fix: preserve deterministic source unit structure

AGENTS.md

@@ -150,7 +150,7 @@ contextual-orchestrator owns model discovery and selection.
   paragraphs, not the authoring application's spacing workaround.
 - Quantity HTML `<sup>`/`<sub>` and caret exponents such as `m^3` become
   Unicode in derived units and React `<sup>`/`<sub>` in the post view
-  (ADR 0119). Never assign the body to `innerHTML`. Do not treat
+  (ADR 0165). Never assign the body to `innerHTML`. Do not treat
   `qty < 50` or a leading footnote `^1` as an exponent.
 - Image descriptions, OCR text, and region evidence are analysis artifacts,
   not buyer-facing prompt instructions. Buyer UI shows the source content and

CHANGELOG.d/2.12.9-quantity-superscripts.md

@@ -5,4 +5,4 @@
 - Post popups now show cubic metres and similar quantities as superscripts
   and subscripts (`12 m³`, `H₂O`) instead of flattened `m^3` or `m3`.
   Semantic units store Unicode so embeddings keep the exponent. Comparison
-  operators and leading footnote carets stay literal (ADR 0119).
+  operators and leading footnote carets stay literal (ADR 0165).

docs/adr/0119-quantity-script-display.md → docs/adr/0165-quantity-script-display.md

@@ -1,4 +1,4 @@
-# ADR 0119: Render quantity superscripts as text runs, Unicode in units
+# ADR 0165: Render quantity superscripts as text runs, Unicode in units
 
 - Status: Accepted
 - Date: 2026-08-22

docs/adr/README.md

@@ -10,7 +10,7 @@ decision from them.
 | Supporting document | Normative ADR |
 |---|---|
 | [`product-technical-gap-baseline.md`](../product-technical-gap-baseline.md) | Product/technical traceability projection across the ADR set; ADRs remain normative |
-| [`lineage-bi-research-notes.md`](../lineage-bi-research-notes.md) | [0084](0084-lineage-research-grounding.md), [0062](0062-semantic-unit-embedding.md), [0064](0064-lineage-evidence-and-tree-assembly.md), [0119](0119-quantity-script-display.md) |
+| [`lineage-bi-research-notes.md`](../lineage-bi-research-notes.md) | [0084](0084-lineage-research-grounding.md), [0062](0062-semantic-unit-embedding.md), [0064](0064-lineage-evidence-and-tree-assembly.md), [0165](0165-quantity-script-display.md) |
 | [`PROV_O_IMPLEMENTATION.md`](../PROV_O_IMPLEMENTATION.md) | [0065](0065-prov-o-provenance-boundary.md) |
 | [`PROV_O_IMPLEMENTATION_MATRIX.md`](../PROV_O_IMPLEMENTATION_MATRIX.md) | [0065](0065-prov-o-provenance-boundary.md) |
 | [`image-content-schema.md`](../image-content-schema.md) | [0066](0066-position-preserving-image-content.md) |

docs/lineage-bi-research-notes.md

@@ -386,7 +386,7 @@ hard truncation that silently drops earlier decisions. This is recorded
 here as the citation this feature would build on, not as a claim that
 conversation-level compression is implemented today.
 
-## Quantity scripts in source units (ADR 0119)
+## Quantity scripts in source units (ADR 0165)
 
 Board exports write cubic metres as HTML `<sup>` or as `m^3`. Flattening
 those tags concatenates `m3`, which is a different quantity, and leaving

docs/product-technical-gap-baseline.md

@@ -5,7 +5,7 @@
 - **Table Parsing**: `post=00505695-3e61-1fd1-80c6-86bb61c8ddc5` completely fails at parsing tables.
 - **Indentation**: Incorrect indentation rendering in `post=00505695-7571-1fd1-83c3-d521b187ad5b` and `post=00505695-3e61-1fd1-83c0-497b3c1c455e`.
 - **Image/Table OCR**: `post=00505695-7571-1fd1-83dd-3d22a61a5734` fails text recognition for tables inside images, markdown parsing fails, and image OCR description is too shallow for Ontology & Semantics.
-- **Math/Superscripts**: (Display/unit text, ADR 0119) Quantity HTML
+- **Math/Superscripts**: (Display/unit text, ADR 0165) Quantity HTML
   `<sup>`/`<sub>` and caret exponents such as `m^3` now render as
   superscripts and persist as Unicode in semantic units. Full formula
   ontology grammar remains open.

frontend/src/postBodyDisplay.test.ts

@@ -47,17 +47,21 @@ describe("splitPostBody", () => {
     ]);
   });
 
