kordoc 4.17.2 → 4.17.4

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (142) hide show
  1. package/dist/{-2UA2JHVA.js → -SQRLFE4N.js} +11 -11
  2. package/dist/{chunk-4HRJISJF.js → chunk-4D52VUGA.js} +3 -3
  3. package/dist/{chunk-6OCKAXIS.js → chunk-6DXSRQ65.js} +2 -2
  4. package/dist/{chunk-XZA6X34R.js → chunk-7TE6TK52.js} +2 -2
  5. package/dist/{chunk-KQIYC2ZU.js → chunk-AAQR4RIX.js} +2 -2
  6. package/dist/{chunk-GPC72VSF.cjs → chunk-AFXX3YSP.cjs} +262 -171
  7. package/dist/chunk-AFXX3YSP.cjs.map +1 -0
  8. package/dist/{chunk-6DOCSYEM.cjs → chunk-AOONZC4S.cjs} +3 -3
  9. package/dist/chunk-AOONZC4S.cjs.map +1 -0
  10. package/dist/{chunk-A2W2JQ3S.js → chunk-DD6AY6EI.js} +2 -2
  11. package/dist/{chunk-KZVQK5NS.cjs → chunk-EXBZHI25.cjs} +2 -2
  12. package/dist/chunk-EXBZHI25.cjs.map +1 -0
  13. package/dist/{chunk-FGL4IGOT.js → chunk-FSCVZR4A.js} +4 -4
  14. package/dist/{chunk-HAW2VL62.cjs → chunk-GNORDND5.cjs} +12 -12
  15. package/dist/chunk-GNORDND5.cjs.map +1 -0
  16. package/dist/chunk-GS7T56RP.cjs.map +1 -1
  17. package/dist/{chunk-GYTL6WT3.js → chunk-HQ6TM2JE.js} +3 -3
  18. package/dist/{chunk-WJLICIYV.js → chunk-HY33WBAH.js} +24 -21
  19. package/dist/{chunk-WJLICIYV.js.map → chunk-HY33WBAH.js.map} +1 -1
  20. package/dist/{chunk-O4CB7D2Q.js → chunk-IERSVUW6.js} +2 -2
  21. package/dist/chunk-J55UQC5H.cjs.map +1 -1
  22. package/dist/chunk-KIYHQQVO.cjs +170 -0
  23. package/dist/chunk-KIYHQQVO.cjs.map +1 -0
  24. package/dist/{chunk-SDRKHEVT.js → chunk-NNNRN6BY.js} +255 -164
  25. package/dist/chunk-NNNRN6BY.js.map +1 -0
  26. package/dist/{chunk-MDCTITUG.js → chunk-O5T5OIJM.js} +2 -2
  27. package/dist/{chunk-WL3BWROT.js → chunk-ODVIUG6Q.js} +22 -12
  28. package/dist/chunk-ODVIUG6Q.js.map +1 -0
  29. package/dist/{chunk-C4MVZXB5.js → chunk-P2ZPPRM7.js} +2 -2
  30. package/dist/chunk-PNXJYEOR.js +170 -0
  31. package/dist/chunk-PNXJYEOR.js.map +1 -0
  32. package/dist/{chunk-YGKC6E6A.js → chunk-PTMYCK7R.js} +2 -2
  33. package/dist/{chunk-4TSYMJDL.js → chunk-PVJFCDND.js} +38 -13
  34. package/dist/chunk-PVJFCDND.js.map +1 -0
  35. package/dist/{chunk-ZDX5PIWR.js → chunk-RCS7WXOM.js} +2 -2
  36. package/dist/{chunk-VU26WGE5.js → chunk-SSY6FQLS.js} +200 -97
  37. package/dist/chunk-SSY6FQLS.js.map +1 -0
  38. package/dist/{chunk-JY2NBFIG.js → chunk-TMHMJUYU.js} +302 -75
  39. package/dist/chunk-TMHMJUYU.js.map +1 -0
  40. package/dist/{chunk-5JZEY5IL.js → chunk-VJP7O6BL.js} +2 -2
  41. package/dist/{chunk-2XHXABCX.js → chunk-W3OY3SLY.js} +3 -3
  42. package/dist/chunk-W3OY3SLY.js.map +1 -0
  43. package/dist/{chunk-WWMKIHF7.cjs → chunk-WNF3TSV3.cjs} +129 -104
  44. package/dist/chunk-WNF3TSV3.cjs.map +1 -0
  45. package/dist/{chunk-3N272C5L.js → chunk-YBIWPCUX.js} +39 -14
  46. package/dist/chunk-YBIWPCUX.js.map +1 -0
  47. package/dist/chunks-VNSLMJQD.js +10 -0
  48. package/dist/cli.js +26 -26
  49. package/dist/{formula-5RETDCIC.cjs → formula-U37MTMBD.cjs} +9 -9
  50. package/dist/formula-U37MTMBD.cjs.map +1 -0
  51. package/dist/{formula-MBLN2DG4.js → formula-U3MJXXIL.js} +4 -4
  52. package/dist/{formula-QZZDW5GW.js → formula-WZH5CMA3.js} +3 -3
  53. package/dist/{image-ocr-EREKAS47.js → image-ocr-HYWDBN72.js} +6 -6
  54. package/dist/{image-ocr-PSPVRLB5.cjs → image-ocr-OXU55EW3.cjs} +15 -15
  55. package/dist/image-ocr-OXU55EW3.cjs.map +1 -0
  56. package/dist/{image-ocr-CJSO3QSR.js → image-ocr-V6MICC6R.js} +9 -8
  57. package/dist/{image-ocr-CJSO3QSR.js.map → image-ocr-V6MICC6R.js.map} +1 -1
  58. package/dist/index.cjs +1194 -864
  59. package/dist/index.cjs.map +1 -1
  60. package/dist/index.js +531 -201
  61. package/dist/index.js.map +1 -1
  62. package/dist/mcp.js +22 -22
  63. package/dist/{model-bundle-MA4MMCRJ.js → model-bundle-AHBUQ6SD.js} +5 -5
  64. package/dist/{models-SEJ7LNSN.js → models-UF7B5LLG.js} +5 -5
  65. package/dist/page-range-QHHFTL7X.cjs.map +1 -1
  66. package/dist/{parser-M5WHRX25.js → parser-F6N6ID25.js} +22 -13
  67. package/dist/parser-F6N6ID25.js.map +1 -0
  68. package/dist/{parser-QILYPWSI.cjs → parser-IXRGUYZW.cjs} +207 -198
  69. package/dist/parser-IXRGUYZW.cjs.map +1 -0
  70. package/dist/{parser-VH3F4EQU.js → parser-T4FBG2FR.js} +23 -14
  71. package/dist/parser-T4FBG2FR.js.map +1 -0
  72. package/dist/pdf-ocr-2CITOAIG.cjs +16 -0
  73. package/dist/pdf-ocr-2CITOAIG.cjs.map +1 -0
  74. package/dist/pdf-ocr-RG2MFWBX.js +15 -0
  75. package/dist/pdf-ocr-U7Z4IUMA.js +19 -0
  76. package/dist/{profile-io-5YB627EO.js → profile-io-GM3NVLV3.js} +3 -3
  77. package/dist/{rasterize-V2XHHS2I.js → rasterize-4A3LJXTG.js} +3 -3
  78. package/dist/{redact-doc-5MK5L5IQ.js → redact-doc-SQ7C3LIG.js} +11 -11
  79. package/dist/{render-SYZMHYK6.js → render-6LTQZG4W.js} +7 -7
  80. package/dist/seal-ROSVLTRQ.js +10 -0
