Add runtime type checks with omlish.check and annotate ignored type errors

wrmsr · wrmsr · commit f450c3dad106 · 2026-05-26T11:40:36.000-07:00
Replace manual isinstance calls with check.isinstance across the parser, update imports, add type: ignore comments where needed, simplify list initializations, and tighten docstring and comment formatting. Minor whitespace and line‑break adjustments are also applied.
diff --git a/omxtra/text/pdcmark/blocks/machine.py b/omxtra/text/pdcmark/blocks/machine.py
@@ -19,6 +19,7 @@
   4. Dispatch the remaining line content to leaf-block recognition (ATX heading, hrule, fence,
      HTML block, indented code, paragraph).
 """
+from omlish import check
 from omlish import dataclasses as dc
 
 from ..events import BlockQuote
@@ -229,7 +230,7 @@ def _process_line(self, bl: BufferedLine, events: list[Event]) -> None:
             while (
                 matched_depth > 0
                 and isinstance(self._stack[matched_depth - 1], OpenList)
-                and not self._line_opens_matching_item(self._stack[matched_depth - 1], ls)
+                and not self._line_opens_matching_item(check.isinstance(self._stack[matched_depth - 1], OpenList), ls)
             ):
                 matched_depth -= 1
 
diff --git a/omxtra/text/pdcmark/inlines/emphasis.py b/omxtra/text/pdcmark/inlines/emphasis.py
@@ -43,8 +43,7 @@ def resolve_emphasis(nodes: list[InlineNode]) -> list[InlineNode]:
             match_s = _find_matching_opener(nodes, delim_stack, node)
             if match_s is not None:
                 opener_i = delim_stack[match_s]
-                opener = nodes[opener_i]
-                check.isinstance(opener, DelimNode)
+                opener = check.isinstance(nodes[opener_i], DelimNode)
 
                 # GFM strikethrough: `~` and `~~` both produce <del>; longer runs (`~~~+`) don't pair at all. Equivalent
                 # to the GFM "tilde must be ≤ 2 chars" rule.
diff --git a/omxtra/text/pdcmark/inlines/links.py b/omxtra/text/pdcmark/inlines/links.py
@@ -95,8 +95,7 @@ def resolve_links(
                 continue
             match_s = len(stack) - 1
             entry = stack[match_s]
-            opener = nodes[entry.node_index]
-            check.isinstance(opener, LinkOpenNode)
+            opener = check.isinstance(nodes[entry.node_index], LinkOpenNode)
 
             # Inner children = nodes strictly between opener and closer.
             children = nodes[entry.node_index + 1:i]
diff --git a/omxtra/text/pdcmark/inlines/parser.py b/omxtra/text/pdcmark/inlines/parser.py
@@ -104,6 +104,7 @@ def _resolve_emphasis_recursive(nodes: list[InlineNode]) -> None:
 
 
 def _emit_events(nodes: ta.Iterable[InlineNode], out: list[Event]) -> None:
+    tag: Tag
     for n in nodes:
         if isinstance(n, TextNode):
             if n.text:
@@ -124,7 +125,6 @@ def _emit_events(nodes: ta.Iterable[InlineNode], out: list[Event]) -> None:
         elif isinstance(n, HardBreakNode):
             out.append(HardBreak(offset=n.offset))
         elif isinstance(n, EmphasisGroup):
-            tag: Tag
             if n.kind == 'strong':
                 tag = Strong()
             elif n.kind == 'strikethrough':
@@ -135,7 +135,7 @@ def _emit_events(nodes: ta.Iterable[InlineNode], out: list[Event]) -> None:
             _emit_events(n.children, out)
             out.append(End(offset=n.offset, tag=tag))
         elif isinstance(n, LinkGroup):
-            tag: Tag = (
+            tag = (
                 Image(link_type=n.link_type, dest_url=n.dest_url, title=n.title, id=n.id)
                 if n.is_image
                 else Link(link_type=n.link_type, dest_url=n.dest_url, title=n.title, id=n.id)
diff --git a/omxtra/text/pdcmark/scanning/links.py b/omxtra/text/pdcmark/scanning/links.py
@@ -101,7 +101,7 @@ def scan_link_destination(text: str, start: int, *, max_parens: int = 32) -> Lin
     # Bare form: balanced parens; terminates at first ASCII control / space / unbalanced ')'.
     i = start
     depth = 0
-    out: list[str] = []
+    out = []
     while i < n:
         c = text[i]
         if ord(c) < 0x20 or c == ' ':
diff --git a/omxtra/text/pdcmark/tests/conftest.py b/omxtra/text/pdcmark/tests/conftest.py
@@ -3,8 +3,8 @@
 import pytest
 
