@clazic/kordoc 2.4.3 → 2.4.5

package/dist/index.js

@@ -1961,6 +1961,13 @@ var require_cfb = __commonJS({
 });
 
 // src/ocr/auto-detect.ts
+var auto_detect_exports = {};
+__export(auto_detect_exports, {
+  detectAvailableOcr: () => detectAvailableOcr,
+  getAutoFallbackChain: () => getAutoFallbackChain,
+  getTesseractFallbackMessage: () => getTesseractFallbackMessage,
+  validateOcrMode: () => validateOcrMode
+});
 import { execSync } from "child_process";
 function detectAvailableOcr() {
   for (const cli of CLI_PRIORITY) {
@@ -1977,6 +1984,14 @@ function isCliInstalled(name) {
     return false;
   }
 }
+function getAutoFallbackChain() {
+  const chain = [];
+  for (const cli of CLI_PRIORITY) {
+    if (isCliInstalled(cli)) chain.push(cli);
+  }
+  chain.push("tesseract");
+  return chain;
+}
 function validateOcrMode(mode) {
   if (mode === "auto" || mode === "off" || mode === "tesseract") return;
   if (!isCliInstalled(mode)) {
@@ -2681,29 +2696,44 @@ async function ocrPagesBatch(doc, provider, pageFilter, effectivePageCount, warn
   let processed = 0;
   const batchTasks = pageBatches.map((batchPageNums, batchIdx) => async () => {
     const pageBlocks = [];
+    const batchImages = [];
     try {
-      const batchImages = [];
       for (const pageNum of batchPageNums) {
         const page = await doc.getPage(pageNum);
         const image = await renderPageToPng(page);
         batchImages.push({ image, pageNum });
       }
-      const results = await provider.processBatch(batchImages);
-      for (const { pageNum } of batchImages) {
-        const result = results.get(pageNum);
-        pageBlocks.push({
-          pageNum,
-          blocks: result ? ocrResultToBlocks(result, pageNum) : []
-        });
-      }
-    } catch (err) {
-      const range = `${batchPageNums[0]}-${batchPageNums[batchPageNums.length - 1]}`;
-      warnings?.push({
-        message: `\uBC30\uCE58 OCR \uC2E4\uD328 (\uD398\uC774\uC9C0 ${range}): ${err instanceof Error ? err.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`,
-        code: "OCR_PAGE_FAILED"
-      });
+    } catch (renderErr) {
+      const rendered = new Set(batchImages.map((b) => b.pageNum));
       for (const pageNum of batchPageNums) {
-        pageBlocks.push({ pageNum, blocks: [] });
+        if (!rendered.has(pageNum)) pageBlocks.push({ pageNum, blocks: [] });
+      }
+    }
+    if (batchImages.length > 0) {
+      try {
+        const results = await provider.processBatch(batchImages);
+        for (const { pageNum } of batchImages) {
+          const result = results.get(pageNum);
+          pageBlocks.push({
+            pageNum,
+            blocks: result ? ocrResultToBlocks(result, pageNum) : []
+          });
+        }
+      } catch (err) {
+        const range = `${batchPageNums[0]}-${batchPageNums[batchPageNums.length - 1]}`;
+        warnings?.push({
+          message: `\uBC30\uCE58 OCR \uC2E4\uD328 (\uD398\uC774\uC9C0 ${range}): ${err instanceof Error ? err.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"} \u2014 \uB2E8\uC77C \uD398\uC774\uC9C0\uB85C \uC7AC\uC2DC\uB3C4`,
+          code: "OCR_PAGE_FAILED"
+        });
+        for (const { image, pageNum } of batchImages) {
+          try {
+            const singleResult = await provider.processBatch([{ image, pageNum }]);
+            const r = singleResult.get(pageNum);
+            pageBlocks.push({ pageNum, blocks: r ? ocrResultToBlocks(r, pageNum) : [] });
+          } catch {
+            pageBlocks.push({ pageNum, blocks: [] });
+          }
+        }
       }
     }
     processed += batchPageNums.length;
@@ -2790,7 +2820,7 @@ import JSZip2 from "jszip";
 import { DOMParser } from "@xmldom/xmldom";
 
