kordoc 2.2.6 → 2.4.0

package/dist/{parser-43IAQ5KE.js → parser-XRUZEFZT.js}

@@ -6,7 +6,7 @@ import {
   blocksToMarkdown,
   safeMax,
   safeMin
-} from "./chunk-NL5XLN5R.js";
+} from "./chunk-XG5CQUSC.js";
 import {
   parsePageRange
 } from "./chunk-SBVRCJFH.js";
@@ -2186,10 +2186,11 @@ function detectSpecialKoreanTables(blocks) {
   return result;
 }
 function removeHeaderFooterBlocks(blocks, pageHeights, warnings) {
-  const ZONE_RATIO = 0.1;
+  const ZONE_RATIO = 0.12;
   const MIN_REPEAT = 3;
-  const headerTexts = /* @__PURE__ */ new Map();
-  const footerTexts = /* @__PURE__ */ new Map();
+  const Y_BUCKET = 5;
+  const topEntries = [];
+  const bottomEntries = [];
   for (let bi = 0; bi < blocks.length; bi++) {
     const b = blocks[bi];
     if (!b.bbox || !b.pageNumber || !b.text?.trim()) continue;
@@ -2197,49 +2198,51 @@ function removeHeaderFooterBlocks(blocks, pageHeights, warnings) {
     if (!ph) continue;
     const blockTop = ph - (b.bbox.y + b.bbox.height);
     const blockBottom = ph - b.bbox.y;
-    if (blockBottom <= ph * ZONE_RATIO) {
-      const arr = footerTexts.get(b.pageNumber) || [];
-      arr.push(b.text.trim());
-      footerTexts.set(b.pageNumber, arr);
-    } else if (blockTop >= ph * (1 - ZONE_RATIO)) {
-      const arr = headerTexts.get(b.pageNumber) || [];
-      arr.push(b.text.trim());
-      headerTexts.set(b.pageNumber, arr);
-    }
-  }
-  const repeatedPatterns = /* @__PURE__ */ new Set();
-  for (const textsMap of [headerTexts, footerTexts]) {
+    const entry = { blockIdx: bi, page: b.pageNumber, y: b.bbox.y, text: b.text.trim() };
+    if (blockBottom <= ph * ZONE_RATIO) bottomEntries.push(entry);
+    else if (blockTop >= ph * (1 - ZONE_RATIO)) topEntries.push(entry);
+  }
+  const removeSet = /* @__PURE__ */ new Set();
+  for (const entries of [topEntries, bottomEntries]) {
+    if (entries.length === 0) continue;
     const patternCount = /* @__PURE__ */ new Map();
-    for (const [, texts] of textsMap) {
-      for (const t of texts) {
-        const normalized = t.replace(/\d+/g, "#");
-        patternCount.set(normalized, (patternCount.get(normalized) || 0) + 1);
+    const patternPages = /* @__PURE__ */ new Map();
+    for (const e of entries) {
+      const norm = e.text.replace(/\d+/g, "#");
+      patternCount.set(norm, (patternCount.get(norm) || 0) + 1);
+      const pages = patternPages.get(norm) || /* @__PURE__ */ new Set();
+      pages.add(e.page);
+      patternPages.set(norm, pages);
+    }
+    const repeatedPatterns = /* @__PURE__ */ new Set();
+    for (const [p, count] of patternCount) {
+      if (count >= MIN_REPEAT && (patternPages.get(p)?.size ?? 0) >= MIN_REPEAT) {
+        repeatedPatterns.add(p);
       }
     }
-    for (const [pattern, count] of patternCount) {
-      if (count >= MIN_REPEAT) repeatedPatterns.add(pattern);
-    }
-  }
-  if (repeatedPatterns.size === 0) return [];
-  const removeIndices = [];
-  for (let bi = 0; bi < blocks.length; bi++) {
-    const b = blocks[bi];
-    if (!b.bbox || !b.pageNumber || !b.text?.trim()) continue;
-    const ph = pageHeights.get(b.bbox.page) || pageHeights.get(b.pageNumber);
-    if (!ph) continue;
-    const blockTop = ph - (b.bbox.y + b.bbox.height);
-    const blockBottom = ph - b.bbox.y;
-    const inZone = blockBottom <= ph * ZONE_RATIO || blockTop >= ph * (1 - ZONE_RATIO);
-    if (!inZone) continue;
-    const normalized = b.text.trim().replace(/\d+/g, "#");
-    if (repeatedPatterns.has(normalized)) {
-      removeIndices.push(bi);
+    const bucketPages = /* @__PURE__ */ new Map();
+    for (const e of entries) {
+      const bucket = Math.round(e.y / Y_BUCKET);
+      const pages = bucketPages.get(bucket) || /* @__PURE__ */ new Set();
+      pages.add(e.page);
+      bucketPages.set(bucket, pages);
+    }
+    const repeatedBuckets = /* @__PURE__ */ new Set();
+    for (const [b, pages] of bucketPages) {
+      if (pages.size >= MIN_REPEAT) repeatedBuckets.add(b);
+    }
+    for (const e of entries) {
+      const norm = e.text.replace(/\d+/g, "#");
+      const bucket = Math.round(e.y / Y_BUCKET);
+      if (repeatedPatterns.has(norm) || repeatedBuckets.has(bucket)) {
+        removeSet.add(e.blockIdx);
+      }
     }
   }
-  if (removeIndices.length > 0) {
-    warnings.push({ message: `${removeIndices.length}\uAC1C \uBA38\uB9AC\uAE00/\uBC14\uB2E5\uAE00 \uC694\uC18C \uC81C\uAC70\uB428`, code: "HIDDEN_TEXT_FILTERED" });
+  if (removeSet.size > 0) {
+    warnings.push({ message: `${removeSet.size}\uAC1C \uBA38\uB9AC\uAE00/\uBC14\uB2E5\uAE00 \uC694\uC18C \uC81C\uAC70\uB428`, code: "HIDDEN_TEXT_FILTERED" });
   }
-  return removeIndices;
+  return [...removeSet].sort((a, b) => a - b);
 }
 function mergeKoreanLines(text) {
   if (!text) return "";
@@ -2275,4 +2278,4 @@ export {
   extractPdfMetadataOnly,
   parsePdfDocument
 };
-//# sourceMappingURL=parser-43IAQ5KE.js.map
+//# sourceMappingURL=parser-XRUZEFZT.js.map