         if nums:
             for i in range(len(dsu)):
                 numstr = "".join([ch for ch in dsu[i][0] if ch.isdigit()])
                 try:
                     n = int(numstr)
                 except ValueError:
                     n = 0;
                 dsu[i][0] = n
 
 
         dsu.sort()
         return SList([t[1] for t in dsu])
 
 
 # FIXME: We need to reimplement type specific displayhook and then add this
 # back as a custom printer. This should also be moved outside utils into the
 # core.
 
 # def print_slist(arg):
 #     """ Prettier (non-repr-like) and more informative printer for SList """
 #     print "SList (.p, .n, .l, .s, .grep(), .fields(), sort() available):"
 #     if hasattr(arg,  'hideonce') and arg.hideonce:
 #         arg.hideonce = False
 #         return
 #
 #     nlprint(arg)
 #
 # print_slist = result_display.when_type(SList)(print_slist)
 
 
 def esc_quotes(strng):
     """Return the input string with single and double quotes escaped out"""
 
     return strng.replace('"','\\"').replace("'","\\'")
 
 
 def qw(words,flat=0,sep=None,maxsplit=-1):
     """Similar to Perl's qw() operator, but with some more options.
 
     qw(words,flat=0,sep=' ',maxsplit=-1) -> words.split(sep,maxsplit)
 
     words can also be a list itself, and with flat=1, the output will be
     recursively flattened.
 
     Examples:
 
     >>> qw('1 2')
     ['1', '2']
 
     >>> qw(['a b','1 2',['m n','p q']])
     [['a', 'b'], ['1', '2'], [['m', 'n'], ['p', 'q']]]
 
     >>> qw(['a b','1 2',['m n','p q']],flat=1)
     ['a', 'b', '1', '2', 'm', 'n', 'p', 'q']
     """
 
     if isinstance(words, basestring):
         return [word.strip() for word in words.split(sep,maxsplit)
                 if word and not word.isspace() ]
     if flat:
         return flatten(map(qw,words,[1]*len(words)))
     return map(qw,words)
 
 
 def qwflat(words,sep=None,maxsplit=-1):
     """Calls qw(words) in flat mode. It's just a convenient shorthand."""
     return qw(words,1,sep,maxsplit)
 
 
 def qw_lol(indata):
     """qw_lol('a b') -> [['a','b']],
     otherwise it's just a call to qw().
 
     We need this to make sure the modules_some keys *always* end up as a
     list of lists."""
 
     if isinstance(indata, basestring):
         return [qw(indata)]
     else:
         return qw(indata)
 
 
 def grep(pat,list,case=1):
     """Simple minded grep-like function.
     grep(pat,list) returns occurrences of pat in list, None on failure.
 
     It only does simple string matching, with no support for regexps. Use the
     option case=0 for case-insensitive matching."""
 
     # This is pretty crude. At least it should implement copying only references
     # to the original data in case it's big. Now it copies the data for output.
     out=[]
     if case:
         for term in list:
             if term.find(pat)>-1: out.append(term)
     else:
         lpat=pat.lower()
         for term in list:
             if term.lower().find(lpat)>-1: out.append(term)
 
     if len(out): return out
     else: return None
 
 
 def dgrep(pat,*opts):
     """Return grep() on dir()+dir(__builtins__).
 
     A very common use of grep() when working interactively."""
 
     return grep(pat,dir(__main__)+dir(__main__.__builtins__),*opts)
 
 
 def idgrep(pat):
     """Case-insensitive dgrep()"""
 
     return dgrep(pat,0)
 
 
 def igrep(pat,list):
     """Synonym for case-insensitive grep."""
 
     return grep(pat,list,case=0)
 
 
 def indent(instr,nspaces=4, ntabs=0, flatten=False):
     """Indent a string a given number of spaces or tabstops.
 
     indent(str,nspaces=4,ntabs=0) -> indent str by ntabs+nspaces.
 
     Parameters
     ----------
 
     instr : basestring
         The string to be indented.
     nspaces : int (default: 4)
         The number of spaces to be indented.
     ntabs : int (default: 0)
         The number of tabs to be indented.
     flatten : bool (default: False)
         Whether to scrub existing indentation.  If True, all lines will be
         aligned to the same indentation.  If False, existing indentation will
         be strictly increased.
 
     Returns
     -------
 
     str|unicode : string indented by ntabs and nspaces.
 