  81. package/dist/{setup-ZPLJENSK.js → setup-DGBJOROA.js} +3 -3
  82. package/dist/visual-6W34SFIW.js +20 -0
  83. package/dist/{watch-FNWHVKK3.js → watch-DJVJRSLE.js} +12 -12
  84. package/package.json +1 -1
  85. package/dist/chunk-2XHXABCX.js.map +0 -1
  86. package/dist/chunk-3N272C5L.js.map +0 -1
  87. package/dist/chunk-4TSYMJDL.js.map +0 -1
  88. package/dist/chunk-6DOCSYEM.cjs.map +0 -1
  89. package/dist/chunk-GPC72VSF.cjs.map +0 -1
  90. package/dist/chunk-HAW2VL62.cjs.map +0 -1
  91. package/dist/chunk-JY2NBFIG.js.map +0 -1
  92. package/dist/chunk-KZVQK5NS.cjs.map +0 -1
  93. package/dist/chunk-SDRKHEVT.js.map +0 -1
  94. package/dist/chunk-UX3HHAWH.js +0 -248
  95. package/dist/chunk-UX3HHAWH.js.map +0 -1
  96. package/dist/chunk-VU26WGE5.js.map +0 -1
  97. package/dist/chunk-WL3BWROT.js.map +0 -1
  98. package/dist/chunk-WWMKIHF7.cjs.map +0 -1
  99. package/dist/chunk-YIANW5QX.cjs +0 -248
  100. package/dist/chunk-YIANW5QX.cjs.map +0 -1
  101. package/dist/chunks-EBWJR55V.js +0 -10
  102. package/dist/formula-5RETDCIC.cjs.map +0 -1
  103. package/dist/image-ocr-PSPVRLB5.cjs.map +0 -1
  104. package/dist/parser-M5WHRX25.js.map +0 -1
  105. package/dist/parser-QILYPWSI.cjs.map +0 -1
  106. package/dist/parser-VH3F4EQU.js.map +0 -1
  107. package/dist/pdf-ocr-3SZUM3CN.cjs +0 -16
  108. package/dist/pdf-ocr-3SZUM3CN.cjs.map +0 -1
  109. package/dist/pdf-ocr-QB2TB5W5.js +0 -18
  110. package/dist/pdf-ocr-XC33L43D.js +0 -15
  111. package/dist/seal-IPKIQHYB.js +0 -10
  112. package/dist/visual-RTTGNOAE.js +0 -20
  113. /package/dist/{-2UA2JHVA.js.map → -SQRLFE4N.js.map} +0 -0
  114. /package/dist/{chunk-4HRJISJF.js.map → chunk-4D52VUGA.js.map} +0 -0
  115. /package/dist/{chunk-6OCKAXIS.js.map → chunk-6DXSRQ65.js.map} +0 -0
  116. /package/dist/{chunk-XZA6X34R.js.map → chunk-7TE6TK52.js.map} +0 -0
  117. /package/dist/{chunk-KQIYC2ZU.js.map → chunk-AAQR4RIX.js.map} +0 -0
  118. /package/dist/{chunk-A2W2JQ3S.js.map → chunk-DD6AY6EI.js.map} +0 -0
  119. /package/dist/{chunk-FGL4IGOT.js.map → chunk-FSCVZR4A.js.map} +0 -0
  120. /package/dist/{chunk-GYTL6WT3.js.map → chunk-HQ6TM2JE.js.map} +0 -0
  121. /package/dist/{chunk-O4CB7D2Q.js.map → chunk-IERSVUW6.js.map} +0 -0
  122. /package/dist/{chunk-MDCTITUG.js.map → chunk-O5T5OIJM.js.map} +0 -0
  123. /package/dist/{chunk-C4MVZXB5.js.map → chunk-P2ZPPRM7.js.map} +0 -0
  124. /package/dist/{chunk-YGKC6E6A.js.map → chunk-PTMYCK7R.js.map} +0 -0
  125. /package/dist/{chunk-ZDX5PIWR.js.map → chunk-RCS7WXOM.js.map} +0 -0
  126. /package/dist/{chunk-5JZEY5IL.js.map → chunk-VJP7O6BL.js.map} +0 -0
  127. /package/dist/{chunks-EBWJR55V.js.map → chunks-VNSLMJQD.js.map} +0 -0
  128. /package/dist/{formula-MBLN2DG4.js.map → formula-U3MJXXIL.js.map} +0 -0
  129. /package/dist/{formula-QZZDW5GW.js.map → formula-WZH5CMA3.js.map} +0 -0
  130. /package/dist/{image-ocr-EREKAS47.js.map → image-ocr-HYWDBN72.js.map} +0 -0
  131. /package/dist/{model-bundle-MA4MMCRJ.js.map → model-bundle-AHBUQ6SD.js.map} +0 -0
  132. /package/dist/{models-SEJ7LNSN.js.map → models-UF7B5LLG.js.map} +0 -0
  133. /package/dist/{pdf-ocr-QB2TB5W5.js.map → pdf-ocr-RG2MFWBX.js.map} +0 -0
  134. /package/dist/{pdf-ocr-XC33L43D.js.map → pdf-ocr-U7Z4IUMA.js.map} +0 -0
  135. /package/dist/{profile-io-5YB627EO.js.map → profile-io-GM3NVLV3.js.map} +0 -0
  136. /package/dist/{rasterize-V2XHHS2I.js.map → rasterize-4A3LJXTG.js.map} +0 -0
  137. /package/dist/{redact-doc-5MK5L5IQ.js.map → redact-doc-SQ7C3LIG.js.map} +0 -0
  138. /package/dist/{render-SYZMHYK6.js.map → render-6LTQZG4W.js.map} +0 -0
  139. /package/dist/{seal-IPKIQHYB.js.map → seal-ROSVLTRQ.js.map} +0 -0
  140. /package/dist/{setup-ZPLJENSK.js.map → setup-DGBJOROA.js.map} +0 -0
  141. /package/dist/{visual-RTTGNOAE.js.map → visual-6W34SFIW.js.map} +0 -0
  142. /package/dist/{watch-FNWHVKK3.js.map → watch-DJVJRSLE.js.map} +0 -0
@@ -0,0 +1,170 @@
1
+ import {
2
+ blocksToMarkdown
3
+ } from "./chunk-NNNRN6BY.js";
4
+
5
+ // src/page-markdown.ts
6
+ function blocksToPages(blocks, render = blocksToMarkdown) {
7
+ const firstNumbered = blocks.find((b) => typeof b.pageNumber === "number")?.pageNumber;
8
+ if (firstNumbered === void 0) return void 0;
9
+ const byPage = /* @__PURE__ */ new Map();
10
+ let current = firstNumbered;
11
+ for (const block of blocks) {
12
+ if (typeof block.pageNumber === "number") current = block.pageNumber;
13
+ const bucket = byPage.get(current);
14
+ if (bucket) bucket.push(block);
15
+ else byPage.set(current, [block]);
16
+ }
17
+ const numbers = [...byPage.keys()];
18
+ const min = Math.min(...numbers);
19
+ const max = Math.max(...numbers);
20
+ const pages = [];
21
+ for (let pageNumber = min; pageNumber <= max; pageNumber++) {
22
+ const pageBlocks = byPage.get(pageNumber);
23