 
-# Resolve a directory from a test-relative path. The pulldown-cmark submodule lives at the repo
-# root; tests reference fixture files there.
+# Resolve a directory from a test-relative path. The pulldown-cmark submodule lives at the repo root; tests reference
+# fixture files there.
 REPO_ROOT = os.path.normpath(os.path.join(os.path.dirname(__file__), '..', '..'))
 PDCMARK_SUBMODULE = os.path.join(REPO_ROOT, 'pulldown-cmark', 'pulldown-cmark')
 
diff --git a/omxtra/text/pdcmark/tests/spec_runner.py b/omxtra/text/pdcmark/tests/spec_runner.py
@@ -9,14 +9,13 @@
     <expected html>
     ````````````````````````````````
 
-The fence is a long run of backticks (>= 3, in practice 32). The opening fence is followed by a
-keyword — `example` for tests that run, `DISABLED example` (or `... DISABLED`) for tests skipped.
-Inside, a `.` on its own line separates the markdown source from the expected HTML. Section titles
-(setext-style `=== ` underlines or `## …` headings) are picked up as group names for nicer pytest
-IDs.
-
-Tabs in spec fixtures are written as the literal arrow character `→` (U+2192) and substituted back
-to `\\t` at load time, matching how the CommonMark spec source is authored.
+The fence is a long run of backticks (>= 3, in practice 32). The opening fence is followed by a keyword — `example` for
+tests that run, `DISABLED example` (or `... DISABLED`) for tests skipped. Inside, a `.` on its own line separates the
+markdown source from the expected HTML. Section titles (setext-style `=== ` underlines or `## …` headings) are picked up
+as group names for nicer pytest IDs.
+
+Tabs in spec fixtures are written as the literal arrow character `→` (U+2192) and substituted back to `\\t` at load
+time, matching how the CommonMark spec source is authored.
 """
 import io
 import os.path
@@ -29,8 +28,8 @@
 ##
 
 
-# Opening fence: at least 3 backticks followed by anything before EOL. We match `example` and
-# `DISABLED example` (in either order, case-insensitive).
+# Opening fence: at least 3 backticks followed by anything before EOL. We match `example` and `DISABLED example` (in
+# either order, case-insensitive).
 _FENCE_RE = re.compile(r'^(`{3,})\s*(.*)$')
 
 _SEPARATOR_LINE = '.'
diff --git a/omxtra/text/pdcmark/tests/test_blocks/test_containers.py b/omxtra/text/pdcmark/tests/test_blocks/test_containers.py
@@ -82,19 +82,19 @@ def test_simple_bullet_list():
     tags = [s.tag.__class__.__name__ for s in starts]
     assert tags == ['List', 'Item', 'Paragraph', 'Item', 'Paragraph']
     list_start = next(e for e in out if isinstance(e, m.Start) and isinstance(e.tag, m.List))
-    assert list_start.tag.start is None  # unordered
+    assert list_start.tag.start is None  # type: ignore  # unordered
 
 
 def test_ordered_list():
     out = feed('1. foo\n2. bar\n')
     list_start = next(e for e in out if isinstance(e, m.Start) and isinstance(e.tag, m.List))
-    assert list_start.tag.start == 1
+    assert list_start.tag.start == 1  # type: ignore
 
 
 def test_ordered_list_arbitrary_start():
     out = feed('42. foo\n')
     list_start = next(e for e in out if isinstance(e, m.Start) and isinstance(e.tag, m.List))
-    assert list_start.tag.start == 42
+    assert list_start.tag.start == 42  # type: ignore
 
 
 def test_different_marker_starts_new_list():
diff --git a/omxtra/text/pdcmark/tests/test_blocks/test_leaf_machine.py b/omxtra/text/pdcmark/tests/test_blocks/test_leaf_machine.py
@@ -44,7 +44,7 @@ def test_atx_emits_heading():
 def test_atx_levels():
     for level in range(1, 7):
         out = feed('#' * level + ' h\n')
-        assert isinstance(out[0], m.Start) and out[0].tag.level == level
+        assert isinstance(out[0], m.Start) and out[0].tag.level == level  # type: ignore
 