     """
     if instr is None:
         return
     ind = '\t'*ntabs+' '*nspaces
     if flatten:
         pat = re.compile(r'^\s*', re.MULTILINE)
     else:
         pat = re.compile(r'^', re.MULTILINE)
     outstr = re.sub(pat, ind, instr)
     if outstr.endswith(os.linesep+ind):
         return outstr[:-len(ind)]
     else:
         return outstr
 
 def native_line_ends(filename,backup=1):
     """Convert (in-place) a file to line-ends native to the current OS.
 
     If the optional backup argument is given as false, no backup of the
     original file is left.  """
 
     backup_suffixes = {'posix':'~','dos':'.bak','nt':'.bak','mac':'.bak'}
 
     bak_filename = filename + backup_suffixes[os.name]
 
     original = open(filename).read()
     shutil.copy2(filename,bak_filename)
     try:
         new = open(filename,'wb')
         new.write(os.linesep.join(original.splitlines()))
         new.write(os.linesep) # ALWAYS put an eol at the end of the file
         new.close()
     except:
         os.rename(bak_filename,filename)
     if not backup:
         try:
             os.remove(bak_filename)
         except:
             pass
 
 
 def list_strings(arg):
     """Always return a list of strings, given a string or list of strings
     as input.
 
     :Examples:
 
         In [7]: list_strings('A single string')
         Out[7]: ['A single string']
 
         In [8]: list_strings(['A single string in a list'])
         Out[8]: ['A single string in a list']
 
         In [9]: list_strings(['A','list','of','strings'])
         Out[9]: ['A', 'list', 'of', 'strings']
     """
 
     if isinstance(arg,basestring): return [arg]
     else: return arg
 
 
 def marquee(txt='',width=78,mark='*'):
     """Return the input string centered in a 'marquee'.
 
     :Examples:
 
         In [16]: marquee('A test',40)
         Out[16]: '**************** A test ****************'
 
         In [17]: marquee('A test',40,'-')
         Out[17]: '---------------- A test ----------------'
 
         In [18]: marquee('A test',40,' ')
         Out[18]: '                 A test                 '
 
     """
     if not txt:
         return (mark*width)[:width]
     nmark = (width-len(txt)-2)//len(mark)//2
     if nmark < 0: nmark =0
     marks = mark*nmark
     return '%s %s %s' % (marks,txt,marks)
 
 
 ini_spaces_re = re.compile(r'^(\s+)')
 
 def num_ini_spaces(strng):
     """Return the number of initial spaces in a string"""
 
     ini_spaces = ini_spaces_re.match(strng)
     if ini_spaces:
         return ini_spaces.end()
     else:
         return 0
 
 
 def format_screen(strng):
     """Format a string for screen printing.
 
     This removes some latex-type format codes."""
     # Paragraph continue
     par_re = re.compile(r'\\$',re.MULTILINE)
     strng = par_re.sub('',strng)
     return strng
 
 
 def dedent(text):
     """Equivalent of textwrap.dedent that ignores unindented first line.
 
     This means it will still dedent strings like:
     '''foo
     is a bar
     '''
 
     For use in wrap_paragraphs.
     """
 
     if text.startswith('\n'):
         # text starts with blank line, don't ignore the first line
         return textwrap.dedent(text)
 
     # split first line
     splits = text.split('\n',1)
     if len(splits) == 1:
         # only one line
         return textwrap.dedent(text)
 
     first, rest = splits
     # dedent everything but the first line
     rest = textwrap.dedent(rest)
     return '\n'.join([first, rest])
 
 
 def wrap_paragraphs(text, ncols=80):
     """Wrap multiple paragraphs to fit a specified width.
 
     This is equivalent to textwrap.wrap, but with support for multiple
     paragraphs, as separated by empty lines.
 
     Returns
     -------
 
     list of complete paragraphs, wrapped to fill `ncols` columns.
     """
     paragraph_re = re.compile(r'\n(\s*\n)+', re.MULTILINE)
     text = dedent(text).strip()
     paragraphs = paragraph_re.split(text)[::2] # every other entry is space
     out_ps = []
     indent_re = re.compile(r'\n\s+', re.MULTILINE)
     for p in paragraphs:
         # presume indentation that survives dedent is meaningful formatting,
         # so don't fill unless text is flush.
         if indent_re.search(p) is None:
             # wrap paragraph
             p = textwrap.fill(p, ncols)
         out_ps.append(p)
     return out_ps
 
 
 def long_substr(data):
     """Return the longest common substring in a list of strings.
     