+ pages.push({ pageNumber, markdown: pageBlocks ? render(pageBlocks) : "" });
24
+ }
25
+ return pages;
26
+ }
27
+
28
+ // src/image/transcode.ts
29
+ import { deflateSync } from "zlib";
30
+ var CRC_TABLE = (() => {
31
+ const table = new Uint32Array(256);
32
+ for (let n = 0; n < 256; n++) {
33
+ let c = n;
34
+ for (let k = 0; k < 8; k++) {
35
+ c = c & 1 ? 3988292384 ^ c >>> 1 : c >>> 1;
36
+ }
37
+ table[n] = c >>> 0;
38
+ }
39
+ return table;
40
+ })();
41
+ function crc32(bytes) {
42
+ let c = 4294967295;
43
+ for (let i = 0; i < bytes.length; i++) {
44
+ c = CRC_TABLE[(c ^ bytes[i]) & 255] ^ c >>> 8;
45
+ }
46
+ return (c ^ 4294967295) >>> 0;
47
+ }
48
+ var BI_RGB = 0;
49
+ var MAX_DIM = 32767;
50
+ var MAX_PIXELS = 36e6;
51
+ function bmpToPng(bmp) {
52
+ if (bmp.length < 54) return null;
53
+ if (bmp[0] !== 66 || bmp[1] !== 77) return null;
54
+ const dv = new DataView(bmp.buffer, bmp.byteOffset, bmp.byteLength);
55
+ const dataOffset = dv.getUint32(10, true);
56
+ const headerSize = dv.getUint32(14, true);
57
+ if (headerSize < 40) return null;
58
+ const width = dv.getInt32(18, true);
59
+ const rawHeight = dv.getInt32(22, true);
60
+ const bitCount = dv.getUint16(28, true);
61
+ const compression = dv.getUint32(30, true);
62
+ if (compression !== BI_RGB) return null;
63
+ if (bitCount !== 24 && bitCount !== 32) return null;
64
+ if (width <= 0 || rawHeight === 0) return null;
65
+ if (width > MAX_DIM || Math.abs(rawHeight) > MAX_DIM) return null;
66
+ const topDown = rawHeight < 0;
67
+ const height = Math.abs(rawHeight);
68
+ if (width * height > MAX_PIXELS) return null;
69
+ const bytesPerPixel = bitCount >> 3;
70
+ const rowStride = width * bytesPerPixel + 3 & ~3;
71
+ if (dataOffset + rowStride * height > bmp.length) return null;
72
+ const rgba = new Uint8Array(width * height * 4);
73
+ let anyAlpha = 0;
74
+ for (let y = 0; y < height; y++) {
75
+ const srcRow = topDown ? y : height - 1 - y;
76
+ let src = dataOffset + srcRow * rowStride;
77
+ let dst = y * width * 4;
78
+ for (let x = 0; x < width; x++) {
79
+ rgba[dst] = bmp[src + 2];
80
+ rgba[dst + 1] = bmp[src + 1];
81
+ rgba[dst + 2] = bmp[src];
82
+ const a = bitCount === 32 ? bmp[src + 3] : 255;
83
+ rgba[dst + 3] = a;
84
+ anyAlpha |= a;
85
+ src += bytesPerPixel;
86
+ dst += 4;
87
+ }
88
+ }
89
+ if (bitCount === 32 && anyAlpha === 0) {
90
+ for (let i = 3; i < rgba.length; i += 4) rgba[i] = 255;
91
+ }
92
+ return encodePng(width, height, rgba);
93
+ }
94
+ var PNG_SIGNATURE = Uint8Array.of(137, 80, 78, 71, 13, 10, 26, 10);
95
+ function chunk(type, data) {
96
+ const body = new Uint8Array(4 + data.length);
97
+ body[0] = type.charCodeAt(0);
98
+ body[1] = type.charCodeAt(1);
99
+ body[2] = type.charCodeAt(2);
100
+ body[3] = type.charCodeAt(3);
101
+ body.set(data, 4);
102
+ const out = new Uint8Array(8 + data.length + 4);
103
+ const dv = new DataView(out.buffer);
104
+ dv.setUint32(0, data.length, false);
105
+ out.set(body, 4);
106
+ dv.setUint32(8 + data.length, crc32(body), false);
107
+ return out;
108
+ }
109
+ function encodePng(width, height, rgba) {
110
+ const ihdr = new Uint8Array(13);
111
+ const iv = new DataView(ihdr.buffer);
112
+ iv.setUint32(0, width, false);
113
+ iv.setUint32(4, height, false);
114
+ ihdr[8] = 8;
115
+ ihdr[9] = 6;
116
+ const stride = width * 4;
117
+ const raw = new Uint8Array((stride + 1) * height);
118
+ for (let y = 0; y < height; y++) {
119
+ const rowStart = y * (stride + 1);
120
+ raw[rowStart] = 0;
121
+ raw.set(rgba.subarray(y * stride, y * stride + stride), rowStart + 1);
122
+ }
123
+ const idat = deflateSync(raw);
124
+ const ihdrChunk = chunk("IHDR", ihdr);
125
+ const idatChunk = chunk("IDAT", idat);
126
+ const iendChunk = chunk("IEND", new Uint8Array(0));
127
+ const out = new Uint8Array(PNG_SIGNATURE.length + ihdrChunk.length + idatChunk.length + iendChunk.length);
128
+ let o = 0;
129
+ out.set(PNG_SIGNATURE, o);
130
+ o += PNG_SIGNATURE.length;
131
+ out.set(ihdrChunk, o);
132
+ o += ihdrChunk.length;
133
+ out.set(idatChunk, o);
134
+ o += idatChunk.length;
135
+ out.set(iendChunk, o);
136
+ return out;
137
+ }
138
+ function inlineImagesIntoMarkdown(markdown, images, opts) {
139
+ const compress = opts?.compress !== false;
140
+ const uris = /* @__PURE__ */ new Map();
141
+ for (const img of images) {
142
+ let bytes = img.data;
143
+ let mime = img.mimeType;
144
+ if (compress && mime === "image/bmp") {
145
+ const png = bmpToPng(img.data);
146
+ if (png) {
147
+ bytes = png;
148
+ mime = "image/png";
149
+ }
150
+ }
151
+ uris.set(img.filename, `data:${mime};base64,${Buffer.from(bytes).toString("base64")}`);
152
+ }
153
+ if (uris.size === 0) return markdown;
154
+ let out = markdown.replace(/!\[image\]\((?:images\/)?([^()\s]+)\)/g, (m, name) => {
155
+ const uri = uris.get(name);