 
 def test_paragraph_emits_paragraph():
@@ -78,9 +78,9 @@ def test_hrule_alternatives():
 
 def test_setext_h1_and_h2():
     out = feed('Title\n===\n')
-    assert isinstance(out[0], m.Start) and out[0].tag.level == 1
+    assert isinstance(out[0], m.Start) and out[0].tag.level == 1  # type: ignore
     out = feed('Title\n---\n')
-    assert isinstance(out[0], m.Start) and out[0].tag.level == 2
+    assert isinstance(out[0], m.Start) and out[0].tag.level == 2  # type: ignore
 
 
 def test_setext_only_applies_to_open_paragraph():
@@ -156,8 +156,8 @@ def test_atx_interrupts_paragraph():
 
 def test_hrule_interrupts_paragraph():
     out = feed('para\n---\n')
-    # Tricky case: `para\n---\n` is actually a setext-H2, NOT paragraph + hrule, per CM spec.
-    # Setext promotion wins because the underline is checked before paragraph-interrupters.
+    # Tricky case: `para\n---\n` is actually a setext-H2, NOT paragraph + hrule, per CM spec. Setext promotion wins
+    # because the underline is checked before paragraph-interrupters.
     assert isinstance(out[0].tag, m.Heading) and out[0].tag.level == 2
 
 
diff --git a/omxtra/text/pdcmark/tests/test_blocks/test_refdefs.py b/omxtra/text/pdcmark/tests/test_blocks/test_refdefs.py
@@ -114,8 +114,8 @@ def test_refdef_then_blank_then_use():
     tags = [type(s.tag).__name__ for s in starts]
     assert tags == ['Paragraph', 'Link']
     link_start = starts[1]
-    assert link_start.tag.dest_url == '/url'
-    assert link_start.tag.link_type == m.LinkType.SHORTCUT
+    assert link_start.tag.dest_url == '/url'  # type: ignore
+    assert link_start.tag.link_type == m.LinkType.SHORTCUT  # type: ignore
     assert 'foo' in refdefs
 
 
@@ -125,17 +125,17 @@ def test_first_definition_wins():
 
 
 def test_setext_paragraph_is_not_refdef():
-    # A paragraph that becomes a setext heading should NOT have refdefs peeled —
-    # `[foo]: /url\n===` is a level-1 heading with content `[foo]: /url`.
+    # A paragraph that becomes a setext heading should NOT have refdefs peeled — `[foo]: /url\n===` is a level-1 heading
+    # with content `[foo]: /url`.
     events, refdefs = _feed('[foo]: /url\n===\n')
     assert any(isinstance(e, m.Start) and isinstance(e.tag, m.Heading) for e in events)
     assert 'foo' not in refdefs
 
 
 def test_non_refdef_first_line_stops_consumption():
     events, refdefs = _feed('not a refdef\n[foo]: /url\n')
-    # The whole thing is ONE paragraph (the second line is paragraph continuation, not a fresh
-    # block). Refdef is NOT peeled because it's mid-paragraph.
+    # The whole thing is ONE paragraph (the second line is paragraph continuation, not a fresh block). Refdef is NOT
+    # peeled because it's mid-paragraph.
     paragraphs = [e for e in events if isinstance(e, m.Start) and isinstance(e.tag, m.Paragraph)]
     assert len(paragraphs) == 1
     assert 'foo' not in refdefs
diff --git a/omxtra/text/pdcmark/tests/test_inlines/test_dos.py b/omxtra/text/pdcmark/tests/test_inlines/test_dos.py
@@ -1,8 +1,7 @@
 """
-Adversarial input tests — the link-reference expansion-bomb that motivates pulldown's
-`link_ref_expansion_limit` (see pulldown-cmark issue #844 and parse.rs::ParserInner). Without the
-fuel guard, an input that repeatedly references the same large refdef can balloon output size
-quadratically.
+Adversarial input tests — the link-reference expansion-bomb that motivates pulldown's `link_ref_expansion_limit` (see
+pulldown-cmark issue #844 and parse.rs::ParserInner). Without the fuel guard, an input that repeatedly references the
+same large refdef can balloon output size quadratically.
 """
 import time
 