-  it("reads CSS box shorthand indentation and markerless footnotes", () => {
+  it("reads CSS box shorthand indentation", () => {
     expect(
       splitPostBody(
         '<ul><li style="margin: 0cm 0cm 0cm 56px">Outer</li></ul>' +
-          '<ul><li style="margin: 0cm 0cm 0cm 80px">Nested</li></ul>' +
-          "<p>*Tier 2: note</p>",
+          '<ul><li style="margin: 0cm 0cm 0cm 80px">Nested</li></ul>',
       ),
     ).toEqual([
       { kind: "text", text: "Outer", indentLevel: 7 },
       { kind: "text", text: "Nested", indentLevel: 10 },
-      { kind: "text", text: "*Tier 2: note", role: "footnote" },
+    ]);
+  });
+
+  it("does not infer a footnote from a bare marker", () => {
+    expect(splitPostBody("<p>*Synthetic list item</p>")).toEqual([
+      { kind: "text", text: "*Synthetic list item" },
     ]);
   });
 
@@ -156,6 +160,55 @@ describe("splitPostBody", () => {
     ]);
   });
 
+  it("normalizes entity-encoded quantity syntax without leaking raw markup", () => {
+    expect(
+      splitPostBody("<p>Reserve 12 m&#94;3 and x&lt;sup&gt;2&lt;/sup&gt; units.</p>"),
+    ).toEqual([{ kind: "text", text: "Reserve 12 m³ and x² units." }]);
+  });
+
+  it("keeps encoded non-script inline markup literal", () => {
+    expect(splitPostBody("<p>Keep &lt;b&gt;bold&lt;/b&gt; literal.</p>")).toEqual([
+      { kind: "text", text: "Keep <b>bold</b> literal." },
+    ]);
+  });
+
+  it("keeps encoded non-script block markup literal", () => {
+    expect(
+      splitPostBody(
+        "<p>Keep &lt;table&gt;&lt;tr&gt;&lt;td&gt;grid&lt;/td&gt;&lt;/tr&gt;&lt;/table&gt; literal.</p>",
+      ),
+    ).toEqual([
+      { kind: "text", text: "Keep <table><tr><td>grid</td></tr></table> literal." },
+    ]);
+  });
+
+  it("normalizes nested-encoded script tags and their inner entity", () => {
+    expect(
+      splitPostBody(
+        "<p>Volume is m&amp;lt;sup&amp;gt;&amp;nbsp;3&amp;lt;/sup&amp;gt;.</p>",
+      ),
+    ).toEqual([{ kind: "text", text: "Volume is m ³." }]);
+  });
+
+  it("normalizes encoded script content wrapped in encoded inline markup", () => {
+    expect(
+      splitPostBody("<p>x&lt;sup&gt;&lt;span&gt;2&lt;/span&gt;&lt;/sup&gt;</p>"),
+    ).toEqual([{ kind: "text", text: "x²" }]);
+  });
+
+  it("keeps encoded script-prefixed custom and namespaced tags literal", () => {
+    expect(
+      splitPostBody(
+        "<p>Keep &lt;sup-note&gt;2&lt;/sup-note&gt; and &lt;sub:item&gt;3&lt;/sub:item&gt; literal.</p>",
+      ),
+    ).toEqual([
+      {
+        kind: "text",
+        text: "Keep <sup-note>2</sup-note> and <sub:item>3</sub:item> literal.",
+      },
+    ]);
+  });
+
   it("matches sup/sub content split across a newline", () => {
     // Pretty-printed source HTML puts tag content on its own line
     // (regression: the regex lacked the dotAll flag, so `.` could not cross
@@ -178,6 +231,10 @@ describe("splitPostBody", () => {
     ]);
   });
 
+  it("keeps mixed script content as a visible fallback", () => {
+    expect(splitPostBody("x<sup>3a</sup>")).toEqual([{ kind: "text", text: "x^3a" }]);
+  });
+
   it("decodes a stored superscript letter deterministically to lowercase", () => {
     // "n" and "N" both encode to the same Unicode "ⁿ" (there is no distinct
     // uppercase superscript N), so decoding must pick one case consistently

