fix(one-sentence-per-line): do not treat ellipsis (...) as sentence end mid-sentence

Treat ... as sentence boundary only when the next word is capitalized
(First... Then). Ellipsis in the middle of one sentence (config... supplied)
no longer triggers. Add processSentenceEnd helper for complexity; add
trailing-space and ellipsis tests; add md_test_files fixtures and
expected_errors.

Author: cypher0n3

markdownlint-rules/one-sentence-per-line.js

@@ -70,6 +70,28 @@ function isSentenceEndChar(ch) {
   return ch === "." || ch === "?" || ch === "!";
 }
 
+/** True when the period at i is part of an ellipsis (two or more consecutive dots). */
+function isPartOfEllipsis(scanned, i) {
+  if (scanned[i] !== ".") return false;
+  return (i > 0 && scanned[i - 1] === ".") || (i + 1 < scanned.length && scanned[i + 1] === ".");
+}
+
+/** Index of the last character of a run of dots starting at i (i is a dot). */
+function endOfEllipsisRun(scanned, i) {
+  let end = i;
+  while (end < scanned.length && scanned[end] === ".") end++;
+  return end;
+}
+
+/** True when the content after the ellipsis (after optional space) starts with an uppercase letter. */
+function capitalAfterEllipsis(scanned, lastDotIndex) {
+  const spaceStart = skipQuotesThenSpace(scanned, lastDotIndex);
+  if (spaceStart === null) return false;
+  const after = spaceStartAndNextWordPos(scanned, spaceStart);
+  if (!after) return false;
+  return /[A-Z]/.test(scanned[after.j]);
+}
+
 function skipQuotesAndSpaces(scanned, j) {
   let pos = j;
   while (pos < scanned.length && (scanned[pos] === "*" || scanned[pos] === "_")) {
@@ -203,6 +225,33 @@ function shouldSkipForSentenceBoundary(ctx) {
   return false;
 }
 
+/**
+ * Process a potential sentence-end at i; return next index and optional boundary.
+ * @returns {{ nextI: number, boundary: number|null }}
+ */
+function processSentenceEnd(scanned, i, abbreviations) {
+  const ch = scanned[i];
+  if (!isSentenceEndChar(ch)) return { nextI: i, boundary: null };
+  if (ch === "." && isPartOfEllipsis(scanned, i)) {
+    const end = endOfEllipsisRun(scanned, i);
+    const lastDotIndex = end - 1;
+    if (capitalAfterEllipsis(scanned, lastDotIndex)) {
+      const boundary = trySentenceBoundary(scanned, lastDotIndex, abbreviations);
+      if (boundary !== null) {
+        const { nextPos: j } = skipQuotesAndSpaces(scanned, lastDotIndex + 1);
+        return { nextI: j - 1, boundary };
+      }
+    }
+    return { nextI: end - 1, boundary: null };
+  }
+  const boundary = trySentenceBoundary(scanned, i, abbreviations);
+  if (boundary !== null) {
+    const { nextPos: j } = skipQuotesAndSpaces(scanned, i + 1);
+    return { nextI: j - 1, boundary };
+  }
+  return { nextI: i, boundary: null };
+}
+
 /**
  * Find all sentence boundary indices (space before each sentence after the first).
  * @param {string} content - Prose content (no list marker)
@@ -237,17 +286,9 @@ function getAllSentenceBoundaries(content, opts) {
       i++;
       continue;
     }
-    if (!isSentenceEndChar(ch)) {
-      i++;
-      continue;
-    }
-    const boundary = trySentenceBoundary(scanned, i, abbreviations);
-    if (boundary !== null) {
-      boundaries.push(boundary);
-      const { nextPos: j } = skipQuotesAndSpaces(scanned, i + 1);
-      i = j - 1;
-    }
-    i++;
+    const { nextI, boundary } = processSentenceEnd(scanned, i, abbreviations);
+    if (boundary !== null) boundaries.push(boundary);
+    i = nextI + 1;
   }
   return boundaries;
 }

md_test_files/README.md

@@ -55,7 +55,8 @@ Each item: **filename** - custom rule(s) that fail; sub-bullet - what the fixtur
 - **negative_no_tables.md** - no-tables
   - GFM table reported; with default config (convert-to none) short message only; suppress via `<!-- no-tables allow -->`.
 - **negative_one_sentence_per_line.md** - one-sentence-per-line
-  - Prose and list lines with multiple sentences (paragraph, bullet, numbered, nested); line with abbreviation (e.g.) not reported.
+  - Prose and list lines with multiple sentences (paragraph, bullet, numbered, nested); line with abbreviation (e.g.) not reported; ellipsis then new sentence (First... Then) reported.
+  - **positive_general.md** has a line with ellipsis in the middle of one sentence (not reported).
 