@@ -14,8 +13,8 @@
 
 
 def test_link_ref_expansion_bomb_stays_bounded():
-    # `[x]: <large>` followed by N copies of `[x]` would, without fuel, produce N copies of <large>
-    # in the rendered output. With fuel, expansion bails after the configured budget.
+    # `[x]: <large>` followed by N copies of `[x]` would, without fuel, produce N copies of <large> in the rendered
+    # output. With fuel, expansion bails after the configured budget.
     big = 'A' * 5000
     refs = '\n[x]\n' * 200
     src = f'[x]: {big}\n{refs}'
diff --git a/omxtra/text/pdcmark/tests/test_inlines/test_tokenize.py b/omxtra/text/pdcmark/tests/test_inlines/test_tokenize.py
@@ -26,8 +26,7 @@ def test_plain_text():
 
 def test_backslash_escape():
     out = _tokens(r'a\*b')
-    # Tokenizer accumulates `a`, `*`, `b` into the same buffer (escape consumes backslash, emits
-    # the punct as text).
+    # Tokenizer accumulates `a`, `*`, `b` into the same buffer (escape consumes backslash, emits the punct as text).
     text_parts = [t.text for t in out if isinstance(t, TextNode)]
     assert ''.join(text_parts) == 'a*b'
     # No DelimNode produced.
diff --git a/omxtra/text/pdcmark/tests/test_offsets.py b/omxtra/text/pdcmark/tests/test_offsets.py
@@ -1,16 +1,16 @@
 """
 Source-offset consistency checks.
 
-Every event carries `offset = (start, end)` — character indices into the absolute input stream.
-We assert a small set of invariants over the CommonMark spec corpus:
+Every event carries `offset = (start, end)` — character indices into the absolute input stream. We assert a small set of
+invariants over the CommonMark spec corpus:
 
   - 0 <= start <= end <= len(input)
-  - For Start/End pairs of the same Tag: enclosing event's start <= every inner offset start, and
-    enclosing event's end >= every inner offset end.
+  - For Start/End pairs of the same Tag: enclosing event's start <= every inner offset start, and enclosing event's end
+    >= every inner offset end.
   - For leaf events (Text, Code, Html, …) the offsets must be inside their enclosing Start/End.
 
-We don't enforce strict monotonicity (Text events can have empty spans, autolinks emit Start/Text/End
-all with the same offset, etc.), but we DO assert no event reports an offset outside [0, len(input)].
+We don't enforce strict monotonicity (Text events can have empty spans, autolinks emit Start/Text/End all with the same
+offset, etc.), but we DO assert no event reports an offset outside [0, len(input)].
 """
 import os.path
 
@@ -47,12 +47,11 @@ def _bad_offsets(events, src_len) -> list[str]:
                 issues.append(f'event #{ix} End with no matching Start')
                 continue
             tag, span, sx = open_stack.pop()
-            # End's offset should be the same (start, end) as the Start's — both span the whole
-            # element.
+            # End's offset should be the same (start, end) as the Start's — both span the whole element.
             if ev.offset[0] != span[0] or ev.offset[1] != span[1]:
-                # NOTE: a couple of constructs (e.g. blockquotes whose final newline is missing
-                # because the document ended on an unterminated line) reasonably have End.end >
-                # Start.end. We tolerate end-difference but not start-difference.
+                # NOTE: a couple of constructs (e.g. blockquotes whose final newline is missing because the document
+                # ended on an unterminated line) reasonably have End.end > Start.end. We tolerate end-difference but not
+                # start-difference.
                 if ev.offset[0] != span[0]:
                     issues.append(
                         f'End #{ix} span {ev.offset} doesn\'t match opening Start #{sx} span {span}',
@@ -78,7 +77,7 @@ def test_offsets_within_bounds_cm(cm_inputs):
     for ix, src in enumerate(cm_inputs):
         events = parse(src)
         for ev in events:
-            s, e = ev.offset
+            s, e = ev.offset  # type: ignore
             if not (0 <= s <= e <= len(src)):
                 failures.append((ix, ev))
                 break
@@ -104,7 +103,7 @@ def test_offset_slice_round_trips_basic():
     events = parse(src)
     text_events = [e for e in events if isinstance(e, m.Text)]
     for t in text_events:
-        # The slice should equal the text content (possibly modulo trivial whitespace —
-        # the inline parser strips leading whitespace from paragraph lines).
+        # The slice should equal the text content (possibly modulo trivial whitespace — the inline parser strips leading
+        # whitespace from paragraph lines).
         slice_ = src[t.offset[0]:t.offset[1]]
         assert t.text in slice_ or slice_.strip() == t.text.strip()
diff --git a/omxtra/text/pdcmark/tests/test_scanners/test_lists.py b/omxtra/text/pdcmark/tests/test_scanners/test_lists.py
@@ -25,7 +25,7 @@ def test_ordered_dot():
 
