👌 Make StateBase.src a plain attribute (~6% faster parsing)

CHANGELOG.md

@@ -2,6 +2,7 @@
 
 ## Unreleased
 
+* 👌 Speed up parsing (~6% on the CommonMark spec) by making `StateBase.src` a plain attribute instead of a property, in [#450](https://github.com/executablebooks/markdown-it-py/pull/450)
 * ✨ Add `--enable-tables` to the CLI for file, standard input and interactive parsing in [#422](https://github.com/executablebooks/markdown-it-py/pull/422)
 * 🐛 Fix CLI interactive mode joining input lines with an extra newline, which split every line into its own paragraph and broke hard line breaks, in [#172](https://github.com/executablebooks/markdown-it-py/issues/172)
 * 🐛 Fix trimming and splitting with the Python whitespace set instead of the CommonMark one, which dropped U+001C–U+001F and U+0085 from paragraphs, headings, table cells and fence info strings and let distinct reference labels resolve each other, in [#418](https://github.com/executablebooks/markdown-it-py/pull/418), thanks to [@Nexory](https://github.com/Nexory)

markdown_it/ruler.py

@@ -30,30 +30,30 @@ class Ruler
 
 
 class StateBase:
+    # `src` is a plain attribute, not a property: rules read it on almost
+    # every step, and a property call there costs ~5% of total parse time.
+    src: str
+
+    # Cache for the deprecated `srcCharCode`, as `(src, char_codes)`.
+    # This is a class-level default, since subclasses do not call `__init__`.
+    _src_char_code: tuple[str, tuple[int, ...]] | None = None
+
     def __init__(self, src: str, md: MarkdownIt, env: EnvType):
         self.src = src
         self.env = env
         self.md = md
 
-    @property
-    def src(self) -> str:
-        return self._src
-
-    @src.setter
-    def src(self, value: str) -> None:
-        self._src = value
-        self._srcCharCode: tuple[int, ...] | None = None
-
     @property
     def srcCharCode(self) -> tuple[int, ...]:
         warnings.warn(
             "StateBase.srcCharCode is deprecated. Use StateBase.src instead.",
             DeprecationWarning,
             stacklevel=2,
         )
-        if self._srcCharCode is None:
-            self._srcCharCode = tuple(ord(c) for c in self._src)
-        return self._srcCharCode
+        # rebuild the cache if `src` has been reassigned since it was built
+        if self._src_char_code is None or self._src_char_code[0] is not self.src:
+            self._src_char_code = (self.src, tuple(ord(c) for c in self.src))
+        return self._src_char_code[1]
 
 
 class RuleOptionsType(TypedDict, total=False):

tests/test_api/test_main.py

@@ -544,3 +544,15 @@ def peek_rule(state, silent):
     # the rule saw the text accumulate one character at a time
     assert seen[:4] == [0, 1, 2, 3]
     assert max(seen) == 9_999
+
+
+def test_src_char_code_follows_src():
+    """The deprecated ``srcCharCode`` is rebuilt when ``src`` is reassigned."""
+    from markdown_it.rules_block.state_block import StateBlock
+
+    state = StateBlock("ab", MarkdownIt(), {}, [])
+    with pytest.warns(DeprecationWarning):
+        assert state.srcCharCode == (97, 98)
+    state.src = "xyz"
+    with pytest.warns(DeprecationWarning):
+        assert list(state.srcCharCode) == [120, 121, 122]