paperlint 2.1.0 → 3.0.0

package/eslint-rules/review-findings-cause.mjs

@@ -1,135 +0,0 @@
-/**
- * `review/findings-cause` — a review report with findings and no analysis of the ROOT CAUSE.
- *
- * ── WHAT THIS RULE IS ──────────────────────────────────────────────────────────
- * Pipeline convention: a paper review does not end with a list of findings, it ends
- * with an answer to "what in the PIPELINE let this through" — a defective skill, a
- * missing skill, a hook, a rule. Otherwise the paper text gets fixed while the tool
- * that missed it stays the same, and the next paper arrives with the same defect.
- *
- * ── WHY THIS IS A MOVE, NOT A NEW RULE ──────────────────────────────────────────
- * First unit of step 9 of the extraction: until 2026-09-15 the check lived in the
- * consumer as `checkReviewFindingsCause` in `.claude/hooks/paper-lint.mjs` — a
- * directory walk plus three regexes over the text.
- *
- * 🔴 AND THE MOVE HERE IS NOT A COPY, IT REMOVES A WHOLE CLASS OF BUGS. The old
- * version counted findings like this:
- *
- *     tableRows  ←  ^\|\s*\d+\s*\|            a line starting with "| <number> |"
- *     boldItems  ←  ^\s*[-*]\s+\*\*             a list item starting bold
- *
- * (the regexes are given WITHOUT trailing slashes and flags on purpose: an
- * "asterisk-slash" sequence inside a block comment closes it — I tripped on this
- * twice in one hour, the second time right here.)
- *
- * That is, a "table row" was recognized by how it's written, not by its markup: the
- * same line inside a ``` fence counted as a finding, a leading space in the cell
- * broke the count, and `*` and `-` had to be listed by hand. In the AST a table row
- * is a `tableRow` node, and all three misses become inexpressible. This is exactly
- * the base rule "parse markdown with a PARSER".
- *
- * ── WHAT STAYS AS CONSUMER DATA ───────────────────────────────────────────────
- * The findings threshold is an option. The "rule from a date" (reports older than a
- * given date are known debt, not a finding) DOES NOT MOVE HERE AT ALL: it's a fact
- * about one consumer's corpus, and its home is that consumer's config via `ignores`,
- * not the mechanism. The mechanism goes in the package, the data stays with the consumer.
- */
-
-/** Numbered cell: the first column of a table row holding a single number. */
-const isNumbered = (row) => {
-  const first = row.children?.[0];
-  const text = (first?.children ?? [])
-    .map((c) => c.value ?? "")
-    .join("")
-    .trim();
-  return /^\d+$/.test(text);
-};
-
-/** Finding item: a list element starting with bold text — the finding's heading. */
-const isBoldItem = (item) => {
-  const para = item.children?.find((c) => c.type === "paragraph");
-  return para?.children?.[0]?.type === "strong";
-};
-
-export default {
-  rules: {
-    "findings-cause": {
-      meta: {
-        type: "suggestion",
-        docs: {
-          description:
-            "a review report with findings names what in the pipeline let them through",
-        },
-        schema: [
-          {
-            type: "object",
-            properties: {
-              minFindings: { type: "integer", minimum: 1 },
-              causeMarker: { type: "string" },
-              sinceCreated: {
-                type: "string",
-                pattern: "^\\d{4}-\\d{2}-\\d{2}$",
-              },
-            },
-            additionalProperties: false,
-          },
-        ],
-        messages: {
-          noCause:
-            "{{count}} findings and not one «{{marker}}» note. First name what in the PIPELINE let them through (a defective skill · a missing skill · a hook · a rule) and fix THAT: the text edit falls out of running the fixed tool, not instead of it.",
-        },
-      },
-      create(context) {
-        const {
-          minFindings = 3,
-          causeMarker = "Cause:",
-          sinceCreated,
-        } = context.options[0] ?? {};
-        let findings = 0;
-        let hasCause = false;
-        let created = "";
-
-        return {
-          // 🔴 "RULE FROM A DATE" — MECHANISM, DATE — DATA. A new check that opens with a
-          // wall of findings on a historical corpus gets muted the same day; so the
-          // consumer needs a way to say "before such-and-such date this is known debt,
-          // not a finding". Measured on the first consumer: 84 reports, 0 new would fire
-          // and 49 old — without this option the rule would have opened with forty-nine
-          // findings. The date is NOT hardcoded: it comes in as an option, read from
-          // `created` in the frontmatter, and a document with no date is treated as OLD
-          // only when the option is set — otherwise a missing frontmatter would become a
-          // way to dodge the rule.
-          yaml(node) {
-            created =
-              /^created:\s*(\d{4}-\d{2}-\d{2})/m.exec(node.value ?? "")?.[1] ??
-              "";
-          },
-          tableRow(node) {
-            if (isNumbered(node)) findings++;
-          },
-          listItem(node) {
-            if (isBoldItem(node)) findings++;
-          },
-          text(node) {
-            if (node.value.includes(causeMarker)) hasCause = true;
-          },
-          "root:exit"(node) {
-            if (hasCause || findings < minFindings) return;
-            if (sinceCreated && (!created || created < sinceCreated)) return;
-            // A finding about the FILE, not a line: what's missing isn't anywhere. So the
-            // position is the start of the document, the only honest place for "this is missing".
-            context.report({
-              loc: {
-                start: { line: 1, column: 0 },
-                end: { line: 1, column: 1 },
-              },
-              messageId: "noCause",
-              data: { count: String(findings), marker: causeMarker },
-              node,
-            });
-          },
-        };
-      },
-    },
-  },
-};