156
+ return uri !== void 0 ? `![image](${uri})` : m;
157
+ });
158
+ out = out.replace(/(<img\b[^>]*\bsrc=")(?:images\/)?([^"]*)(")/g, (m, pre, name, post) => {
159
+ const uri = uris.get(name);
160
+ return uri !== void 0 ? `${pre}${uri}${post}` : m;
161
+ });
162
+ return out;
163
+ }
164
+
165
+ export {
166
+ encodePng,
167
+ inlineImagesIntoMarkdown,
168
+ blocksToPages
169
+ };
170
+ //# sourceMappingURL=chunk-PNXJYEOR.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../src/page-markdown.ts","../src/image/transcode.ts"],"sourcesContent":["/**\n * 페이지별 마크다운 사영 (#68)\n *\n * 문서를 다시 파싱하지 않고 이미 만들어진 `IRBlock.pageNumber` 로만 블록을 갈라\n * 페이지마다 `blocksToMarkdown()` 을 돌린다. 페이지 번호의 신뢰도는 파서가\n * 정하는 것이고(`metadata.pageMode` — \"layout\" 은 실제 페이지, \"section\" 은\n * 섹션 근사), 여기서는 그 값을 그대로 존중한다.\n */\n\nimport type { IRBlock, PageMarkdown } from \"./types.js\"\nimport { blocksToMarkdown } from \"./table/builder.js\"\n\n/**\n * 블록을 페이지 번호로 묶어 페이지별 마크다운을 만든다.\n *\n * - 어떤 블록에도 페이지 번호가 없으면 `undefined` — 페이지 개념이 없는 포맷에\n * \"전부 1페이지\" 같은 없는 사실을 만들어 내지 않는다.\n * - 번호가 없는 블록은 **직전 블록의 페이지로 이어붙인다.** 블록은 읽기 순서라\n * 이게 원문 위치에 가장 가깝고, 무엇보다 어느 페이지에도 안 실려 조용히\n * 사라지는 블록이 생기지 않는다. 선두의 번호 없는 블록은 첫 실번호 페이지로.\n * - 관측된 최소~최대 페이지 사이의 빈 페이지도 항목으로 낸다. 여러 페이지에\n * 걸친 표는 시작 페이지 한 블록이라 중간 페이지가 실제로 비어 있고, 이때\n * 항목을 빼면 소비자가 배열 길이로 페이지 수를 셀 수 없다.\n * - `render` 는 쪽 블록 → 마크다운. 문서 마크다운에 포맷 고유 마무리가 붙는 파서(PDF: 쪽번호 제거·균등배분 정리)는\n * 같은 마무리를 넘겨 쪽 마크다운이 문서 마크다운과 어긋나지 않게 한다.\n */\nexport function blocksToPages(blocks: IRBlock[], render: (blocks: IRBlock[]) => string = blocksToMarkdown): PageMarkdown[] | undefined {\n const firstNumbered = blocks.find(b => typeof b.pageNumber === \"number\")?.pageNumber\n if (firstNumbered === undefined) return undefined\n\n const byPage = new Map<number, IRBlock[]>()\n let current = firstNumbered\n for (const block of blocks) {\n if (typeof block.pageNumber === \"number\") current = block.pageNumber\n const bucket = byPage.get(current)\n if (bucket) bucket.push(block)\n else byPage.set(current, [block])\n }\n\n const numbers = [...byPage.keys()]\n const min = Math.min(...numbers)\n const max = Math.max(...numbers)\n\n const pages: PageMarkdown[] = []\n for (let pageNumber = min; pageNumber <= max; pageNumber++) {\n const pageBlocks = byPage.get(pageNumber)\n pages.push({ pageNumber, markdown: pageBlocks ? render(pageBlocks) : \"\" })\n }\n return pages\n}\n","/**\n * 순수 JS BMP → PNG 트랜스코더 + 마크다운 이미지 인라이너.\n *\n * HWP5 BinData 임베드 이미지는 비압축 BMP(수 MB)인 경우가 많다. AI 에이전트가\n * 자체 완결형 마크다운을 받도록, BMP 를 PNG(무손실 압축)로 변환한 뒤 base64\n * data URI 로 마크다운에 인라인한다.\n *\n * 의존성 최소화 원칙 — Node 내장(node:zlib, Buffer)만 사용한다.\n */\n\nimport { deflateSync } from \"node:zlib\"\n\n// ─── CRC32 (PNG 청크용) ──────────────────────────────\n\nconst CRC_TABLE = (() => {\n const table = new Uint32Array(256)\n for (let n = 0; n < 256; n++) {\n let c = n\n for (let k = 0; k < 8; k++) {\n c = (c & 1) ? (0xedb88320 ^ (c >>> 1)) : (c >>> 1)\n }\n table[n] = c >>> 0\n }\n return table\n})()\n\n/** 표준 PNG CRC32 (다항식 0xEDB88320) */\nfunction crc32(bytes: Uint8Array): number {\n let c = 0xffffffff\n for (let i = 0; i < bytes.length; i++) {\n c = CRC_TABLE[(c ^ bytes[i]) & 0xff] ^ (c >>> 8)\n }\n return (c ^ 0xffffffff) >>> 0\n}\n\n// ─── BMP 디코드 ──────────────────────────────────────\n\nconst BI_RGB = 0\n/** 폭·높이 상한 — 비정상 헤더에 의한 과대 할당 방지 */\nconst MAX_DIM = 0x7fff\n/**\n * 총 픽셀(W×H) 상한 — MAX_DIM 은 각 변만 제한하므로 곱의 폭주(과대 할당 + 긴 deflateSync\n * 동기 블로킹)를 별도로 차단한다. raw 저장 BMP 는 상위 inflate 100MB 캡을 우회하므로 필요.\n * 36MP = A4 전면 600dpi 스캔(≈35MP)까지 허용 — 그 이상은 rgba+raw 로 300MB 를 넘겨\n * MCP 서버 이벤트루프를 수 초 세우므로 원본 바이트 폴백으로 돌린다.\n */\nconst MAX_PIXELS = 36_000_000\n\n/**\n * 비압축 BMP(24/32bpp, BI_RGB) → PNG(8bit RGBA).\n *\n * 지원: BITMAPFILEHEADER + BITMAPINFOHEADER(biSize>=40), biBitCount 24/32,\n * biCompression BI_RGB, bottom-up(양수 biHeight)/top-down(음수), 24bpp 4바이트 행 패딩.\n * 미지원 변형(팔레트/8bpp·RLE·BITFIELDS·OS/2 헤더 등)은 null 을 반환하여\n * 호출부가 원본 바이트로 폴백하게 한다.\n */\nexport function bmpToPng(bmp: Uint8Array): Uint8Array | null {\n // BITMAPFILEHEADER(14) + BITMAPINFOHEADER(40)\n if (bmp.length < 54) return null\n if (bmp[0] !== 0x42 || bmp[1] !== 0x4d) return null // \"BM\"\n\n const dv = new DataView(bmp.buffer, bmp.byteOffset, bmp.byteLength)\n const dataOffset = dv.getUint32(10, true)\n const headerSize = dv.getUint32(14, true)\n if (headerSize < 40) return null // OS/2 BITMAPCOREHEADER 등 미지원\n\n const width = dv.getInt32(18, true)\n const rawHeight = dv.getInt32(22, true)\n const bitCount = dv.getUint16(28, true)\n const compression = dv.getUint32(30, true)\n\n if (compression !== BI_RGB) return null // RLE / BITFIELDS 미지원\n if (bitCount !== 24 && bitCount !