 Note: some negative fixtures intentionally trigger built-in markdownlint rules in addition to custom rules (e.g. MD031/MD032/MD033), so the test suite can assert multiple errors on specific lines.
 

md_test_files/expected_errors.yml

@@ -299,6 +299,9 @@ negative_one_sentence_per_line.md:
     - line: 22
       rule: one-sentence-per-line
       message_contains: one sentence per line
+    - line: 24
+      rule: one-sentence-per-line
+      message_contains: one sentence per line
 
 negative_no_tables.md:
   errors:

md_test_files/negative_one_sentence_per_line.md

@@ -20,3 +20,5 @@ This should not catch; e.g.: here is some text.
 - **No command or path allowlists inside the container.** The sandbox agent runs in an already-sandboxed environment (the container).
 
 This is the first sentence. **Bolded text** rest of the sentence.
+
+First... Then the next sentence.

md_test_files/positive_general.md

@@ -39,6 +39,7 @@ It is intended to be linted explicitly via `npx markdownlint-cli2 md_test_files/
 
 This paragraph has one sentence.
 This is another sentence on its own line.
+Configuration options... are supplied by the orchestrator in the **start bundle**.
 
 ## Lists
 

test-scripts/test_fix_one_sentence_per_line.py

@@ -173,6 +173,21 @@ def test_no_split_within_filenames(self) -> None:
             msg = f"no one-sentence-per-line errors expected: {proc.stderr}"
             self.assertEqual(proc.returncode, 0, msg)
 
+    def test_no_split_on_ellipsis_in_sentence(self) -> None:
+        """Ellipsis (...) in the middle of a sentence does not trigger one-sentence-per-line."""
+        content = """# Doc
+
+Inference connectivity configuration... supplied by the orchestrator in the \
+**PMA managed service start bundle**.
+"""
+        with tempfile.TemporaryDirectory(prefix="fix_one_sentence_") as tmp:
+            path = Path(tmp) / "test.md"
+            path.write_text(content, encoding="utf-8")
+            overrides = {"default": False, RULE: True}
+            proc = _run_markdownlint(path, fix=False, config_overrides=overrides)
+            msg = f"no {RULE} errors expected for ellipsis in sentence: {proc.stderr}"
+            self.assertEqual(proc.returncode, 0, msg)
+
     def test_no_split_on_identifiers_with_periods(self) -> None:
         """Periods in identifiers (e.g. CYNAI.PROJCT) with no space after do not trigger split."""
         content = """# Doc

test/markdownlint-rules/one-sentence-per-line.test.js

@@ -309,6 +309,12 @@ describe("one-sentence-per-line", () => {
       assert.strictEqual(errors.length, 0);
     });
 
+    it("does not split on ellipsis within a single sentence", () => {
+      const lines = ["Inference connectivity configuration... supplied by the orchestrator in the **PMA managed service start bundle**."];
+      const errors = runRule(rule, lines);
+      assert.strictEqual(errors.length, 0, "ellipsis in middle of sentence should not trigger");
+    });
+
     it("splits when ellipsis is not sentence end and real sentence end follows", () => {
       const lines = ["First... Then the next sentence."];
       const errors = runRule(rule, lines);
@@ -339,6 +345,12 @@ describe("one-sentence-per-line", () => {
       assert.strictEqual(errors.length, 0);
     });
 
+    it("does not split when period is followed only by trailing space (no next word)", () => {
+      const lines = ["First. "];
+      const errors = runRule(rule, lines);
+      assert.strictEqual(errors.length, 0, "trailing space after period with no second sentence");
+    });
+
     it("version number 1. not treated as sentence end", () => {
       const lines = ["Use version 1. It is stable."];
       const errors = runRule(rule, lines);