package/eslint-rules/review-findings-cause.mutations.mjs

@@ -1,72 +0,0 @@
-/**
- * Battery for `review/findings-cause`. Three mutations, three different asserts: the rule
- * must FIRE, must STAY SILENT when there's a cause analysis, and the threshold must come
- * in as an option.
- */
-import { fileURLToPath } from "node:url";
-import { dirname, join, resolve } from "node:path";
-import { runMutations } from "../lib/mutation-driver.mjs";
-
-const HERE = dirname(fileURLToPath(import.meta.url));
-const ROOT = resolve(HERE, "..");
-const RULE = join(HERE, "review-findings-cause.mjs");
-const HARNESS = join(HERE, "review-findings-cause.harness.mjs");
-
-process.exit(
-  runMutations({
-    root: ROOT,
-    runner: "node",
-    cases: [
-      {
-        name: "the rule stops reporting",
-        harness: HARNESS,
-        expect: "one finding was expected",
-        disables:
-          "the verdict itself — a report without a cause analysis passes silently",
-        edits: [
-          [
-            RULE,
-            "if (hasCause || findings < minFindings) return;",
-            "if (true) return;",
-          ],
-        ],
-      },
-      {
-        name: 'the "Cause:" marker stops being noticed',
-        harness: HARNESS,
-        expect: "the rule must stay silent",
-        disables:
-          "the exemption — a report WITH a cause analysis starts going red",
-        edits: [
-          [
-            RULE,
-            "if (node.value.includes(causeMarker)) hasCause = true;",
-            "void node;",
-          ],
-        ],
-      },
-      {
-        name: '"rule from a date" stops exempting',
-        harness: HARNESS,
-        expect: "debt, not a finding",
-        disables:
-          "exemption of the historical corpus — the rule opens with a wall of findings",
-        edits: [
-          [
-            RULE,
-            "if (sinceCreated && (!created || created < sinceCreated)) return;",
-            "if (false) return;",
-          ],
-        ],
-      },
-      {
-        name: "the threshold stops being an option",
-        harness: HARNESS,
-        expect: "the threshold is data",
-        disables:
-          "passing the threshold in from outside — the mechanism claims the data for itself",
-        edits: [[RULE, "context.options[0] ?? {}", "{}"]],
-      },
-    ],
-  }),
-);

package/fixtures/paper-stages/authors-ran/PIPELINE-STATUS.md

@@ -1,16 +0,0 @@
----
-stages:
-  - stage: submitted
-    date: 2026-07-22
-    venue: A Venue 2026
-    pdf: versions/2026-07-22-submitted.pdf
-    bytes: 100
----
-# PIPELINE-STATUS
-
-<!-- Evidence of a run lives in the PROSE of a note cell — it has no node of its own, so the
-     rule looks for the marker across the whole scorecard. The fixture repeats that shape on purpose. -->
-
-| id | what | with | ok | when | note |
-|---|---|---|---|---|---|
-| cites | any \cite edit | verify-citations | ☑ | 2026-08-24 | `bib-authors` ran: 27 entries, 1 discrepancy, fixed |

package/fixtures/paper-stages/marker-in-prose/PIPELINE-STATUS.md

@@ -1,17 +0,0 @@
----
-stages:
-  - stage: submitted
-    date: 2026-07-22
-    pdf: versions/a.pdf
-    bytes: 1
----
-# PIPELINE-STATUS
-
-Need to run `bib-authors` before camera-ready — never got round to it.
-
-Exactly the case GREP cannot tell apart from a record of a run: the marker is in the file,
-but the scorecard is empty. An intention read as a fact.
-
-| id | what | ok | note |
-|---|---|---|---|
-| build | pdf builds | ☑ | nothing to note |

package/fixtures/review-findings-cause/clean.md

@@ -1,17 +0,0 @@
----
-title: "Paper review — round 3"
-created: 2026-09-01
----
-
-# Findings
-
-| # | what | where |
-|---|---|---|
-| 1 | the number in the abstract does not match table 2 | §1 |
-| 2 | a reference to a withdrawn work | §4 |
-| 3 | a past-tense promise about something unpublished | §6 |
-
-## Analysis
-
-Cause: the numbers gate read only `paper.md` and never opened `.tex`, so the table
-and the abstract were compared in different files. Fix the gate, not the paragraph.

package/fixtures/review-findings-cause/defect.md

@@ -1,14 +0,0 @@
----
-title: "Paper review — round 3"
-created: 2026-09-01
----
-
-# Findings
-
-| # | what | where |
-|---|---|---|
-| 1 | the number in the abstract does not match table 2 | §1 |
-| 2 | a reference to a withdrawn work | §4 |
-| 3 | a past-tense promise about something unpublished | §6 |
-
-Text edits follow.

package/fixtures/review-findings-cause/old-debt.md

@@ -1,14 +0,0 @@
----
-title: "Paper review — an old report"
-created: 2026-01-01
----
-
-# Findings
-
-| # | what | where |
-|---|---|---|
-| 1 | the number does not match | §1 |
-| 2 | a reference to a withdrawn work | §4 |
-| 3 | a promise about something unpublished | §6 |
-
-There is no analysis of causes here — the report predates the rule.

package/fixtures/review-findings-cause/quiet-in-fence.md

@@ -1,16 +0,0 @@
----
-title: "A note on the report format"
-created: 2026-09-01
----
-
-An example of WHAT a findings table looks like — not itself a finding:
-
-```markdown
-| # | what | where |
-|---|---|---|
-| 1 | example | §1 |
-| 2 | example | §2 |
-| 3 | example | §3 |
-```
-
-There must be no analysis of causes here: this describes the format, it is not a report.