== 32) return null // 팔레트 / 8bpp 등 미지원\n if (width <= 0 || rawHeight === 0) return null\n if (width > MAX_DIM || Math.abs(rawHeight) > MAX_DIM) return null\n\n const topDown = rawHeight < 0\n const height = Math.abs(rawHeight)\n // 총 픽셀 상한 검사 — rgba/raw 할당·deflate 전에 조기 차단 (각 변은 MAX_DIM 이하여도 곱은 폭주 가능)\n if (width * height > MAX_PIXELS) return null\n const bytesPerPixel = bitCount >> 3\n // 행 스트라이드 — 4바이트 경계 패딩 (24bpp 에서 유효)\n const rowStride = (width * bytesPerPixel + 3) & ~3\n if (dataOffset + rowStride * height > bmp.length) return null // 잘린 파일\n\n const rgba = new Uint8Array(width * height * 4)\n let anyAlpha = 0\n\n for (let y = 0; y < height; y++) {\n // bottom-up 은 첫 행이 이미지 맨 아래 → 뒤집어 읽는다\n const srcRow = topDown ? y : height - 1 - y\n let src = dataOffset + srcRow * rowStride\n let dst = y * width * 4\n for (let x = 0; x < width; x++) {\n // BMP 픽셀은 BGR(A) 순\n rgba[dst] = bmp[src + 2] // R\n rgba[dst + 1] = bmp[src + 1] // G\n rgba[dst + 2] = bmp[src] // B\n const a = bitCount === 32 ? bmp[src + 3] : 255\n rgba[dst + 3] = a\n anyAlpha |= a\n src += bytesPerPixel\n dst += 4\n }\n }\n\n // 32bpp BI_RGB 의 4번째 바이트는 \"미정의\" — 전부 0 이면 전면 투명으로 오해되므로\n // 불투명(255)으로 승격한다. 실제 알파가 있으면(anyAlpha!=0) 그대로 보존.\n if (bitCount === 32 && anyAlpha === 0) {\n for (let i = 3; i < rgba.length; i += 4) rgba[i] = 255\n }\n\n return encodePng(width, height, rgba)\n}\n\n// ─── PNG 인코드 (8bit RGBA, color type 6) ────────────\n\nconst PNG_SIGNATURE = Uint8Array.of(0x89, 0x50, 0x4e, 0x47, 0x0d, 0x0a, 0x1a, 0x0a)\n\n/** [4B 길이][4B 타입][데이터][4B CRC] 청크 조립. CRC 는 (타입+데이터) 기준. */\nfunction chunk(type: string, data: Uint8Array): Uint8Array {\n const body = new Uint8Array(4 + data.length)\n body[0] = type.charCodeAt(0)\n body[1] = type.charCodeAt(1)\n body[2] = type.charCodeAt(2)\n body[3] = type.charCodeAt(3)\n body.set(data, 4)\n\n const out = new Uint8Array(8 + data.length + 4)\n const dv = new DataView(out.buffer)\n dv.setUint32(0, data.length, false) // 길이(데이터만, big-endian)\n out.set(body, 4)\n dv.setUint32(8 + data.length, crc32(body), false)\n return out\n}\n\n/** 8bit RGBA 픽셀 버퍼 → PNG (PDF 이미지 XObject 추출에서도 사용) */\nexport function encodePng(width: number, height: number, rgba: Uint8Array): Uint8Array {\n // IHDR: width, height, bit depth 8, color type 6(RGBA), compression 0, filter 0, interlace 0\n const ihdr = new Uint8Array(13)\n const iv = new DataView(ihdr.buffer)\n iv.setUint32(0, width, false)\n iv.setUint32(4, height, false)\n ihdr[8] = 8\n ihdr[9] = 6\n\n // IDAT: 각 스캔라인 앞에 필터 바이트 0(None) 을 붙인 raw 데이터를 zlib deflate\n const stride = width * 4\n const raw = new Uint8Array((stride + 1) * height)\n for (let y = 0; y < height; y++) {\n const rowStart = y * (stride + 1)\n raw[rowStart] = 0 // filter: None\n raw.set(rgba.subarray(y * stride, y * stride + stride), rowStart + 1)\n }\n const idat = deflateSync(raw)\n\n const ihdrChunk = chunk(\"IHDR\", ihdr)\n const idatChunk = chunk(\"IDAT\", idat)\n const iendChunk = chunk(\"IEND\", new Uint8Array(0))\n\n const out = new Uint8Array(PNG_SIGNATURE.length + ihdrChunk.length + idatChunk.length + iendChunk.length)\n let o = 0\n out.set(PNG_SIGNATURE, o); o += PNG_SIGNATURE.length\n out.set(ihdrChunk, o); o += ihdrChunk.length\n out.set(idatChunk, o); o += idatChunk.length\n out.set(iendChunk, o)\n return out\n}\n\n// ─── 마크다운 이미지 인라이너 ────────────────────────\n\ninterface InlineImage {\n filename: string\n data: Uint8Array\n mimeType: string\n}\n\n/**\n * 마크다운의 이미지 참조를 base64 data URI 로 치환해 자체 완결형 마크다운을 만든다.\n * 두 방출 형태를 모두 커버한다:\n * - `![image](FILENAME)` / `![image](images/FILENAME)` — 일반 문단 (CLI 접두사 포함)\n * - `<img src=\"FILENAME\">` — 병합/중첩 표 셀 (table/builder.ts HTML 표 경로).\n * 이걸 빼면 CLI `--inline-images` 가 이미지 저장을 생략하면서 표 셀 이미지만\n * dangling 참조로 유실된다.\n *\n * @param opts.compress 기본 true. `image/bmp` 는 PNG 로 트랜스코딩 후 인라인하며,\n * 트랜스코딩 실패 시 원본 바이트+원본 MIME 으로 폴백한다.\n */\nexport function inlineImagesIntoMarkdown(\n markdown: string,\n images: ReadonlyArray<InlineImage>,\n opts?: { compress?: boolean },\n): string {\n const compress = opts?.compress !== false\n // 파일명 → data URI 사전 구축 후 단일 패스 치환 — 이미지 k개마다 전문 재스캔하던\n // O(k²) 제거. 맵에 없는 파일명 참조는 원문 그대로 둔다 (기존 계약 동일).\n const uris = new Map<string, string>()\n for (const img of images) {\n let bytes = img.data\n let mime = img.mimeType\n if (compress && mime === \"image/bmp\") {\n const png = bmpToPng(img.data)\n if (png) {\n bytes = png\n mime = \"image/png\"\n }\n }\n uris.set(img.filename, `data:${mime};base64,${Buffer.from(bytes).toString(\"base64\")}`)\n }\n if (uris.size === 0) return markdown\n // FILENAME 과 images/FILENAME 접두사를 모두 정확한 파일명 기준으로 치환\n let out = markdown.replace(/!\\[image\\]\\((?:images\\/)?([^()\\s]+)\\)/g, (m, name: string) => {\n const uri = uris.get(name)\n return uri !== undefined ? `![image](${uri})` : m\n })\n out = out.replace(/(<img\\b[^>]*\\bsrc=\")(?:images\\/)?