For Python 2.7 and 3.4 pytest 5 doesn't work anymore

Line continuation characters are valid inside of strings, but weren't
handled correctly in certain cases with f-strings, due to some small
tokenizer bugs.

This pull request to address those issues, and adds tests to validate
the new logic.

CHANGELOG.rst

@@ -3,6 +3,12 @@
 Changelog
 ---------
 
+0.5.1 (2019-07-13)
+++++++++++++++++++
+
+- Fix: Some unicode identifiers were not correctly tokenized
+- Fix: Line continuations in f-strings are now working
+
 0.5.0 (2019-06-20)
 ++++++++++++++++++
 
@@ -17,19 +23,19 @@ Changelog
 - Python 3.8 support
 - FileIO support, it's now possible to use abstract file IO, support is alpha
 
-0.3.4 (2018-02-13)
+0.3.4 (2019-02-13)
 +++++++++++++++++++
 
 - Fix an f-string tokenizer error
 
-0.3.3 (2018-02-06)
+0.3.3 (2019-02-06)
 +++++++++++++++++++
 
 - Fix async errors in the diff parser
 - A fix in iter_errors
 - This is a very small bugfix release
 
-0.3.2 (2018-01-24)
+0.3.2 (2019-01-24)
 +++++++++++++++++++
 
 - 20+ bugfixes in the diff parser and 3 in the tokenizer

parso/__init__.py

@@ -43,7 +43,7 @@ from parso.grammar import Grammar, load_grammar
 from parso.utils import split_lines, python_bytes_to_unicode
 
 
-__version__ = '0.5.0'
+__version__ = '0.5.1'
 
 
 def parse(code=None, **kwargs):

parso/grammar.py

@@ -57,7 +57,8 @@ class Grammar(object):
         :param str path: The path to the file you want to open. Only needed for caching.
         :param bool cache: Keeps a copy of the parser tree in RAM and on disk
             if a path is given. Returns the cached trees if the corresponding
-            files on disk have not changed.
+            files on disk have not changed. Note that this stores pickle files
+            on your file system (e.g. for Linux in ``~/.cache/parso/``).
         :param bool diff_cache: Diffs the cached python module against the new
             code and tries to parse only the parts that have changed. Returns
             the same (changed) module that is found in cache. Using this option

parso/python/tokenize.py

@@ -23,6 +23,9 @@ from parso._compatibility import py_version
 from parso.utils import split_lines
 
 
+# Maximum code point of Unicode 6.0: 0x10ffff (1,114,111)
+MAX_UNICODE = '\U0010ffff'
+
 STRING = PythonTokenTypes.STRING
 NAME = PythonTokenTypes.NAME
 NUMBER = PythonTokenTypes.NUMBER
@@ -51,8 +54,13 @@ if py_version >= 30:
     # Python 3 has str.isidentifier() to check if a char is a valid identifier
     is_identifier = str.isidentifier
 else:
-    namechars = string.ascii_letters + '_'
-    is_identifier = lambda s: s in namechars
+    # Python 2 doesn't, but it's not that important anymore and if you tokenize
+    # Python 2 code with this, it's still ok. It's just that parsing Python 3
+    # code with this function is not 100% correct.
+    # This just means that Python 2 code matches a few identifiers too much,
+    # but that doesn't really matter.
+    def is_identifier(s):
+        return True
 
 
 def group(*choices, **kwargs):
@@ -118,9 +126,9 @@ def _get_token_collection(version_info):
         return result
 
 
-fstring_string_single_line = _compile(r'(?:[^{}\r\n]+|\{\{|\}\})+')
+fstring_string_single_line = _compile(r'(?:\{\{|\}\}|\\(?:\r\n?|\n)|[^{}\r\n])+')
 fstring_string_multi_line = _compile(r'(?:[^{}]+|\{\{|\}\})+')
-fstring_format_spec_single_line = _compile(r'[^{}\r\n]+')
+fstring_format_spec_single_line = _compile(r'(?:\\(?:\r\n?|\n)|[^{}\r\n])+')
 fstring_format_spec_multi_line = _compile(r'[^{}]+')
 