 def test_ordered_paren():
     m = scan_list_marker('42) item')
-    assert m.is_ordered and m.start == 42 and m.char == ')'
+    assert m.is_ordered and m.start == 42 and m.char == ')'  # type: ignore
 
 
 def test_ordered_max_digits():
diff --git a/omxtra/text/pdcmark/tests/test_spec_commonmark.py b/omxtra/text/pdcmark/tests/test_spec_commonmark.py
@@ -101,8 +101,8 @@ def test_cm_section_full_pass(cm_cases, section):
 # A small set of CM spec case indices that exercise specific M1 features and must pass. If any fails, something specific
 # has regressed.
 _CURATED_CASE_INDICES = (
-    # Numbers reference upstream CM 0.30 spec.txt indices. Cases listed here are exact-match
-    # smoke tests covering features we explicitly want green on each milestone.
+    # Numbers reference upstream CM 0.30 spec.txt indices. Cases listed here are exact-match smoke tests covering
+    # features we explicitly want green on each milestone.
     # M1 — block-level:
     1,    # tab in indented code
     43,   # thematic break (* / - / _)
diff --git a/omxtra/text/pdcmark/tests/test_spec_gfm.py b/omxtra/text/pdcmark/tests/test_spec_gfm.py
@@ -1,7 +1,7 @@
 """
-Integration tests against the GFM-extension spec fixtures (vendored under the pulldown-cmark
-submodule's third_party/GitHub/). Tables / strikethrough / tasklists are required to be enabled
-via the GFM preset; without enabling the extensions these tests are not meaningful.
+Integration tests against the GFM-extension spec fixtures (vendored under the pulldown-cmark submodule's
+third_party/GitHub/). Tables / strikethrough / tasklists are required to be enabled via the GFM preset; without enabling
+the extensions these tests are not meaningful.
 """
 import os.path
 
diff --git a/omxtra/text/pdcmark/tests/test_streaming/test_basic.py b/omxtra/text/pdcmark/tests/test_streaming/test_basic.py
@@ -65,8 +65,7 @@ def test_tentative_updates_across_chunks():
 def test_open_code_block_tentative():
     sp = StreamingParser()
     fo = sp.feed('```py\nx = 1\n')
-    # The fence opened and one line of body went in. Tentative shows the open code block as if
-    # closed now.
+    # The fence opened and one line of body went in. Tentative shows the open code block as if closed now.
     assert any(isinstance(e, m.Start) and isinstance(e.tag, m.FencedCodeBlock) for e in fo.tentative)
     texts = [e.text for e in fo.tentative if isinstance(e, m.Text)]
     assert texts == ['x = 1\n']
@@ -76,7 +75,7 @@ def test_committed_appended_only():
     """The same event should NOT appear in both committed and tentative across feeds."""
 
     sp = StreamingParser()
-    all_committed = []
+    all_committed: list = []
     chunks = ['# heading\n', 'paragraph ', 'more\n', '\n', 'second para\n']
     for chunk in chunks:
         fo = sp.feed(chunk)
@@ -98,10 +97,7 @@ def test_crlf_handling():
 
 
 def test_cr_at_chunk_boundary():
-    """
-    CR ending one chunk + LF starting the next must form a single CRLF terminator, not two
-    separate line breaks.
-    """
+    """CR ending one chunk + LF starting the next must form a single CRLF terminator, not two separate line breaks."""
 
     sp = StreamingParser()
     sp.feed('a\r')
@@ -111,7 +107,7 @@ def test_cr_at_chunk_boundary():
     # Reconstruct from feeds. (We need to capture them all.)
     sp = StreamingParser()
     parts = ['a\r', '\nb\n']
-    out = []
+    out: list = []
     for p in parts:
         out.extend(sp.feed(p).committed)
     out.extend(sp.finish().committed)
diff --git a/omxtra/text/pdcmark/tests/test_streaming/test_equivalence.py b/omxtra/text/pdcmark/tests/test_streaming/test_equivalence.py