([^\"]*)(\")/g, (m, pre: string, name: string, post: string) => {\n const uri = uris.get(name)\n return uri !== undefined ? `${pre}${uri}${post}` : m\n })\n return out\n}\n"],"mappings":";;;;;AA0BO,SAAS,cAAc,QAAmB,SAAwC,kBAA8C;AACrI,QAAM,gBAAgB,OAAO,KAAK,OAAK,OAAO,EAAE,eAAe,QAAQ,GAAG;AAC1E,MAAI,kBAAkB,OAAW,QAAO;AAExC,QAAM,SAAS,oBAAI,IAAuB;AAC1C,MAAI,UAAU;AACd,aAAW,SAAS,QAAQ;AAC1B,QAAI,OAAO,MAAM,eAAe,SAAU,WAAU,MAAM;AAC1D,UAAM,SAAS,OAAO,IAAI,OAAO;AACjC,QAAI,OAAQ,QAAO,KAAK,KAAK;AAAA,QACxB,QAAO,IAAI,SAAS,CAAC,KAAK,CAAC;AAAA,EAClC;AAEA,QAAM,UAAU,CAAC,GAAG,OAAO,KAAK,CAAC;AACjC,QAAM,MAAM,KAAK,IAAI,GAAG,OAAO;AAC/B,QAAM,MAAM,KAAK,IAAI,GAAG,OAAO;AAE/B,QAAM,QAAwB,CAAC;AAC/B,WAAS,aAAa,KAAK,cAAc,KAAK,cAAc;AAC1D,UAAM,aAAa,OAAO,IAAI,UAAU;AACxC,UAAM,KAAK,EAAE,YAAY,UAAU,aAAa,OAAO,UAAU,IAAI,GAAG,CAAC;AAAA,EAC3E;AACA,SAAO;AACT;;;ACvCA,SAAS,mBAAmB;AAI5B,IAAM,aAAa,MAAM;AACvB,QAAM,QAAQ,IAAI,YAAY,GAAG;AACjC,WAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,QAAI,IAAI;AACR,aAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,UAAK,IAAI,IAAM,aAAc,MAAM,IAAO,MAAM;AAAA,IAClD;AACA,UAAM,CAAC,IAAI,MAAM;AAAA,EACnB;AACA,SAAO;AACT,GAAG;AAGH,SAAS,MAAM,OAA2B;AACxC,MAAI,IAAI;AACR,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,QAAI,WAAW,IAAI,MAAM,CAAC,KAAK,GAAI,IAAK,MAAM;AAAA,EAChD;AACA,UAAQ,IAAI,gBAAgB;AAC9B;AAIA,IAAM,SAAS;AAEf,IAAM,UAAU;AAOhB,IAAM,aAAa;AAUZ,SAAS,SAAS,KAAoC;AAE3D,MAAI,IAAI,SAAS,GAAI,QAAO;AAC5B,MAAI,IAAI,CAAC,MAAM,MAAQ,IAAI,CAAC,MAAM,GAAM,QAAO;AAE/C,QAAM,KAAK,IAAI,SAAS,IAAI,QAAQ,IAAI,YAAY,IAAI,UAAU;AAClE,QAAM,aAAa,GAAG,UAAU,IAAI,IAAI;AACxC,QAAM,aAAa,GAAG,UAAU,IAAI,IAAI;AACxC,MAAI,aAAa,GAAI,QAAO;AAE5B,QAAM,QAAQ,GAAG,SAAS,IAAI,IAAI;AAClC,QAAM,YAAY,GAAG,SAAS,IAAI,IAAI;AACtC,QAAM,WAAW,GAAG,UAAU,IAAI,IAAI;AACtC,QAAM,cAAc,GAAG,UAAU,IAAI,IAAI;AAEzC,MAAI,gBAAgB,OAAQ,QAAO;AACnC,MAAI,aAAa,MAAM,aAAa,GAAI,QAAO;AAC/C,MAAI,SAAS,KAAK,cAAc,EAAG,QAAO;AAC1C,MAAI,QAAQ,WAAW,KAAK,IAAI,SAAS,IAAI,QAAS,QAAO;AAE7D,QAAM,UAAU,YAAY;AAC5B,QAAM,SAAS,KAAK,IAAI,SAAS;AAEjC,MAAI,QAAQ,SAAS,WAAY,QAAO;AACxC,QAAM,gBAAgB,YAAY;AAElC,QAAM,YAAa,QAAQ,gBAAgB,IAAK,CAAC;AACjD,MAAI,aAAa,YAAY,SAAS,IAAI,OAAQ,QAAO;AAEzD,QAAM,OAAO,IAAI,WAAW,QAAQ,SAAS,CAAC;AAC9C,MAAI,WAAW;AAEf,WAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAE/B,UAAM,SAAS,UAAU,IAAI,SAAS,IAAI;AAC1C,QAAI,MAAM,aAAa,SAAS;AAChC,QAAI,MAAM,IAAI,QAAQ;AACtB,aAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAE9B,WAAK,GAAG,IAAI,IAAI,MAAM,CAAC;AACvB,WAAK,MAAM,CAAC,IAAI,IAAI,MAAM,CAAC;AAC3B,WAAK,MAAM,CAAC,IAAI,IAAI,GAAG;AACvB,YAAM,IAAI,aAAa,KAAK,IAAI,MAAM,CAAC,IAAI;AAC3C,WAAK,MAAM,CAAC,IAAI;AAChB,kBAAY;AACZ,aAAO;AACP,aAAO;AAAA,IACT;AAAA,EACF;AAIA,MAAI,aAAa,MAAM,aAAa,GAAG;AACrC,aAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK,EAAG,MAAK,CAAC,IAAI;AAAA,EACrD;AAEA,SAAO,UAAU,OAAO,QAAQ,IAAI;AACtC;AAIA,IAAM,gBAAgB,WAAW,GAAG,KAAM,IAAM,IAAM,IAAM,IAAM,IAAM,IAAM,EAAI;AAGlF,SAAS,MAAM,MAAc,MAA8B;AACzD,QAAM,OAAO,IAAI,WAAW,IAAI,KAAK,MAAM;AAC3C,OAAK,CAAC,IAAI,KAAK,WAAW,CAAC;AAC3B,OAAK,CAAC,IAAI,KAAK,WAAW,CAAC;AAC3B,OAAK,CAAC,IAAI,KAAK,WAAW,CAAC;AAC3B,OAAK,CAAC,IAAI,KAAK,WAAW,CAAC;AAC3B,OAAK,IAAI,MAAM,CAAC;AAEhB,QAAM,MAAM,IAAI,WAAW,IAAI,KAAK,SAAS,CAAC;AAC9C,QAAM,KAAK,IAAI,SAAS,IAAI,MAAM;AAClC,KAAG,UAAU,GAAG,KAAK,QAAQ,KAAK;AAClC,MAAI,IAAI,MAAM,CAAC;AACf,KAAG,UAAU,IAAI,KAAK,QAAQ,MAAM,IAAI,GAAG,KAAK;AAChD,SAAO;AACT;AAGO,SAAS,UAAU,OAAe,QAAgB,MAA8B;AAErF,QAAM,OAAO,IAAI,WAAW,EAAE;AAC9B,QAAM,KAAK,IAAI,SAAS,KAAK,MAAM;AACnC,KAAG,UAAU,GAAG,OAAO,KAAK;AAC5B,KAAG,UAAU,GAAG,QAAQ,KAAK;AAC7B,OAAK,CAAC,IAAI;AACV,OAAK,CAAC,IAAI;AAGV,QAAM,SAAS,QAAQ;AACvB,QAAM,MAAM,IAAI,YAAY,SAAS,KAAK,MAAM;AAChD,WAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAC/B,UAAM,WAAW,KAAK,SAAS;AAC/B,QAAI,QAAQ,IAAI;AAChB,QAAI,IAAI,KAAK,SAAS,IAAI,QAAQ,IAAI,SAAS,MAAM,GAAG,WAAW,CAAC;AAAA,EACtE;AACA,QAAM,OAAO,YAAY,GAAG;AAE5B,QAAM,YAAY,MAAM,QAAQ,IAAI;AACpC,QAAM,YAAY,MAAM,QAAQ,IAAI;AACpC,QAAM,YAAY,MAAM,QAAQ,IAAI,WAAW,CAAC,CAAC;AAEjD,QAAM,MAAM,IAAI,WAAW,cAAc,SAAS,UAAU,SAAS,UAAU,SAAS,UAAU,MAAM;AACxG,MAAI,IAAI;AACR,MAAI,IAAI,eAAe,CAAC;AAAG,OAAK,cAAc;AAC9C,MAAI,IAAI,WAAW,CAAC;AAAG,OAAK,UAAU;AACtC,MAAI,IAAI,WAAW,CAAC;AAAG,OAAK,UAAU;AACtC,MAAI,IAAI,WAAW,CAAC;AACpB,SAAO;AACT;AAqBO,SAAS,yBACd,UACA,QACA,MACQ;AACR,QAAM,WAAW,MAAM,aAAa;AAGpC,QAAM,OAAO,oBAAI,IAAoB;AACrC,aAAW,OAAO,QAAQ;AACxB,QAAI,QAAQ,IAAI;AAChB,QAAI,OAAO,IAAI;AACf,QAAI,YAAY,SAAS,aAAa;AACpC,YAAM,MAAM,SAAS,IAAI,IAAI;AAC7B,UAAI,KAAK;AACP,gBAAQ;AACR,eAAO;AAAA,MACT;AAAA,IACF;AACA,SAAK,IAAI,IAAI,UAAU,QAAQ,IAAI,WAAW,OAAO,KAAK,KAAK,EAAE,SAAS,QAAQ,CAAC,EAAE;AAAA,EACvF;AACA,MAAI,KAAK,SAAS,EAAG,QAAO;AAE5B,MAAI,MAAM,SAAS,QAAQ,0CAA0C,CAAC,GAAG,SAAiB;AACxF,UAAM,MAAM,KAAK,IAAI,IAAI;AACzB,WAAO,QAAQ,SAAY,YAAY,GAAG,MAAM;AAAA,EAClD,CAAC;AACD,QAAM,IAAI,QAAQ,gDAAgD,CAAC,GAAG,KAAa,MAAc,SAAiB;AAChH,UAAM,MAAM,KAAK,IAAI,IAAI;AACzB,WAAO,QAAQ,SAAY,GAAG,GAAG,GAAG,GAAG,GAAG,IAAI,KAAK;AAAA,EACrD,CAAC;AACD,SAAO;AACT;","names":[]}
