From 130dc1d5012c453c10f039e3f4e159d5ab7e5fad Mon Sep 17 00:00:00 2001
From: Nicholas Ward <nicholaspward@outlook.com>
Date: Mon, 17 Aug 2026 18:10:27 -0500
Subject: [PATCH] fix(ingest): resolve Docling stacked fraction OCR artifacts
 into clean measurements

---
 scripts/ingest-docling-book.mjs | 30 ++++++++++++++++++++----------
 1 file changed, 20 insertions(+), 10 deletions(-)

diff --git a/scripts/ingest-docling-book.mjs b/scripts/ingest-docling-book.mjs
index 801f781..6de7440 100644
--- a/scripts/ingest-docling-book.mjs
+++ b/scripts/ingest-docling-book.mjs
@@ -222,13 +222,23 @@ function parseMetricAmount(str) {
   return null;
 }
 
+export function cleanFractionText(str) {
+  if (!str) return "";
+  return str
+    .replace(/[\r\n]+/g, " ")
+    .replace(/(\d+)\s*\/\s*\1\s*\/\s*(\d+)/g, (m, a, b) => `${a}/${b}`)
+    .replace(/(\d+)\s*\/\s*(\d+)/g, (m, a, b) => `${a}/${b}`)
+    .replace(/\s+/g, " ")
+    .trim();
+}
+
 function parseUsFallback(text, name) {
   if (!text) {
     if (/zest/i.test(name)) return { quantity: 1, unit_id: "each", notes: "Zest of 1" };
     if (/vanilla bean/i.test(name)) return { quantity: 1, unit_id: "each", notes: "1 bean" };
     return { quantity: 1, unit_id: "gram", notes: "To taste / as needed" };
   }
-  const s = text.replace(/[\r\n\s]+/g, " ").trim().toLowerCase();
+  const s = cleanFractionText(text).toLowerCase();
   if (s.includes("1/8") || s.includes("⅛")) return { quantity: 0.6, unit_id: "gram", notes: "⅛ tsp" };
   if (s.includes("1/4") || s.includes("¼")) return { quantity: 1.25, unit_id: "gram", notes: "¼ tsp" };
   if (s.includes("1/2") || s.includes("½")) return { quantity: 2.5, unit_id: "gram", notes: "½ tsp" };
@@ -244,23 +254,23 @@ function parseUsFallback(text, name) {
 
   const ozMatch = s.match(/^([\d.]+)\s*oz$/);
   if (ozMatch && parseFloat(ozMatch[1]) > 0) {
-    return { quantity: Math.round(parseFloat(ozMatch[1]) * 28.3495 * 100) / 100, unit_id: "gram", notes: text };
+    return { quantity: Math.round(parseFloat(ozMatch[1]) * 28.3495 * 100) / 100, unit_id: "gram", notes: cleanFractionText(text) };
   }
 
-  return { quantity: 1, unit_id: "gram", notes: text };
+  return { quantity: 1, unit_id: "gram", notes: cleanFractionText(text) };
 }
 
 function cleanIngredientName(raw) {
-  let name = raw.trim().replace(/^[\s•\-\*]+/, "");
+  let cleaned = cleanFractionText(raw).replace(/^[\s•\-\*]+/, "");
   let notes = null;
 
-  const parenMatch = name.match(/^([^(]+)\s*\(([^)]+)\)$/);
+  const parenMatch = cleaned.match(/^([^(]+)\s*\(([^)]+)\)$/);
   if (parenMatch) {
-    name = parenMatch[1].trim();
+    cleaned = parenMatch[1].trim();
     notes = parenMatch[2].trim();
   }
 
-  return { name, notes };
+  return { name: cleaned, notes };
 }
 
 function inferIngredientId(name) {
@@ -397,9 +407,9 @@ export function parseDoclingBook(doclingJson, targetPages = null) {
         const row = grid.get(r);
         if (!row) continue;
 
-        const ingText = row.get(0) || "";
-        const metricText = row.get(1) || "";
-        const usText = row.get(2) || "";
+        const ingText = cleanFractionText(row.get(0) || "");
+        const metricText = cleanFractionText(row.get(1) || "");
+        const usText = cleanFractionText(row.get(2) || "");
 
         if (/total weight/i.test(ingText)) {
           const parsedTotal = parseMetricAmount(metricText) || parseMetricAmount(usText);