frontend/src/postBodyDisplay.ts

@@ -20,7 +20,6 @@ const BLOCK_TAG =
   /<\/?(?:article|blockquote|div|h[1-6]|li|ol|p|section|table|tbody|td|tfoot|th|thead|tr|ul|w:p|w:tbl|w:tr|w:tc)\b[^>]*>/gi;
 const WORD_INDENT_TAG = /<w:ind\b[^>]*\/?\s*>/gi;
 const LIST_ITEM_START = /^\s*(?:[-*•·]\s+|[*†‡](?=\S)|(?:\d{1,3}|[A-Za-z가-힣])[.)]\s+|[①-⑳]\s+)/;
-const FOOTNOTE_START = /^\s*[*†‡](?=\S)/;
 const INDENT_MARKER = "\u0001lw-indent:";
 const INDENT_MARKER_END = "\u0002";
 const INDENT_MARKER_PATTERN = /lw-indent:(\d+)/g;
@@ -156,6 +155,9 @@ const SUPER_UNI_TO_ASCII = buildUnicodeToAsciiTable(SUPER_ASCII_TO_UNI);
 const SUB_UNI_TO_ASCII = buildUnicodeToAsciiTable(SUB_ASCII_TO_UNI);
 const CARET_EXPONENT =
   /(?<=[A-Za-z0-9µμ°ΩÅåÅ)])\^(?:\{([+-]?\d{1,3}|[nNiI])\}|([+-]?\d{1,3}|[nNiI]))/g;
+const ENCODED_CARET = /&(?:amp;)*(?:#0*94|#x0*5e);/gi;
+const ENCODED_SCRIPT_TAG =
+  /&(?:amp;)*(?:lt|#0*60|#x0*3c);\s*\/?\s*(?:sup|sub)(?=\s|\/|&(?:amp;)*(?:gt|#0*62|#x0*3e);).*?&(?:amp;)*(?:gt|#0*62|#x0*3e);/gis;
 
 function applyUnicodeScript(text: string, kind: "super" | "sub"): string {
   const table = kind === "super" ? SUPER_ASCII_TO_UNI : SUB_ASCII_TO_UNI;
@@ -174,17 +176,25 @@ function applyUnicodeScript(text: string, kind: "super" | "sub"): string {
 function replaceHtmlScripts(text: string): string {
   return text
     .replace(/<sup\b[^>]*>(.*?)<\/sup>/gis, (_match, inner: string) =>
-      applyUnicodeScript(decodeHtmlEntities(String(inner).replace(/<[^>]+>/g, "")), "super"),
+      applyUnicodeScript(decodeHtmlEntities(String(inner)).replace(/<[^>]+>/g, ""), "super"),
     )
     .replace(/<sub\b[^>]*>(.*?)<\/sub>/gis, (_match, inner: string) =>
-      applyUnicodeScript(decodeHtmlEntities(String(inner).replace(/<[^>]+>/g, "")), "sub"),
+      applyUnicodeScript(decodeHtmlEntities(String(inner)).replace(/<[^>]+>/g, ""), "sub"),
     );
 }
 
+function decodeScriptEntities(text: string): string {
+  return text
+    .replace(ENCODED_SCRIPT_TAG, (tag) => decodeHtmlEntities(tag))
+    .replace(ENCODED_CARET, (caret) => decodeHtmlEntities(caret));
+}
+
 export function normalizeScriptText(text: string): string {
-  return replaceHtmlScripts(text).replace(CARET_EXPONENT, (_match, braced: string, bare: string) =>
-    applyUnicodeScript(braced || bare, "super"),
+  const withCarets = decodeScriptEntities(text).replace(
+    CARET_EXPONENT,
+    (_match, braced: string, bare: string) => applyUnicodeScript(braced || bare, "super"),
   );
+  return replaceHtmlScripts(withCarets);
 }
 
 export type ScriptRun = { text: string; script?: "super" | "sub" };
@@ -239,7 +249,7 @@ export function splitScriptRuns(text: string): ScriptRun[] {
 }
 
 function stripHtmlTags(text: string): string {
-  const withScripts = replaceHtmlScripts(text);
+  const withScripts = normalizeScriptText(text);
   const withBoundaries = withScripts
     .replace(BREAK_TAG, "\n")
     .replace(BLOCK_TAG, (tag) => {
@@ -250,8 +260,7 @@ function stripHtmlTags(text: string): string {
   const withoutTags = withBoundaries.replace(HTML_TAG, (tag) =>
     /^<\/?w:/i.test(tag) ? "" : " ",
   );
-  const decoded = decodeHtmlEntities(withoutTags);
-  return normalizeScriptText(decoded)
+  return decodeHtmlEntities(withoutTags)
     .split("\n")
     .map((line) => {
       if (!line.trim()) return "";
@@ -358,7 +367,6 @@ function pushText(segments: PostBodySegment[], raw: string, indentUnit: number):
         kind: "text",
         text: normalized,
         ...(indentLevel > 0 ? { indentLevel } : {}),
-        ...(FOOTNOTE_START.test(normalized) ? { role: "footnote" as const } : {}),
       });
     }
   }