@@ -1,5 +1,5 @@
1
1
  // src/utils.ts
2
- var VERSION = true ? "4.17.2" : "0.0.0-dev";
2
+ var VERSION = true ? "4.17.4" : "0.0.0-dev";
3
3
  function toArrayBuffer(buf) {
4
4
  if (buf.byteOffset === 0 && buf.byteLength === buf.buffer.byteLength) {
5
5
  return buf.buffer;
@@ -154,4 +154,4 @@ export {
154
154
  OPTIONAL_DEP_INSTALL_HINT,
155
155
  classifyError
156
156
  };
157
- //# sourceMappingURL=chunk-YGKC6E6A.js.map
157
+ //# sourceMappingURL=chunk-PTMYCK7R.js.map
@@ -5,11 +5,11 @@ import {
5
5
  HEADING_RATIO_H2,
6
6
  HEADING_RATIO_H3,
7
7
  tidyScriptTags
8
- } from "./chunk-WL3BWROT.js";
8
+ } from "./chunk-ODVIUG6Q.js";
9
9
  import {
10
10
  safeMax,
11
11
  safeMin
12
- } from "./chunk-A2W2JQ3S.js";
12
+ } from "./chunk-DD6AY6EI.js";
13
13
 
14
14
  // src/pdf/line-extract.ts
15
15
  import { OPS } from "pdfjs-dist/legacy/build/pdf.mjs";
@@ -2097,6 +2097,7 @@ var hasBatchim = (c) => {
2097
2097
  return k >= 0 && k < 11172 && k % 28 !== 0;
2098
2098
  };
2099
2099
  var CJK = /[\u3000-\u303F\u3040-\u30FF\u3400-\u4DBF\u4E00-\u9FFF\uF900-\uFAFF\uFF00-\uFFEF]/;
2100
+ var unspacedCjkBoundary = (prev, next) => CJK.test(prev.trimEnd().slice(-1)) && CJK.test(next.trimStart().charAt(0)) && !/[가-힣ㄱ-ㅎㅏ-ㅣ]/.test(prev + next);
2100
2101
  var CLOSE_TAIL = /[’”」』)\]〉》>]+$/;
2101
2102
  var TAIL = "[.,)\u300D\u300F\u2019\u201D]*";
2102
2103
  var PARTICLE = new RegExp(`^(?:\uB2E4\\.|\uC758|\uC5D0|\uC5D0\uC11C|\uC5D0\uB294|\uC5D0\uAC8C|\uC5D0\uB3C4|\uC5D0\uC11C\uB294|\uAE4C\uC9C0|\uBD80\uD130|\uCC98\uB7FC|\uB9C8\uB2E4|\uB9CC|\uB3C4|\uB4E4|\uB4E4\uC774|\uB4E4\uC758|\uB4E4\uC740|\uB4E4\uC744|\uB4E4\uC5D0\uAC8C|\uC11C\uB294|\uC11C\uB3C4|\uB77C\uACE0|\uB77C\uBA70|\uC740|\uB294|\uC744|\uB97C|\uC73C\uB85C|\uB85C|\uC73C\uB85C\uC11C|\uB85C\uC11C|\uC73C\uB85C\uC368|\uB85C\uC368)${TAIL}$`);
@@ -2146,7 +2147,7 @@ function wrapJoiner(prevText, nextText, lex) {
2146
2147
  const prev = prevText.replace(MARKUP, "").trimEnd(), next = nextText.replace(MARKUP, "").trimStart();
2147
2148
  const a = prev[prev.length - 1], b = next[0];
2148
2149
  if (!a || !b || /[,;:!?]/.test(a) || DATE_DAY_END.test(prev) && b !== "(") return " ";
2149
- if (CJK.test(a) && CJK.test(b) && !/[가-힣ㄱ-ㅎㅏ-ㅣ]/.test(prev + next)) return "";
2150
+ if (unspacedCjkBoundary(prev, next)) return "";
2150
2151
  let s = prev.length;
2151
2152
  while (s > 0 && !/\s/.test(prev[s - 1])) s--;
2152
2153
  const left = prev.slice(s), right = next.match(/^\S+/)[0];
@@ -2189,17 +2190,32 @@ function joinPageBreakWraps(blocks, lex) {
2189
2190
  }
2190
2191
  }
2191
2192
  var HANGING = /[,。、:;!?)」』】〕》〉.]\s*$/;
2193
+ var MULTI_LEVEL_NUMBER = /^\d{1,2}(?:\.\d{1,3})+\.?\s/;
2192
2194
  var BODY_FULL_TOL = 0.25;
2193
2195
  var BODY_MIN_WIDTH_EM = 12;
2194
2196
  var BODY_MAX_PITCH_EM = 2;
2195
2197
  var BODY_PITCH_REL = 1.05;
2196
2198
  var BODY_MAX_PITCH_ABS_EM = 3.5;
2197
2199
  function bodyLineJoins(lines, lex) {
2200
+ lines = lines.map((l) => {
2201
+ const text = l.text.replace(MARKUP, "");
2202
+ return text === l.text ? l : { ...l, text };
2203
+ });
2198
2204
  let right = -Infinity;
2199
2205
  for (const l of lines) if (l.right > right) right = l.right;
2200
2206
  let inner = -Infinity;
2201
2207
  for (const l of lines) if (!HANGING.test(l.text) && l.right > inner) inner = l.right;
2202
2208
  const full = (l) => right - l.right < BODY_FULL_TOL * l.fontSize || right - inner <= 1.2 * l.fontSize && inner - l.right < BODY_FULL_TOL * l.fontSize;
2209
+ const nextWordNoRoom = (a, b) => {
2210
+ const w = b.text.trim().match(/^\S+/)?.[0];
2211
+ const n = [...b.text.trim()].length;
2212
+ if (!w || n === 0) return false;
2213
+ const em = (b.right - b.left) / n;
2214
+ const advance = unspacedCjkBoundary(a.text, b.text) ? 1 : [...w].length + 1;
2215
+ return right - a.right < advance * em;
2216
+ };
2217
+ const candidates = lines.filter((l, i) => full(l) || i + 1 < lines.length && nextWordNoRoom(l, lines[i + 1]));
2218
+ const ragged = candidates.filter(full).length * 2 < candidates.length;
2203
2219
  const pitch = (k) => lines[k].y - lines[k + 1].y;
2204
2220
  const sameSize = (k) => Math.abs(lines[k + 1].fontSize - lines[k].fontSize) <= 0.15 * lines[k].fontSize;
2205
2221
  let p1 = Infinity, p2 = Infinity, i1 = -1;
@@ -2218,8 +2234,9 @@ function bodyLineJoins(lines, lex) {
2218
2234
  const fs = a.fontSize;
2219
2235
  const others = i === i1 ? p2 : p1;
2220
2236
  const maxPitch = Number.isFinite(others) ? Math.min(BODY_MAX_PITCH_ABS_EM * fs, Math.max(BODY_MAX_PITCH_EM * fs, others * BODY_PITCH_REL)) : BODY_MAX_PITCH_EM * fs;
2221
- const wraps = fs > 0 && full(a) && a.right - a.left >= BODY_MIN_WIDTH_EM * fs && a.y - b.y > 0 && a.y - b.y < maxPitch && Math.abs(b.fontSize - fs) <= 0.15 * fs && !startsNewItem(a.text, b.text);