     Credit: http://stackoverflow.com/questions/2892931/longest-common-substring-from-more-than-two-strings-python
     """
     substr = ''
     if len(data) > 1 and len(data[0]) > 0:
         for i in range(len(data[0])):
             for j in range(len(data[0])-i+1):
                 if j > len(substr) and all(data[0][i:i+j] in x for x in data):
                     substr = data[0][i:i+j]
-    else:
+    elif len(data) == 1:
         substr = data[0]
     return substr
 
 
 def strip_email_quotes(text):
     """Strip leading email quotation characters ('>').
 
     Removes any combination of leading '>' interspersed with whitespace that
     appears *identically* in all lines of the input text.
 
     Parameters
     ----------
     text : str
 
     Examples
     --------
 
     Simple uses::
 
         In [2]: strip_email_quotes('> > text')
         Out[2]: 'text'
 
         In [3]: strip_email_quotes('> > text\\n> > more')
         Out[3]: 'text\\nmore'
 
     Note how only the common prefix that appears in all lines is stripped::
 
         In [4]: strip_email_quotes('> > text\\n> > more\\n> more...')
         Out[4]: '> text\\n> more\\nmore...'
 
     So if any line has no quote marks ('>') , then none are stripped from any
     of them ::
     
         In [5]: strip_email_quotes('> > text\\n> > more\\nlast different')
         Out[5]: '> > text\\n> > more\\nlast different'
     """
     lines = text.splitlines()
     matches = set()
     for line in lines:
         prefix = re.match(r'^(\s*>[ >]*)', line)
         if prefix:
             matches.add(prefix.group(1))
         else:
             break
     else:
         prefix = long_substr(list(matches))
         if prefix:
             strip = len(prefix)
             text = '\n'.join([ ln[strip:] for ln in lines])
     return text
 
 
 class EvalFormatter(Formatter):
     """A String Formatter that allows evaluation of simple expressions.
     
     Note that this version interprets a : as specifying a format string (as per
     standard string formatting), so if slicing is required, you must explicitly
     create a slice.
     
     This is to be used in templating cases, such as the parallel batch
     script templates, where simple arithmetic on arguments is useful.
 
     Examples
     --------
     
     In  [1]: f = EvalFormatter()
     In  [2]: f.format('{n//4}', n=8)
     Out [2]: '2'
     
     In  [3]: f.format("{greeting[slice(2,4)]}", greeting="Hello")
     Out [3]: 'll'
     """
     def get_field(self, name, args, kwargs):
         v = eval(name, kwargs)
         return v, name
 
 
 @skip_doctest_py3
 class FullEvalFormatter(Formatter):
     """A String Formatter that allows evaluation of simple expressions.
     
     Any time a format key is not found in the kwargs,
     it will be tried as an expression in the kwargs namespace.
     
     Note that this version allows slicing using [1:2], so you cannot specify
     a format string. Use :class:`EvalFormatter` to permit format strings.
     
     Examples
     --------
     
     In [1]: f = FullEvalFormatter()
     In [2]: f.format('{n//4}', n=8)
     Out[2]: u'2'
     
     In [3]: f.format('{list(range(5))[2:4]}')
     Out[3]: u'[2, 3]'
 
     In [4]: f.format('{3*2}')
     Out[4]: u'6'
     """
     # copied from Formatter._vformat with minor changes to allow eval
     # and replace the format_spec code with slicing
     def _vformat(self, format_string, args, kwargs, used_args, recursion_depth):
         if recursion_depth < 0:
             raise ValueError('Max string recursion exceeded')
         result = []
         for literal_text, field_name, format_spec, conversion in \
                 self.parse(format_string):
 
             # output the literal text
             if literal_text:
                 result.append(literal_text)
 
             # if there's a field, output it
             if field_name is not None:
                 # this is some markup, find the object and do
                 # the formatting
 
                 if format_spec:
                     # override format spec, to allow slicing:
                     field_name = ':'.join([field_name, format_spec])
 
                 # eval the contents of the field for the object
                 # to be formatted
                 obj = eval(field_name, kwargs)
 
                 # do any conversion on the resulting object
                 obj = self.convert_field(obj, conversion)
 
                 # format the object and append to the result
                 result.append(self.format_field(obj, ''))
 
         return u''.join(py3compat.cast_unicode(s) for s in result)
 
 
 @skip_doctest_py3
 class DollarFormatter(FullEvalFormatter):
     """Formatter allowing Itpl style $foo replacement, for names and attribute
     access only. Standard {foo} replacement also works, and allows full
     evaluation of its arguments.
 
