@clazic/kordoc 2.4.2 → 2.4.4

package/dist/auto-detect-2YGFYQCN.js

@@ -0,0 +1,15 @@
+#!/usr/bin/env node
+import {
+  detectAvailableOcr,
+  getAutoFallbackChain,
+  getTesseractFallbackMessage,
+  validateOcrMode
+} from "./chunk-7NOZFYH6.js";
+import "./chunk-ZWE3DS7E.js";
+export {
+  detectAvailableOcr,
+  getAutoFallbackChain,
+  getTesseractFallbackMessage,
+  validateOcrMode
+};
+//# sourceMappingURL=auto-detect-2YGFYQCN.js.map

package/dist/{chunk-I3HO5HLQ.js → chunk-5AXJRBBK.js}

@@ -1,4 +1,7 @@
 #!/usr/bin/env node
+import {
+  markdownToBlocks
+} from "./chunk-4PP34NVQ.js";
 import {
   KordocError,
   classifyError,
@@ -6,7 +9,7 @@ import {
   precheckZipSize,
   sanitizeHref,
   toArrayBuffer
-} from "./chunk-CMZPKEJ7.js";
+} from "./chunk-KEDUF24M.js";
 import {
   parsePageRange
 } from "./chunk-MOL7MDBG.js";
@@ -16,9 +19,6 @@ import {
 import {
   createCliOcrProvider
 } from "./chunk-JOGAFNIL.js";
-import {
-  markdownToBlocks
-} from "./chunk-4PP34NVQ.js";
 import {
   __commonJS,
   __require,
@@ -5432,66 +5432,93 @@ async function parsePdfDocument(buffer, options) {
         warnings.push({ page: i, message: `\uD398\uC774\uC9C0 ${i} \uD30C\uC2F1 \uC2E4\uD328: ${pageErr instanceof Error ? pageErr.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`, code: "PARTIAL_PARSE" });
       }
     };
-    const sampleCount = Math.min(5, targetPageNums.length);
-    for (let si = 0; si < sampleCount; si++) {
+    const SAMPLE_SIZE = Math.min(10, targetPageNums.length);
+    const sampledIndices = /* @__PURE__ */ new Set();
+    if (targetPageNums.length <= SAMPLE_SIZE) {
+      for (let i = 0; i < targetPageNums.length; i++) sampledIndices.add(i);
+    } else {
+      for (let i = 0; i < SAMPLE_SIZE; i++) {
+        const idx = Math.round(i * (targetPageNums.length - 1) / (SAMPLE_SIZE - 1));
+        sampledIndices.add(idx);
+      }
+    }
+    for (const si of sampledIndices) {
       await parseSinglePage(targetPageNums[si]);
     }
-    const sampleParsed = parsedPages || sampleCount;
+    const sampleParsed = parsedPages || sampledIndices.size;
     const isImageBased = totalChars / Math.max(sampleParsed, 1) < 10;
     if (!isImageBased) {
-      for (let si = sampleCount; si < targetPageNums.length; si++) {
-        await parseSinglePage(targetPageNums[si]);
+      for (let si = 0; si < targetPageNums.length; si++) {
+        if (!sampledIndices.has(si)) {
+          await parseSinglePage(targetPageNums[si]);
+        }
       }
     }
     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 } = await import("./resolve-QA3VACUP.js");
-          const concurrency = options?.ocrConcurrency ?? 1;
-          const batchSize = options?.ocrBatchSize;
-          ocrProvider = await resolveOcrProvider(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, ${totalChars}\uC790)`), { isImageBased: true });
       }
-      if (ocrProvider) {
-        let ocrBlocks = [];
+      const { resolveOcrProvider } = await import("./resolve-TZVGVOVD.js");
+      const { ocrPages } = await import("./provider-HE727F7Z.js");
+      const tryProvider = async (provider) => {
         try {
-          const { ocrPages } = await import("./provider-HE727F7Z.js");
-          const concurrency = options?.ocrConcurrency ?? 1;
-          ocrBlocks = await ocrPages(doc, ocrProvider, pageFilter, effectivePageCount, warnings, concurrency, options?.onProgress);
+          return await ocrPages(doc, provider, pageFilter, effectivePageCount, warnings, concurrency, options?.onProgress);
         } catch {
+          return [];
         } finally {
-          const terminable = ocrProvider;
+          const terminable = provider;
           if (typeof terminable.terminate === "function") {
             await terminable.terminate().catch(() => {
             });
           }
         }
-        if (ocrBlocks.length > 0) {
-          const ocrMarkdown = blocksToMarkdown(ocrBlocks);
-          return {
-            markdown: ocrMarkdown,
-            blocks: ocrBlocks,
-            metadata,
-            warnings: warnings.length > 0 ? warnings : void 0,
-            isImageBased: true
-          };
+      };
+      let ocrBlocks = [];
+      if (options?.ocr) {
+        ocrBlocks = await tryProvider(options.ocr);
+      } else if (ocrMode === "auto") {
+        const { getAutoFallbackChain } = await import("./auto-detect-2YGFYQCN.js");
+        for (const mode of getAutoFallbackChain()) {
+          try {
+            const provider = await resolveOcrProvider(mode, warnings, concurrency, batchSize);
+            const blocks2 = await tryProvider(provider);
+            if (blocks2.length > 0) {
+              ocrBlocks = blocks2;
+              break;
+            }
+            warnings.push({ message: `OCR: '${mode}' \uACB0\uACFC \uC5C6\uC74C, \uB2E4\uC74C \uC5D4\uC9C4\uC73C\uB85C \uC2DC\uB3C4`, code: "OCR_CLI_FALLBACK" });
+          } catch {
+          }
+        }
+      } else {
+        try {
+          const provider = await resolveOcrProvider(ocrMode, warnings, concurrency, batchSize);
+          ocrBlocks = await tryProvider(provider);
+        } 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);
@@ -9768,4 +9795,4 @@ export {
 cfb/cfb.js:
   (*! crc32.js (C) 2014-present SheetJS -- http://sheetjs.com *)
 */
-//# sourceMappingURL=chunk-I3HO5HLQ.js.map
+//# sourceMappingURL=chunk-5AXJRBBK.js.map