 
@@ -130,7 +138,16 @@ def _create_token_collection(version_info):
     Whitespace = r'[ \f\t]*'
     whitespace = _compile(Whitespace)
     Comment = r'#[^\r\n]*'
-    Name = r'\w+'
+    # Python 2 is pretty much not working properly anymore, we just ignore
+    # parsing unicode properly, which is fine, I guess.
+    if version_info[0] == 2:
+        Name = r'([A-Za-z_0-9]+)'
+    elif sys.version_info[0] == 2:
+        # Unfortunately the regex engine cannot deal with the regex below, so
+        # just use this one.
+        Name = r'(\w+)'
+    else:
+        Name = u'([A-Za-z_0-9\u0080-' + MAX_UNICODE + ']+)'
 
     if version_info >= (3, 6):
         Hexnumber = r'0[xX](?:_?[0-9a-fA-F])+'
@@ -340,7 +357,9 @@ def _find_fstring_string(endpats, fstring_stack, line, lnum, pos):
 
     new_pos = pos
     new_pos += len(string)
-    if allow_multiline and (string.endswith('\n') or string.endswith('\r')):
+    # even if allow_multiline is False, we still need to check for trailing
+    # newlines, because a single-line f-string can contain line continuations
+    if string.endswith('\n') or string.endswith('\r'):
         tos.previous_lines += string
         string = ''
     else:
@@ -510,6 +529,24 @@ def tokenize_lines(lines, version_info, start_pos=(1, 0)):
             if (initial in numchars or                      # ordinary number
                     (initial == '.' and token != '.' and token != '...')):
                 yield PythonToken(NUMBER, token, spos, prefix)
+            elif pseudomatch.group(3) is not None:            # ordinary name
+                if token in always_break_tokens:
+                    fstring_stack[:] = []
+                    paren_level = 0
+                    # We only want to dedent if the token is on a new line.
+                    if re.match(r'[ \f\t]*$', line[:start]):
+                        while True:
+                            indent = indents.pop()
+                            if indent > start:
+                                yield PythonToken(DEDENT, '', spos, '')
+                            else:
+                                indents.append(indent)
+                                break
+                if is_identifier(token):
+                    yield PythonToken(NAME, token, spos, prefix)
+                else:
+                    for t in _split_illegal_unicode_name(token, spos, prefix):
+                        yield t  # yield from Python 2
             elif initial in '\r\n':
                 if any(not f.allow_multiline() for f in fstring_stack):
                     # Would use fstring_stack.clear, but that's not available
@@ -564,20 +601,6 @@ def tokenize_lines(lines, version_info, start_pos=(1, 0)):
             elif token in fstring_pattern_map:  # The start of an fstring.
                 fstring_stack.append(FStringNode(fstring_pattern_map[token]))
                 yield PythonToken(FSTRING_START, token, spos, prefix)
-            elif is_identifier(initial):                      # ordinary name
-                if token in always_break_tokens:
-                    fstring_stack[:] = []
-                    paren_level = 0
-                    # We only want to dedent if the token is on a new line.
-                    if re.match(r'[ \f\t]*$', line[:start]):
-                        while True:
-                            indent = indents.pop()
-                            if indent > start:
-                                yield PythonToken(DEDENT, '', spos, '')
-                            else:
-                                indents.append(indent)
-                                break
-                yield PythonToken(NAME, token, spos, prefix)
             elif initial == '\\' and line[start:] in ('\\\n', '\\\r\n', '\\\r'):  # continued stmt
                 additional_prefix += prefix + line[start:]
                 break
@@ -613,6 +636,39 @@ def tokenize_lines(lines, version_info, start_pos=(1, 0)):
     yield PythonToken(ENDMARKER, '', end_pos, additional_prefix)
 
