bpo-25324: copy tok_name before changing it (#1608)

* add test to check if were modifying token

* copy list so import tokenize doesnt have side effects on token

* shorten line

* add tokenize tokens to token.h to get them to show up in token

* move ERRORTOKEN back to its previous location, and fix nitpick

* copy comments from token.h automatically

* fix whitespace and make more pythonic

* change to fix comments from @Haypo

* update token.rst and Misc/NEWS

* change wording

* some more wording changes

Author: albertjan

Doc/library/token.rst

@@ -101,10 +101,18 @@ The token constants are:
           AWAIT
           ASYNC
           ERRORTOKEN
+          COMMENT
+          NL
+          ENCODING
           N_TOKENS
           NT_OFFSET
 
    .. versionchanged:: 3.5
       Added :data:`AWAIT` and :data:`ASYNC` tokens. Starting with
       Python 3.7, "async" and "await" will be tokenized as :data:`NAME`
       tokens, and :data:`AWAIT` and :data:`ASYNC` will be removed.
+
+   .. versionchanged:: 3.7
+      Added :data:`COMMENT`, :data:`NL` and :data:`ENCODING` to bring
+      the tokens in the C code in line with the tokens needed in
+      :mod:`tokenize` module. These tokens aren't used by the C tokenizer.

Include/token.h

@@ -67,7 +67,11 @@ extern "C" {
 #define AWAIT		54
 #define ASYNC		55
 #define ERRORTOKEN	56
-#define N_TOKENS	57
+/* These aren't used by the C tokenizer but are needed for tokenize.py */
+#define COMMENT		57
+#define NL		58
+#define ENCODING		59
+#define N_TOKENS	60
 
 /* Special definitions for cooperation with parser */
 

Lib/test/test_tokenize.py

@@ -1343,13 +1343,13 @@ def assertExactTypeEqual(self, opstr, *optypes):
         tokens = list(tokenize(BytesIO(opstr.encode('utf-8')).readline))
         num_optypes = len(optypes)
         self.assertEqual(len(tokens), 2 + num_optypes)
-        self.assertEqual(token.tok_name[tokens[0].exact_type],
-                         token.tok_name[ENCODING])
+        self.assertEqual(tok_name[tokens[0].exact_type],
+                         tok_name[ENCODING])
         for i in range(num_optypes):
-            self.assertEqual(token.tok_name[tokens[i + 1].exact_type],
-                             token.tok_name[optypes[i]])
-        self.assertEqual(token.tok_name[tokens[1 + num_optypes].exact_type],
-                         token.tok_name[token.ENDMARKER])
+            self.assertEqual(tok_name[tokens[i + 1].exact_type],
+                             tok_name[optypes[i]])
+        self.assertEqual(tok_name[tokens[1 + num_optypes].exact_type],
+                         tok_name[token.ENDMARKER])
 
     def test_exact_type(self):
         self.assertExactTypeEqual('()', token.LPAR, token.RPAR)

Lib/token.py

@@ -63,11 +63,17 @@
 ATEQUAL = 50
 RARROW = 51
 ELLIPSIS = 52
+# Don't forget to update the table _PyParser_TokenNames in tokenizer.c!
 OP = 53
 AWAIT = 54
 ASYNC = 55
 ERRORTOKEN = 56
-N_TOKENS = 57
+# These aren't used by the C tokenizer but are needed for tokenize.py
+COMMENT = 57
+NL = 58
+ENCODING = 59
+N_TOKENS = 60
+# Special definitions for cooperation with parser
 NT_OFFSET = 256
 #--end constants--
 
@@ -102,15 +108,26 @@ def _main():
     with fp:
         lines = fp.read().split("\n")
     prog = re.compile(
-        "#define[ \t][ \t]*([A-Z0-9][A-Z0-9_]*)[ \t][ \t]*([0-9][0-9]*)",
+        r"#define[ \t][ \t]*([A-Z0-9][A-Z0-9_]*)[ \t][ \t]*([0-9][0-9]*)",
         re.IGNORECASE)
+    comment_regex = re.compile(
+        r"^\s*/\*\s*(.+?)\s*\*/\s*$",
+        re.IGNORECASE)
+
     tokens = {}
+    prev_val = None
     for line in lines:
         match = prog.match(line)
         if match:
             name, val = match.group(1, 2)
             val = int(val)
-            tokens[val] = name          # reverse so we can sort them...
+            tokens[val] = {'token': name}          # reverse so we can sort them...
+            prev_val = val
+        else:
+            comment_match = comment_regex.match(line)
+            if comment_match and prev_val is not None:
+                comment = comment_match.group(1)
+                tokens[prev_val]['comment'] = comment
     keys = sorted(tokens.keys())
     # load the output skeleton from the target:
     try:
@@ -127,8 +144,10 @@ def _main():
         sys.stderr.write("target does not contain format markers")
         sys.exit(3)
     lines = []
-    for val in keys:
-        lines.append("%s = %d" % (tokens[val], val))
+    for key in keys:
+        lines.append("%s = %d" % (tokens[key]["token"], key))
+        if "comment" in tokens[key]:
+            lines.append("# %s" % tokens[key]["comment"])
     format[start:end] = lines
     try:
         fp = open(outFileName, 'w')

Lib/tokenize.py

@@ -38,17 +38,10 @@
 blank_re = re.compile(br'^[ \t\f]*(?:[#\r\n]|$)', re.ASCII)
 
 import token
-__all__ = token.__all__ + ["COMMENT", "tokenize", "detect_encoding",
-                           "NL", "untokenize", "ENCODING", "TokenInfo"]
+__all__ = token.__all__ + ["tokenize", "detect_encoding",
+                           "untokenize", "TokenInfo"]
 del token
 
-COMMENT = N_TOKENS
-tok_name[COMMENT] = 'COMMENT'
-NL = N_TOKENS + 1
-tok_name[NL] = 'NL'
-ENCODING = N_TOKENS + 2
-tok_name[ENCODING] = 'ENCODING'
-N_TOKENS += 3
 EXACT_TOKEN_TYPES = {
     '(':   LPAR,
     ')':   RPAR,

Misc/NEWS

@@ -10,6 +10,10 @@ What's New in Python 3.7.0 alpha 1?
 Core and Builtins
 -----------------
 
+- bpo-25324: Tokens needed for parsing in Python moved to C. ``COMMENT``,
+  ``NL`` and ``ENCODING``. This way the tokens and tok_names in the token
+  module don't get changed when you import the tokenize module.
+
 - bpo-29104: Fixed parsing backslashes in f-strings.
 
 - bpo-27945: Fixed various segfaults with dict when input collections are

Parser/tokenizer.c

@@ -106,6 +106,9 @@ const char *_PyParser_TokenNames[] = {
     "AWAIT",
     "ASYNC",
     "<ERRORTOKEN>",
+    "COMMENT",
+    "NL",
+    "ENCODING"
     "<N_TOKENS>"
 };