Add fix_oldstr_wrap and refactor fix_division_safe

Author: edschofield

libfuturize/fixer_util.py

@@ -13,6 +13,7 @@
 from lib2to3.pytree import Leaf, Node
 from lib2to3.pygram import python_symbols as syms, python_grammar
 from lib2to3.pygram import token
+from lib2to3.fixer_util import (Node, Call, Name, syms, Comma, Number)
 import re
 
 
@@ -429,3 +430,23 @@ def is_shebang_comment(node):
     return bool(re.match(SHEBANG_REGEX, node.prefix))
 
 
+def wrap_in_fn_call(fn_name, args, prefix=None):
+    """
+    Example:
+    >>> wrap_in_fn_call("oldstr", (arg,))
+    oldstr(arg)
+
+    >>> wrap_in_fn_call("olddiv", (arg1, arg2))
+    olddiv(arg1, arg2)
+    """
+    assert len(args) > 0
+    if len(args) == 1:
+        newargs = args
+    elif len(args) == 2:
+        expr1, expr2 = args
+        newargs = [expr1, Comma(), expr2]
+    else:
+        assert NotImplementedError('write me')
+    return Call(Name(fn_name), newargs, prefix=prefix)
+
+

libfuturize/fixes/fix_division_safe.py

@@ -11,10 +11,10 @@
 """
 
 from lib2to3 import fixer_base
-from lib2to3.fixer_util import (Node, Call, Name, parenthesize,
-                                syms, LParen, RParen, Comma, Number)
+from lib2to3.fixer_util import syms
 from libfuturize.fixer_util import token, future_import, touch_import_top
 
+
 def match_division(node):
     u"""
     __future__.division redefines the meaning of a single slash for division,
@@ -25,10 +25,6 @@ def match_division(node):
                                   not node.prev_sibling.type == slash
 
 
-def parenthesize_two(expr1, expr2):
-    return Node(syms.atom, [LParen(), expr1, Comma(), expr2, RParen()])
-
-
 class FixDivisionSafe(fixer_base.BaseFix):
     # BM_compatible = True
     run_order = 4    # this seems to be ignored?
@@ -48,13 +44,15 @@ def match(self, node):
                     len(node.children) == 3 and
                     match_division(node.children[1])):
             expr1, expr2 = node.children[0], node.children[2]
-        return expr1, expr2
+            return expr1, expr2
+        else:
+            return False
 
     def transform(self, node, results):
         future_import(u"division", node)
         touch_import_top(u'past.utils', u'old_div', node)
         expr1, expr2 = results[0].clone(), results[1].clone()
         # Strip any leading space for the first number:
         expr1.prefix = u''
-        return Call(Name("old_div"), [expr1, Comma(), expr2], prefix=node.prefix)
+        return wrap_in_fn_call("old_div", (expr1, expr2), prefix=node.prefix)
 

libfuturize/fixes/fix_oldstr_wrap.py

@@ -0,0 +1,80 @@
+"""
+For the ``future`` package.
+
+Adds this import line:
+
+    from past.builtins import str as oldstr
+
+at the top and wraps any unadorned string literals 'abc' or explicit byte-string
+literals b'abc' in oldstr() calls so the code has the same behaviour on Py3 as
+on Py2.6/2.7.
+"""
+
+from __future__ import unicode_literals
+import re
+from lib2to3 import fixer_base
+from lib2to3.pgen2 import token
+from lib2to3.fixer_util import syms
+from libfuturize.fixer_util import (future_import, touch_import_top,
+                                    wrap_in_fn_call)
+
+
+_literal_re = re.compile(r"[^uUrR]?[\'\"]")
+
+
+class FixOldstrWrap(fixer_base.BaseFix):
+    BM_compatible = True
+    PATTERN = "STRING"
+
+    def transform(self, node, results):
+        import pdb
+        pdb.set_trace()
+        if node.type == token.STRING:
+            touch_import_top(u'past.types', u'oldstr', node)
+            if _literal_re.match(node.value):
+                new = node.clone()
+                # Strip any leading space or comments:
+                # TODO: check: do we really want to do this?
+                new.prefix = u''
+                new.value = u'b' + new.value
+                wrapped = wrap_in_fn_call("oldstr", [new], prefix=node.prefix)
+                return wrapped
+
+    def transform(self, node, results):
+        expr1, expr2 = results[0].clone(), results[1].clone()
+        # Strip any leading space for the first number:
+        expr1.prefix = u''
+        return wrap_in_fn_call("old_div", expr1, expr2, prefix=node.prefix)
+
+
+class FixDivisionSafe(fixer_base.BaseFix):
+    # BM_compatible = True
+    run_order = 4    # this seems to be ignored?
+
+    _accept_type = token.SLASH
+
+    PATTERN = """
+    term<(not('/') any)+ '/' ((not('/') any))>
+    """
+
+    def match(self, node):
+        u"""
+        Since the tree needs to be fixed once and only once if and only if it
+        matches, then we can start discarding matches after we make the first.
+        """
+        if (node.type == self.syms.term and 
+                    len(node.children) == 3 and
+                    match_division(node.children[1])):
+            expr1, expr2 = node.children[0], node.children[2]
+            return expr1, expr2
+        else:
+            return False
+
+    def transform(self, node, results):
+        future_import(u"division", node)
+        touch_import_top(u'past.utils', u'old_div', node)
+        expr1, expr2 = results[0].clone(), results[1].clone()
+        # Strip any leading space for the first number:
+        expr1.prefix = u''
+        return wrap_in_fn_call("old_div", expr1, expr2, prefix=node.prefix)
+