 
+def _split_illegal_unicode_name(token, start_pos, prefix):
+    def create_token():
+        return PythonToken(ERRORTOKEN if is_illegal else NAME, found, pos, prefix)
+
+    found = ''
+    is_illegal = False
+    pos = start_pos
+    for i, char in enumerate(token):
+        if is_illegal:
+            if is_identifier(char):
+                yield create_token()
+                found = char
+                is_illegal = False
+                prefix = ''
+                pos = start_pos[0], start_pos[1] + i
+            else:
+                found += char
+        else:
+            new_found = found + char
+            if is_identifier(new_found):
+                found = new_found
+            else:
+                if found:
+                    yield create_token()
+                    prefix = ''
+                    pos = start_pos[0], start_pos[1] + i
+                found = char
+                is_illegal = True
+
+    if found:
+        yield create_token()
+
+
 if __name__ == "__main__":
     if len(sys.argv) >= 2:
         path = sys.argv[1]

parso/python/tree.py

@@ -43,7 +43,10 @@ Parser Tree Classes
 """
 
 import re
-from collections import Mapping
+try:
+    from collections.abc import Mapping
+except ImportError:
+    from collections import Mapping
 
 from parso._compatibility import utf8_repr, unicode
 from parso.tree import Node, BaseNode, Leaf, ErrorNode, ErrorLeaf, \

test/test_diff_parser.py

@@ -974,10 +974,12 @@ def test_random_unicode_characters(differ):
     Those issues were all found with the fuzzer.
     """
     differ.initialize('')
-    differ.parse(u'\x1dĔBϞɛˁşʑ˳˻ȣſéÎ\x90̕ȟòwʘ\x1dĔBϞɛˁşʑ˳˻ȣſéÎ', parsers=1, expect_error_leaves=True)
+    differ.parse(u'\x1dĔBϞɛˁşʑ˳˻ȣſéÎ\x90̕ȟòwʘ\x1dĔBϞɛˁşʑ˳˻ȣſéÎ', parsers=1,
+                 expect_error_leaves=True)
     differ.parse(u'\r\r', parsers=1)
     differ.parse(u"˟Ę\x05À\r   rúƣ@\x8a\x15r()\n", parsers=1, expect_error_leaves=True)
-    differ.parse(u'a\ntaǁ\rGĒōns__\n\nb', parsers=1)
+    differ.parse(u'a\ntaǁ\rGĒōns__\n\nb', parsers=1,
+                 expect_error_leaves=sys.version_info[0] == 2)
     s = '        if not (self, "_fi\x02\x0e\x08\n\nle"):'
     differ.parse(s, parsers=1, expect_error_leaves=True)
     differ.parse('')

test/test_fstring.py

@@ -12,33 +12,57 @@ def grammar():
 
 @pytest.mark.parametrize(
     'code', [
-        '{1}',
-        '{1:}',
-        '',
-        '{1!a}',
-        '{1!a:1}',
-        '{1:1}',
-        '{1:1.{32}}',
-        '{1::>4}',
-        '{foo} {bar}',
-        '{x:{y}}',
-        '{x:{y:}}',
-        '{x:{y:1}}',
+        # simple cases
+        'f"{1}"',
+        'f"""{1}"""',
+        'f"{foo} {bar}"',
+
+        # empty string
+        'f""',
+        'f""""""',
+
+        # empty format specifier is okay
+        'f"{1:}"',
+
+        # use of conversion options
+        'f"{1!a}"',
+        'f"{1!a:1}"',
+
+        # format specifiers
+        'f"{1:1}"',
+        'f"{1:1.{32}}"',
+        'f"{1::>4}"',
+        'f"{x:{y}}"',
+        'f"{x:{y:}}"',
+        'f"{x:{y:1}}"',
 
