v0.0.13

fix highlighting hang with empty begin end rules

README.md

@@ -12,7 +12,7 @@ a text editor, eventually...
 
 ### why is it called babi?
 
-I usually use the text editor `nano`, frequently I typo this.  on a qwerty
+I used to use the text editor `nano`, frequently I typo this.  on a qwerty
 keyboard, when the right hand is shifted left by one, `nano` becomes `babi`.
 
 ### quitting babi

babi/highlight.py

@@ -273,6 +273,7 @@ class CompiledRegsetRule(CompiledRule, Protocol):
 class Entry(NamedTuple):
     scope: Tuple[str, ...]
     rule: CompiledRule
+    start: Tuple[str, int]
     reg: _Reg = ERR_REG
     boundary: bool = False
 
@@ -284,7 +285,7 @@ def _inner_capture_parse(
         scope: Scope,
         rule: CompiledRule,
 ) -> Regions:
-    state = State.root(Entry(scope + rule.name, rule))
+    state = State.root(Entry(scope + rule.name, rule, (s, 0)))
     _, regions = highlight_line(compiler, state, s, first_line=False)
     return tuple(
         r._replace(start=r.start + start, end=r.end + start) for r in regions
@@ -440,7 +441,8 @@ class EndRule(NamedTuple):
 
         boundary = match.end() == len(match.string)
         reg = make_reg(expand_escaped(match, self.end))
-        state = state.push(Entry(next_scope, self, reg, boundary))
+        start = (match.string, match.start())
+        state = state.push(Entry(next_scope, self, start, reg, boundary))
         regions = _captures(compiler, scope, match, self.begin_captures)
         return state, True, regions
 
@@ -455,7 +457,16 @@ class EndRule(NamedTuple):
         if m.start() > pos:
             ret.append(Region(pos, m.start(), state.cur.scope))
         ret.extend(_captures(compiler, state.cur.scope, m, self.end_captures))
-        return state.pop(), m.end(), False, tuple(ret)
+        # this is probably a bug in the grammar, but it pushed and popped at
+        # the same position.
+        # we'll advance the highlighter by one position to get past the loop
+        # this appears to be what vs code does as well
+        if state.entries[-1].start == (m.string, m.end()):
+            ret.append(Region(m.end(), m.end() + 1, state.cur.scope))
+            end = m.end() + 1
+        else:
+            end = m.end()
+        return state.pop(), end, False, tuple(ret)
 
     def search(
             self,
@@ -501,7 +512,9 @@ class WhileRule(NamedTuple):
 
         boundary = match.end() == len(match.string)
         reg = make_reg(expand_escaped(match, self.while_))
-        state = state.push_while(self, Entry(next_scope, self, reg, boundary))
+        start = (match.string, match.start())
+        entry = Entry(next_scope, self, start, reg, boundary)
+        state = state.push_while(self, entry)
         regions = _captures(compiler, scope, match, self.begin_captures)
         return state, True, regions
 
@@ -541,7 +554,7 @@ class Compiler:
         self._rule_to_grammar: Dict[_Rule, Grammar] = {}
         self._c_rules: Dict[_Rule, CompiledRule] = {}
         root = self._compile_root(grammar)
-        self.root_state = State.root(Entry(root.name, root))
+        self.root_state = State.root(Entry(root.name, root, ('', 0)))
 
     def _visit_rule(self, grammar: Grammar, rule: _Rule) -> _Rule:
         self._rule_to_grammar[rule] = grammar

babi/main.py

@@ -2,6 +2,7 @@ import argparse
 import curses
 import os
 import re
+import signal
 import sys
 from typing import List
 from typing import Optional
@@ -138,6 +139,11 @@ def main(argv: Optional[Sequence[str]] = None) -> int:
     else:
         stdin = ''
 