2222
- out.push(wraps ? wrapJoiner(a.text, b.text, lex) : "\n");
2237
+ const wordWrap = ragged && !full(a) && nextWordNoRoom(a, b) && !MULTI_LEVEL_NUMBER.test(b.text.trimStart());
2238
+ const wraps = fs > 0 && (full(a) || wordWrap) && a.right - a.left >= BODY_MIN_WIDTH_EM * fs && a.y - b.y > 0 && a.y - b.y < maxPitch && Math.abs(b.fontSize - fs) <= 0.15 * fs && !startsNewItem(a.text, b.text);
2239
+ out.push(!wraps ? "\n" : wordWrap ? unspacedCjkBoundary(a.text, b.text) ? "" : " " : wrapJoiner(a.text, b.text, lex));
2223
2240
  }
2224
2241
  return out;
2225
2242
  }
@@ -3268,15 +3285,18 @@ function detectTypographyHeadings(blocks) {
3268
3285
  const faceWeight = /* @__PURE__ */ new Map();
3269
3286
  for (const block of page) {
3270
3287
  if (block.type !== "paragraph" || !block.text || !block.style?.fontName) continue;
3271
- const face = block.style.fontName;
3272
- faceWeight.set(face, (faceWeight.get(face) ?? 0) + block.text.length);
3288
+ const faces = FACE_CHARS.get(block);
3289
+ if (faces) for (const [face, n] of faces) faceWeight.set(face, (faceWeight.get(face) ?? 0) + n);
3290
+ else faceWeight.set(block.style.fontName, (faceWeight.get(block.style.fontName) ?? 0) + block.text.length);
3273
3291
  }
3274
3292
  const bodyFace = [...faceWeight].sort((a, b) => b[1] - a[1])[0]?.[0];
3275
3293
  if (!bodyFace || (faceWeight.get(bodyFace) ?? 0) < 250) continue;
3276
3294
  const sizeWeight = /* @__PURE__ */ new Map();
3277
3295
  for (const block of page) {
3278
- if (block.type !== "paragraph" || block.style?.fontName !== bodyFace || !block.style.fontSize) continue;
3279
- sizeWeight.set(block.style.fontSize, (sizeWeight.get(block.style.fontSize) ?? 0) + block.text.length);
3296
+ if (block.type !== "paragraph" || !block.style?.fontSize) continue;
3297
+ const faces = FACE_CHARS.get(block);
3298
+ const chars = faces ? faces.get(bodyFace) ?? 0 : block.style.fontName === bodyFace ? block.text?.length ?? 0 : 0;
3299
+ if (chars) sizeWeight.set(block.style.fontSize, (sizeWeight.get(block.style.fontSize) ?? 0) + chars);
3280
3300
  }
3281
3301
  const bodySize = [...sizeWeight].sort((a, b) => b[1] - a[1])[0]?.[0] ?? 0;
3282
3302
  for (let i = 0; i < page.length; i++) {
@@ -3385,7 +3405,9 @@ function detectRepeatedPageLabels(blocks) {
3385
3405
  for (const page of byPage.values()) {
3386
3406
  const bodyChars = /* @__PURE__ */ new Map();
3387
3407
  for (const b of page) if (b.type === "paragraph" && b.text && b.style?.fontName) {
3388
- bodyChars.set(b.style.fontName, (bodyChars.get(b.style.fontName) ?? 0) + b.text.length);
3408
+ const faces = FACE_CHARS.get(b);
3409
+ if (faces) for (const [face, n] of faces) bodyChars.set(face, (bodyChars.get(face) ?? 0) + n);
3410
+ else bodyChars.set(b.style.fontName, (bodyChars.get(b.style.fontName) ?? 0) + b.text.length);
3389
3411
  }
3390
3412
  const bodyFace = [...bodyChars].sort((a, b) => b[1] - a[1])[0]?.[0];
3391
3413
  const candidates = page.filter((b) => b.type === "paragraph" && b.text && b.bbox && b.style?.fontName && b.style.fontName !== bodyFace && b.bbox.height <= (b.style.fontSize ?? 0) * 1.6 && b.text.trim().length >= 5 && b.text.trim().length <= 80);
@@ -7113,9 +7135,8 @@ function extractBlocksWithGrids(items, pageNum, pageWidth, pageHeight, grids, ho
7113
7135
  continue;
7114
7136
  }
7115
7137
  let lastLeft = -1;
7116
- const beside = (b) => !!b.bbox && b.bbox.x + b.bbox.width <= box.x - 3 && b.bbox.y + b.bbox.height >= box.y && b.bbox.y <= box.y + box.height;
7117
7138
  units.forEach((unit, index) => {
7118
- if (unit.some(beside) && unit.every((b) => beside(b) || !!b.bbox && b.bbox.y + b.bbox.height < box.y)) lastLeft = index;
7139
+ if (isProseBesidePanel(unit, box)) lastLeft = index;
7119
7140
  });
7120
7141
  const current = units.findIndex((unit) => unit.includes(sidebar));
7121
7142
  if (lastLeft > current) {
@@ -7131,6 +7152,10 @@ function extractBlocksWithGrids(items, pageNum, pageWidth, pageHeight, grids, ho
7131
7152
  }
7132
7153
  return mergeAdjacentTableBlocks(ordered);
7133
7154
  }
7155
+ function isProseBesidePanel(unit, box) {
7156
+ const beside = (b) => !!b.bbox && b.bbox.x + b.bbox.width <= box.x - 3 && b.bbox.y + b.bbox.height >= box.y && b.bbox.y <= box.y + box.height;
7157
+ return unit.some(beside) && unit.every((b) => beside(b) || !!b.bbox && b.bbox.y < box.y);
7158
+ }
7134
7159
  var SIDEBAR_ITEMS = /* @__PURE__ */ new WeakMap();
7135
7160
  function clusterTableBlock(cr, source, pageNum, horizontals = []) {
7136
7161
  const b = cr.bbox;
@@ -7404,4 +7429,4 @@ export {
7404
7429
  joinCellEdges,
7405
7430
  extractPageBlocksWithLines
7406
7431
  };
7407
- //# sourceMappingURL=chunk-4TSYMJDL.js.map
7432
+ //# sourceMappingURL=chunk-PVJFCDND.js.map