 // src/utils.ts
-var VERSION = true ? "2.4.3" : "0.0.0-dev";
+var VERSION = true ? "2.4.5" : "0.0.0-dev";
 function toArrayBuffer(buf) {
   if (buf.byteOffset === 0 && buf.byteLength === buf.buffer.byteLength) {
     return buf.buffer;
@@ -6342,53 +6372,86 @@ async function parsePdfDocument(buffer, options) {
     }
     const parsedPageCount = parsedPages || (pageFilter ? pageFilter.size : effectivePageCount);
     if (isImageBased) {
-      let ocrProvider = options?.ocr ?? null;
       const ocrMode = options?.ocrMode ?? "auto";
-      if (!ocrProvider && ocrMode !== "off") {
-        try {
-          const { resolveOcrProvider: resolveOcrProvider2 } = await Promise.resolve().then(() => (init_resolve(), resolve_exports));
-          const concurrency = options?.ocrConcurrency ?? 1;
-          const batchSize = options?.ocrBatchSize;
-          ocrProvider = await resolveOcrProvider2(ocrMode, warnings, concurrency, batchSize);
-        } catch (resolveErr) {
-          if (ocrMode !== "auto") {
-            throw Object.assign(
-              new KordocError(resolveErr instanceof Error ? resolveErr.message : "OCR \uD504\uB85C\uBC14\uC774\uB354 \uCD08\uAE30\uD654 \uC2E4\uD328"),
-              { isImageBased: true }
-            );
-          }
-        }
+      const concurrency = options?.ocrConcurrency ?? 1;
+      const batchSize = options?.ocrBatchSize;
+      if (ocrMode === "off") {
+        throw Object.assign(new KordocError(`\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`), { isImageBased: true });
       }
-      if (ocrProvider) {
-        let ocrBlocks = [];
+      const { resolveOcrProvider: resolveOcrProvider2 } = await Promise.resolve().then(() => (init_resolve(), resolve_exports));
+      const { ocrPages: ocrPages2 } = await Promise.resolve().then(() => (init_provider(), provider_exports));
+      const tryProvider = async (provider, filter) => {
         try {
-          const { ocrPages: ocrPages2 } = await Promise.resolve().then(() => (init_provider(), provider_exports));
-          const concurrency = options?.ocrConcurrency ?? 1;
-          ocrBlocks = await ocrPages2(doc, ocrProvider, pageFilter, effectivePageCount, warnings, concurrency, options?.onProgress);
+          return await ocrPages2(doc, provider, filter, effectivePageCount, warnings, concurrency, options?.onProgress);
         } catch {
+          return [];
         } finally {
-          const terminable = ocrProvider;
+          const terminable = provider;
           if (typeof terminable.terminate === "function") {
             await terminable.terminate().catch(() => {
             });
           }
         }
-        if (ocrBlocks.length > 0) {
-          const ocrMarkdown = blocksToMarkdown(ocrBlocks);
-          return {
-            markdown: ocrMarkdown,
-            blocks: ocrBlocks,
-            metadata,
-            warnings: warnings.length > 0 ? warnings : void 0,
-            isImageBased: true
-          };
+      };
+      let ocrBlocks = [];
+      if (options?.ocr) {
+        ocrBlocks = await tryProvider(options.ocr, pageFilter);
+      } else if (ocrMode === "auto") {
+        const { getAutoFallbackChain: getAutoFallbackChain2 } = await Promise.resolve().then(() => (init_auto_detect(), auto_detect_exports));
+        const pendingPages = /* @__PURE__ */ new Set();
+        for (let i = 1; i <= effectivePageCount; i++) {
+          if (!pageFilter || pageFilter.has(i)) pendingPages.add(i);
+        }
+        const allOcrBlocks = [];
+        for (const mode of getAutoFallbackChain2()) {
+          if (pendingPages.size === 0) break;
+          try {
+            const modeFilter = pendingPages.size < effectivePageCount ? new Set(pendingPages) : pageFilter;
+            const provider = await resolveOcrProvider2(mode, warnings, concurrency, batchSize);
+            const blocks2 = await tryProvider(provider, modeFilter);
+            if (blocks2.length > 0) {
+              for (const b of blocks2) {
+                if (b.pageNumber !== void 0) pendingPages.delete(b.pageNumber);
+              }
+              for (const b of blocks2) allOcrBlocks.push(b);
+              if (pendingPages.size > 0) {
+                warnings.push({
+                  message: `OCR: '${mode}' \uC644\uB8CC (${pendingPages.size}\uD398\uC774\uC9C0 \uBBF8\uCC98\uB9AC \u2192 \uB2E4\uC74C \uC5D4\uC9C4\uC73C\uB85C \uC7AC\uC2DC\uB3C4)`,
+                  code: "OCR_CLI_FALLBACK"
+                });
+              }
+            } else {
+              warnings.push({ message: `OCR: '${mode}' \uACB0\uACFC \uC5C6\uC74C, \uB2E4\uC74C \uC5D4\uC9C4\uC73C\uB85C \uC2DC\uB3C4`, code: "OCR_CLI_FALLBACK" });
+            }
+          } catch {
+          }
+        }
+        ocrBlocks = allOcrBlocks;
+      } else {
+        try {
+          const provider = await resolveOcrProvider2(ocrMode, warnings, concurrency, batchSize);
+          ocrBlocks = await tryProvider(provider, pageFilter);
+        } catch (resolveErr) {
+          throw Object.assign(
+            new KordocError(resolveErr instanceof Error ? resolveErr.message : "OCR \uD504\uB85C\uBC14\uC774\uB354 \uCD08\uAE30\uD654 \uC2E4\uD328"),
+            { isImageBased: true }
+          );
         }
       }
-      if (ocrMode === "off") {
-        throw Object.assign(new KordocError(`\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`), { isImageBased: true });
+      if (ocrBlocks.length > 0) {
+        const ocrMarkdown = blocksToMarkdown(ocrBlocks);
+        return {
+          markdown: ocrMarkdown,
+          blocks: ocrBlocks,
+          metadata,
+          warnings: warnings.length > 0 ? warnings : void 0,
+          isImageBased: true
+        };
       }
-      const errMsg = ocrMode ? `\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF \u2014 OCR \uC2E4\uD328 (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)` : `\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`;
-      throw Object.assign(new KordocError(errMsg), { isImageBased: true });
+      throw Object.assign(
+        new KordocError(`\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF \u2014 OCR \uC2E4\uD328 (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`),
+        { isImageBased: true }
+      );
     }
     if (options?.removeHeaderFooter !== false && parsedPageCount >= 3) {
       const removed = removeHeaderFooterBlocks(blocks, pageHeights, warnings);