+    # ignore backgrounding signals, we'll handle those in curses
+    # fixes a problem with ^Z on termination which would break the terminal
+    if sys.platform != 'win32':  # pragma: win32 no cover  # pragma: no branch
+        signal.signal(signal.SIGTSTP, signal.SIG_IGN)
+
     with perf_log(args.perf_log) as perf, make_stdscr() as stdscr:
         if args.key_debug:
             return _key_debug(stdscr, perf)

babi/reg.py

@@ -6,80 +6,36 @@ from typing import Tuple
 
 import onigurumacffi
 
-from babi.cached_property import cached_property
-
 _BACKREF_RE = re.compile(r'((?<!\\)(?:\\\\)*)\\([0-9]+)')
 
 
-def _replace_esc(s: str, chars: str) -> str:
-    """replace the given escape sequences of `chars` with \\uffff"""
-    for c in chars:
-        if f'\\{c}' in s:
-            break
-    else:
-        return s
-
-    b = []
-    i = 0
-    length = len(s)
-    while i < length:
-        try:
-            sbi = s.index('\\', i)
-        except ValueError:
-            b.append(s[i:])
-            break
-        if sbi > i:
-            b.append(s[i:sbi])
-        b.append('\\')
-        i = sbi + 1
-        if i < length:
-            if s[i] in chars:
-                b.append('\uffff')
-            else:
-                b.append(s[i])
-        i += 1
-    return ''.join(b)
+_FLAGS = {
+    # (first_line, boundary)
+    (False, False): (
+        onigurumacffi.OnigSearchOption.NOT_END_STRING |
+        onigurumacffi.OnigSearchOption.NOT_BEGIN_STRING |
+        onigurumacffi.OnigSearchOption.NOT_BEGIN_POSITION
+    ),
+    (False, True): (
+        onigurumacffi.OnigSearchOption.NOT_END_STRING |
+        onigurumacffi.OnigSearchOption.NOT_BEGIN_STRING
+    ),
+    (True, False): (
+        onigurumacffi.OnigSearchOption.NOT_END_STRING |
+        onigurumacffi.OnigSearchOption.NOT_BEGIN_POSITION
+    ),
+    (True, True): onigurumacffi.OnigSearchOption.NOT_END_STRING,
+}
 
 
 class _Reg:
     def __init__(self, s: str) -> None:
         self._pattern = s
+        self._reg = onigurumacffi.compile(self._pattern)
 
     def __repr__(self) -> str:
         return f'{type(self).__name__}({self._pattern!r})'
 
-    @cached_property
-    def _reg(self) -> onigurumacffi._Pattern:
-        return onigurumacffi.compile(_replace_esc(self._pattern, 'z'))
-
-    @cached_property
-    def _reg_no_A(self) -> onigurumacffi._Pattern:
-        return onigurumacffi.compile(_replace_esc(self._pattern, 'Az'))
-
-    @cached_property
-    def _reg_no_G(self) -> onigurumacffi._Pattern:
-        return onigurumacffi.compile(_replace_esc(self._pattern, 'Gz'))
-
-    @cached_property
-    def _reg_no_A_no_G(self) -> onigurumacffi._Pattern:
-        return onigurumacffi.compile(_replace_esc(self._pattern, 'AGz'))
-
-    def _get_reg(
-            self,
-            first_line: bool,
-            boundary: bool,
-    ) -> onigurumacffi._Pattern:
-        if boundary:
-            if first_line:
-                return self._reg
-            else:
-                return self._reg_no_A
-        else:
-            if first_line:
-                return self._reg_no_G
-            else:
-                return self._reg_no_A_no_G
-
     def search(
             self,
             line: str,
@@ -87,7 +43,7 @@ class _Reg:
             first_line: bool,
             boundary: bool,
     ) -> Optional[Match[str]]:
-        return self._get_reg(first_line, boundary).search(line, pos)
+        return self._reg.search(line, pos, flags=_FLAGS[first_line, boundary])
 
     def match(
             self,
@@ -96,36 +52,18 @@ class _Reg:
             first_line: bool,
             boundary: bool,
     ) -> Optional[Match[str]]:
-        return self._get_reg(first_line, boundary).match(line, pos)
+        return self._reg.match(line, pos, flags=_FLAGS[first_line, boundary])
 
 
 class _RegSet:
     def __init__(self, *s: str) -> None:
         self._patterns = s
+        self._set = onigurumacffi.compile_regset(*self._patterns)
 
     def __repr__(self) -> str:
         args = ', '.join(repr(s) for s in self._patterns)
         return f'{type(self).__name__}({args})'
 