         # Escapes
-        '{{}}',
-        '{{{1}}}',
-        '{{{1}',
-        '1{{2{{3',
-        '}}',
+        'f"{{}}"',
+        'f"{{{1}}}"',
+        'f"{{{1}"',
+        'f"1{{2{{3"',
+        'f"}}"',
 
         # New Python 3.8 syntax f'{a=}'
-        '{a=}',
-        '{a()=}',
+        'f"{a=}"',
+        'f"{a()=}"',
+
+        # multiline f-string
+        'f"""abc\ndef"""',
+        'f"""abc{\n123}def"""',
+
+        # a line continuation inside of an fstring_string
+        'f"abc\\\ndef"',
+        'f"\\\n{123}\\\n"',
+
+        # a line continuation inside of an fstring_expr
+        'f"{\\\n123}"',
+
+        # a line continuation inside of an format spec
+        'f"{123:.2\\\nf}"',
     ]
 )
 def test_valid(code, grammar):
-    code = 'f"""%s"""' % code
     module = grammar.parse(code, error_recovery=False)
     fstring = module.children[0]
     assert fstring.type == 'fstring'
@@ -47,23 +71,34 @@ def test_valid(code, grammar):
 
 @pytest.mark.parametrize(
     'code', [
-        '}',
-        '{',
-        '{1!{a}}',
-        '{!{a}}',
-        '{}',
-        '{:}',
-        '{:}}}',
-        '{:1}',
-        '{!:}',
-        '{!}',
-        '{!a}',
-        '{1:{}}',
-        '{1:{:}}',
+        # an f-string can't contain unmatched curly braces
+        'f"}"',
+        'f"{"',
+        'f"""}"""',
+        'f"""{"""',
+
+        # invalid conversion characters
+        'f"{1!{a}}"',
+        'f"{!{a}}"',
+
+        # The curly braces must contain an expression
+        'f"{}"',
+        'f"{:}"',
+        'f"{:}}}"',
+        'f"{:1}"',
+        'f"{!:}"',
+        'f"{!}"',
+        'f"{!a}"',
+
+        # invalid (empty) format specifiers
+        'f"{1:{}}"',
+        'f"{1:{:}}"',
+
+        # a newline without a line continuation inside a single-line string
+        'f"abc\ndef"',
     ]
 )
 def test_invalid(code, grammar):
-    code = 'f"""%s"""' % code
     with pytest.raises(ParserSyntaxError):
         grammar.parse(code, error_recovery=False)
 
@@ -95,6 +130,7 @@ def test_tokenize_start_pos(code, positions):
             """),
         'f"foo',
         'f"""foo',
+        'f"abc\ndef"',
     ]
 )
 def test_roundtrip(grammar, code):

test/test_tokenize.py

@@ -1,5 +1,6 @@
 # -*- coding: utf-8    # This file contains Unicode characters.
 
+import sys
 from textwrap import dedent
 
 import pytest
@@ -16,6 +17,7 @@ from parso.python.tokenize import PythonToken
 NAME = PythonTokenTypes.NAME
 NEWLINE = PythonTokenTypes.NEWLINE
 STRING = PythonTokenTypes.STRING
+NUMBER = PythonTokenTypes.NUMBER
 INDENT = PythonTokenTypes.INDENT
 DEDENT = PythonTokenTypes.DEDENT
 ERRORTOKEN = PythonTokenTypes.ERRORTOKEN
@@ -140,7 +142,7 @@ def test_identifier_contains_unicode():
     else:
         # Unicode tokens in Python 2 seem to be identified as operators.
         # They will be ignored in the parser, that's ok.
-        assert unicode_token[0] == OP
+        assert unicode_token[0] == ERRORTOKEN
 
 
 def test_quoted_strings():
@@ -228,16 +230,29 @@ def test_endmarker_end_pos():
     check('a\\')
 
