Group quantized literals in regex optimization

CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## [Unreleased][unreleased]
 
+### Fixed
+
+* Group quantized 2+ char literals in regex optimization ([#54])
+
 
 ## [v0.5.3][]
 
@@ -215,3 +219,4 @@ descent parser and a work-in-progress state-machine parser.
 [#38]: https://github.com/goodmami/pe/issues/38
 [#44]: https://github.com/goodmami/pe/issues/44
 [#46]: https://github.com/goodmami/pe/issues/46
+[#54]: https://github.com/goodmami/pe/issues/54

docs/guides/using-flags.md

@@ -125,7 +125,7 @@ pe._errors.ParseError:
   line 0, character 8
     "one" 1 "two
             ^
-ParseError: `(?=(?P<_5>"(?=(?P<_2>[^"]*))(?P=_2)"|\-?(?=(?P<_4>0|[1-9](?=(?P<_3>[0-9]*))(?P=_3)))(?P=_4)))(?P=_5)`
+ParseError: `(?=(?P<_5>"(?=(?P<_2>[^"]*))(?P=_2)"|(?:\-)?(?=(?P<_4>0|[1-9](?=(?P<_3>[0-9]*))(?P=_3)))(?P=_4)))(?P=_5)`
 
 ```
 

pe/_optimize.py

@@ -242,7 +242,7 @@ def _regex_optional(defn, defs, grpid):
     subdef = defn.args[0]
     d = _regex(defn.args[0], defs, grpid)
     if d.op == RGX:
-        subpat = d.args[0] if subdef.op in (DOT, LIT, CLS) else f'(?:{d.args[0]})'
+        subpat = _regex_maybe_group(subdef.op, d.args[0])
         return Regex(f'{subpat}?')
     else:
         return Optional(d)
@@ -252,7 +252,7 @@ def _regex_star(defn, defs, grpid):
     subdef = defn.args[0]
     d = _regex(subdef, defs, grpid)
     if d.op == RGX:
-        subpat = d.args[0] if subdef.op in (DOT, LIT, CLS) else f'(?:{d.args[0]})'
+        subpat = _regex_maybe_group(subdef.op, d.args[0])
         gid = f'_{next(grpid)}'
         return Regex(f'(?=(?P<{gid}>{subpat}*))(?P={gid})')
     else:
@@ -263,13 +263,19 @@ def _regex_plus(defn, defs, grpid):
     subdef = defn.args[0]
     d = _regex(defn.args[0], defs, grpid)
     if d.op == RGX:
-        subpat = d.args[0] if subdef.op in (DOT, LIT, CLS) else f'(?:{d.args[0]})'
+        subpat = _regex_maybe_group(subdef.op, d.args[0])
         gid = f'_{next(grpid)}'
         return Regex(f'(?=(?P<{gid}>{subpat}+))(?P={gid})')
     else:
         return Plus(d)
 
 
+def _regex_maybe_group(op: Operator, arg: str) -> str:
+    if op in (DOT, CLS) or (op == LIT and len(arg) == 1):
+        return arg
+    return f'(?:{arg})'
+
+
 def _regex_and(defn, defs, grpid):
     d = _regex(defn.args[0], defs, grpid)
     if d.op == RGX:

pe/packrat.py

@@ -87,7 +87,7 @@ def match(self,
             else:
                 return None
 
-        args = tuple(args or ())
+        args = list(args or ())
         if kwargs is None:
             kwargs = {}
 

test/test_regression.py

@@ -40,3 +40,36 @@ def test_capture_repeated(parser):
     m3 = pe.match('(~"a")+', 'aaa', parser=parser)
     assert m3.group() == 'aaa'
     assert m3.groups() == ('a', 'a', 'a')
+
+
+@pytest.mark.parametrize('parser', ['packrat', 'machine', 'machine-python'])
+def test_regex_optimized_quantified_literal(parser):
+    """https://github.com/goodmami/pe/issues/54"""
+
+    p = pe.compile("Start <- 'ab'? 'c'", parser=parser, flags=pe.REGEX)
+    assert p.match("c", flags=pe.NONE).group(0) == "c"
+    assert p.match("abc", flags=pe.NONE).group(0) == "abc"
+    assert p.match("ababc", flags=pe.NONE) is None
+    assert p.match("ac", flags=pe.NONE) is None
+    assert p.match("abac", flags=pe.NONE) is None
+
+    p = pe.compile("Start <- 'ab'* 'c'", parser=parser, flags=pe.REGEX)
+    assert p.match("c", flags=pe.NONE).group(0) == "c"
+    assert p.match("abc", flags=pe.NONE).group(0) == "abc"
+    assert p.match("ababc", flags=pe.NONE).group(0) == "ababc"
+    assert p.match("ac", flags=pe.NONE) is None
+    assert p.match("abac", flags=pe.NONE) is None
+
+    p = pe.compile("Start <- 'ab'+ 'c'", parser=parser, flags=pe.REGEX)
+    assert p.match("c", flags=pe.NONE) is None
+    assert p.match("abc", flags=pe.NONE).group(0) == "abc"
+    assert p.match("ababc", flags=pe.NONE).group(0) == "ababc"
+    assert p.match("ac", flags=pe.NONE) is None
+    assert p.match("abac", flags=pe.NONE) is None
+
+    p = pe.compile("Start <- 'ab'{2} 'c'", parser=parser, flags=pe.REGEX)
+    assert p.match("c", flags=pe.NONE) is None
+    assert p.match("abc", flags=pe.NONE) is None
+    assert p.match("ababc", flags=pe.NONE).group(0) == "ababc"
+    assert p.match("ac", flags=pe.NONE) is None
+    assert p.match("abac", flags=pe.NONE) is None