-    @cached_property
-    def _set(self) -> onigurumacffi._RegSet:
-        return onigurumacffi.compile_regset(*self._patterns)
-
-    @cached_property
-    def _set_no_A(self) -> onigurumacffi._RegSet:
-        patterns = (_replace_esc(p, 'A') for p in self._patterns)
-        return onigurumacffi.compile_regset(*patterns)
-
-    @cached_property
-    def _set_no_G(self) -> onigurumacffi._RegSet:
-        patterns = (_replace_esc(p, 'G') for p in self._patterns)
-        return onigurumacffi.compile_regset(*patterns)
-
-    @cached_property
-    def _set_no_A_no_G(self) -> onigurumacffi._RegSet:
-        patterns = (_replace_esc(p, 'AG') for p in self._patterns)
-        return onigurumacffi.compile_regset(*patterns)
-
     def search(
             self,
             line: str,
@@ -133,16 +71,7 @@ class _RegSet:
             first_line: bool,
             boundary: bool,
     ) -> Tuple[int, Optional[Match[str]]]:
-        if boundary:
-            if first_line:
-                return self._set.search(line, pos)
-            else:
-                return self._set_no_A.search(line, pos)
-        else:
-            if first_line:
-                return self._set_no_G.search(line, pos)
-            else:
-                return self._set_no_A_no_G.search(line, pos)
+        return self._set.search(line, pos, flags=_FLAGS[first_line, boundary])
 
 
 def expand_escaped(match: Match[str], s: str) -> str:
@@ -151,4 +80,4 @@ def expand_escaped(match: Match[str], s: str) -> str:
 
 make_reg = functools.lru_cache(maxsize=None)(_Reg)
 make_regset = functools.lru_cache(maxsize=None)(_RegSet)
-ERR_REG = make_reg(')this pattern always triggers an error when used(')
+ERR_REG = make_reg('$ ^')

setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = babi
-version = 0.0.11
+version = 0.0.13
 description = a text editor
 long_description = file: README.md
 long_description_content_type = text/markdown
@@ -24,7 +24,7 @@ packages = find:
 install_requires =
     babi-grammars
     identify
-    onigurumacffi>=0.0.10
+    onigurumacffi>=0.0.18
     importlib_metadata>=1;python_version<"3.8"
     windows-curses;sys_platform=="win32"
 python_requires = >=3.6.1

tests/highlight_test.py

@@ -637,3 +637,25 @@ def test_backslash_z(compiler_state):
     assert regions2 == (
         Region(0, 6, ('test', 'comment')),
     )
+
+
+def test_buggy_begin_end_grammar(compiler_state):
+    # before this would result in an infinite loop of start / end
+    compiler, state = compiler_state({
+        'scopeName': 'test',
+        'patterns': [
+            {
+                'begin': '(?=</style)',
+                'end': '(?=</style)',
+                'name': 'css',
+            },
+        ],
+    })
+
+    state, regions = highlight_line(compiler, state, 'test </style', True)
+
+    assert regions == (
+        Region(0, 5, ('test',)),
+        Region(5, 6, ('test', 'css')),
+        Region(6, 12, ('test',)),
+    )

tests/reg_test.py

@@ -35,9 +35,8 @@ def test_reg_other_escapes_left_untouched():
 def test_reg_not_out_of_bounds_at_end():
     # the only way this is triggerable is with an illegal regex, we'd rather
     # produce an error about the regex being wrong than an IndexError
-    reg = _Reg('\\A\\')
     with pytest.raises(onigurumacffi.OnigError) as excinfo:
-        reg.search('\\', 0, first_line=False, boundary=False)
+        _Reg('\\A\\')
     msg, = excinfo.value.args
     assert msg == 'end pattern at escape'