 
+xfail_py2 = dict(marks=[pytest.mark.xfail(sys.version_info[0] == 2, reason='Python 2')])
+
+
 @pytest.mark.parametrize(
     ('code', 'types'), [
+        # Indentation
         (' foo', [INDENT, NAME, DEDENT]),
         ('  foo\n bar', [INDENT, NAME, NEWLINE, ERROR_DEDENT, NAME, DEDENT]),
         ('  foo\n bar \n baz', [INDENT, NAME, NEWLINE, ERROR_DEDENT, NAME,
                                 NEWLINE, ERROR_DEDENT, NAME, DEDENT]),
         (' foo\nbar', [INDENT, NAME, NEWLINE, DEDENT, NAME]),
+
+        # Name stuff
+        ('1foo1', [NUMBER, NAME]),
+        pytest.param(
+            u'மெல்லினம்', [NAME],
+            **xfail_py2),
+        pytest.param(u'²', [ERRORTOKEN], **xfail_py2),
+        pytest.param(u'ä²ö', [NAME, ERRORTOKEN, NAME], **xfail_py2),
+        pytest.param(u'ää²¹öö', [NAME, ERRORTOKEN, NAME], **xfail_py2),
     ]
 )
-def test_indentation(code, types):
+def test_token_types(code, types):
     actual_types = [t.type for t in _get_token_list(code)]
     assert actual_types == types + [ENDMARKER]
 
@@ -330,13 +345,46 @@ def test_backslash():
         ('f" "{}', [FSTRING_START, FSTRING_STRING, FSTRING_END, OP, OP]),
         (r'f"\""', [FSTRING_START, FSTRING_STRING, FSTRING_END]),
         (r'f"\""', [FSTRING_START, FSTRING_STRING, FSTRING_END]),
+
+        # format spec
         (r'f"Some {x:.2f}{y}"', [FSTRING_START, FSTRING_STRING, OP, NAME, OP,
                                  FSTRING_STRING, OP, OP, NAME, OP, FSTRING_END]),
+
+        # multiline f-string
+        ('f"""abc\ndef"""', [FSTRING_START, FSTRING_STRING, FSTRING_END]),
+        ('f"""abc{\n123}def"""', [
+            FSTRING_START, FSTRING_STRING, OP, NUMBER, OP, FSTRING_STRING,
+            FSTRING_END
+        ]),
+
+        # a line continuation inside of an fstring_string
+        ('f"abc\\\ndef"', [
+            FSTRING_START, FSTRING_STRING, FSTRING_END
+        ]),
+        ('f"\\\n{123}\\\n"', [
+            FSTRING_START, FSTRING_STRING, OP, NUMBER, OP, FSTRING_STRING,
+            FSTRING_END
+        ]),
+
+        # a line continuation inside of an fstring_expr
+        ('f"{\\\n123}"', [FSTRING_START, OP, NUMBER, OP, FSTRING_END]),
+
+        # a line continuation inside of an format spec
+        ('f"{123:.2\\\nf}"', [
+            FSTRING_START, OP, NUMBER, OP, FSTRING_STRING, OP, FSTRING_END
+        ]),
+
+        # a newline without a line continuation inside a single-line string is
+        # wrong, and will generate an ERRORTOKEN
+        ('f"abc\ndef"', [
+            FSTRING_START, FSTRING_STRING, NEWLINE, NAME, ERRORTOKEN
+        ]),
+
+        # a more complex example
         (r'print(f"Some {x:.2f}a{y}")', [
             NAME, OP, FSTRING_START, FSTRING_STRING, OP, NAME, OP,
             FSTRING_STRING, OP, FSTRING_STRING, OP, NAME, OP, FSTRING_END, OP
         ]),
-
     ]
 )
 def test_fstring(code, types, version_ge_py36):

tox.ini

@@ -4,6 +4,7 @@ envlist = {py26,py27,py33,py34,py35,py36,py37}
 extras = testing
 deps =
     py26,py33: pytest>=3.0.7,<3.3
+    py27,py34: pytest<5
     py26,py33: setuptools<37
     coverage: coverage
 setenv =