     Examples
     --------
     In [1]: f = DollarFormatter()
     In [2]: f.format('{n//4}', n=8)
     Out[2]: u'2'
     
     In [3]: f.format('23 * 76 is $result', result=23*76)
     Out[3]: u'23 * 76 is 1748'
     
     In [4]: f.format('$a or {b}', a=1, b=2)
     Out[4]: u'1 or 2'
     """
     _dollar_pattern = re.compile("(.*?)\$(\$?[\w\.]+)")
     def parse(self, fmt_string):
         for literal_txt, field_name, format_spec, conversion \
                     in Formatter.parse(self, fmt_string):
             
             # Find $foo patterns in the literal text.
             continue_from = 0
             txt = ""
             for m in self._dollar_pattern.finditer(literal_txt):
                 new_txt, new_field = m.group(1,2)
                 # $$foo --> $foo
                 if new_field.startswith("$"):
                     txt += new_txt + new_field
                 else:
                     yield (txt + new_txt, new_field, "", None)
                     txt = ""
                 continue_from = m.end()
             
             # Re-yield the {foo} style pattern
             yield (txt + literal_txt[continue_from:], field_name, format_spec, conversion)
 
 #-----------------------------------------------------------------------------
 # Utils to columnize a list of string
 #-----------------------------------------------------------------------------
 
 def _chunks(l, n):
     """Yield successive n-sized chunks from l."""
     for i in xrange(0, len(l), n):
         yield l[i:i+n]
 
 
 def _find_optimal(rlist , separator_size=2 , displaywidth=80):
     """Calculate optimal info to columnize a list of string"""
     for nrow in range(1, len(rlist)+1) :
         chk = map(max,_chunks(rlist, nrow))
         sumlength = sum(chk)
         ncols = len(chk)
         if sumlength+separator_size*(ncols-1) <= displaywidth :
             break;
     return {'columns_numbers' : ncols,
             'optimal_separator_width':(displaywidth - sumlength)/(ncols-1) if (ncols -1) else 0,
             'rows_numbers' : nrow,
             'columns_width' : chk
            }
 
 
 def _get_or_default(mylist, i, default=None):
     """return list item number, or default if don't exist"""
     if i >= len(mylist):
         return default
     else :
         return mylist[i]
 
 
 @skip_doctest
 def compute_item_matrix(items, empty=None, *args, **kwargs) :
     """Returns a nested list, and info to columnize items
 
     Parameters :
     ------------
 
     items :
         list of strings to columize
     empty : (default None)
         default value to fill list if needed
     separator_size : int (default=2)
         How much caracters will be used as a separation between each columns.
     displaywidth : int (default=80)
         The width of the area onto wich the columns should enter
 
     Returns :
     ---------
 
     Returns a tuple of (strings_matrix, dict_info)
 
     strings_matrix :
 
         nested list of string, the outer most list contains as many list as
         rows, the innermost lists have each as many element as colums. If the
         total number of elements in `items` does not equal the product of
         rows*columns, the last element of some lists are filled with `None`.
 
     dict_info :
         some info to make columnize easier:
 
         columns_numbers : number of columns
         rows_numbers    : number of rows
         columns_width   : list of with of each columns
         optimal_separator_width : best separator width between columns
 
     Exemple :
     ---------
 
     In [1]: l = ['aaa','b','cc','d','eeeee','f','g','h','i','j','k','l']
        ...: compute_item_matrix(l,displaywidth=12)
     Out[1]:
         ([['aaa', 'f', 'k'],
         ['b', 'g', 'l'],
         ['cc', 'h', None],
         ['d', 'i', None],
         ['eeeee', 'j', None]],
         {'columns_numbers': 3,
         'columns_width': [5, 1, 1],
         'optimal_separator_width': 2,
         'rows_numbers': 5})
 
     """
     info = _find_optimal(map(len, items), *args, **kwargs)
     nrow, ncol = info['rows_numbers'], info['columns_numbers']
     return ([[ _get_or_default(items, c*nrow+i, default=empty) for c in range(ncol) ] for i in range(nrow) ], info)
 
 
 def columnize(items, separator='  ', displaywidth=80):
     """ Transform a list of strings into a single string with columns.
 
     Parameters
     ----------
     items : sequence of strings
         The strings to process.
 
     separator : str, optional [default is two spaces]
         The string that separates columns.
 
     displaywidth : int, optional [default is 80]
         Width of the display in number of characters.
 
     Returns
     -------
     The formatted string.
     """
     if not items :
         return '\n'
     matrix, info = compute_item_matrix(items, separator_size=len(separator), displaywidth=displaywidth)
     fmatrix = [filter(None, x) for x in matrix]
     sjoin = lambda x : separator.join([ y.ljust(w, ' ') for y, w in zip(x, info['columns_width'])])
     return '\n'.join(map(sjoin, fmatrix))+'\n'