kordoc 4.2.9 → 4.3.1

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 (166) hide show
  1. package/README.md +922 -885
  2. package/dist/{-ECUCTQTR.js → -RH6BEVTT.js} +28 -20
  3. package/dist/{chunk-J2RZ444H.js → chunk-3FFPHTXD.js} +1 -1
  4. package/dist/chunk-3FFPHTXD.js.map +1 -0
  5. package/dist/{chunk-PATPZTH7.cjs → chunk-5F5UCACZ.cjs} +6 -6
  6. package/dist/chunk-5F5UCACZ.cjs.map +1 -0
  7. package/dist/{chunk-5RPYBC2Q.js → chunk-6GBLFQMA.js} +1 -1
  8. package/dist/chunk-6GBLFQMA.js.map +1 -0
  9. package/dist/{chunk-GEPINOL6.js → chunk-6XPXIG57.js} +2 -2
  10. package/dist/chunk-6XPXIG57.js.map +1 -0
  11. package/dist/{chunk-4ZWKFDUC.cjs → chunk-7DZPPUQD.cjs} +44 -13
  12. package/dist/chunk-7DZPPUQD.cjs.map +1 -0
  13. package/dist/{chunk-MHMVFI2N.js → chunk-BPJSTJBN.js} +4 -4
  14. package/dist/chunk-BPJSTJBN.js.map +1 -0
  15. package/dist/{chunk-XU4VWOKL.js → chunk-BYEQDLZS.js} +4 -3
  16. package/dist/chunk-BYEQDLZS.js.map +1 -0
  17. package/dist/{chunk-3PLF5UIS.js → chunk-CD5X6UG5.js} +37 -6
  18. package/dist/chunk-CD5X6UG5.js.map +1 -0
  19. package/dist/chunk-DCZVOIEO.cjs.map +1 -1
  20. package/dist/{chunk-STYN45MS.js → chunk-FUXDD36A.js} +4 -4
  21. package/dist/chunk-FUXDD36A.js.map +1 -0
  22. package/dist/chunk-GS7T56RP.cjs.map +1 -1
  23. package/dist/{chunk-YR4PF56B.js → chunk-GTPEKBN6.js} +2 -2
  24. package/dist/chunk-GTPEKBN6.js.map +1 -0
  25. package/dist/{chunk-27Z4PUZZ.js → chunk-HPRK67YB.js} +38 -7
  26. package/dist/chunk-HPRK67YB.js.map +1 -0
  27. package/dist/{chunk-NIISBGSR.js → chunk-HZU7Y22R.js} +71 -29
  28. package/dist/chunk-HZU7Y22R.js.map +1 -0
  29. package/dist/{chunk-7ZARSBQ3.js → chunk-LEC2HG7A.js} +4 -3
  30. package/dist/chunk-LEC2HG7A.js.map +1 -0
  31. package/dist/{chunk-H7OWEYH2.cjs → chunk-MGGU3ZRK.cjs} +1 -1
  32. package/dist/chunk-MGGU3ZRK.cjs.map +1 -0
  33. package/dist/chunk-MOL7MDBG.js +0 -0
  34. package/dist/{chunk-DZIXKL7E.js → chunk-O5EK52TG.js} +1 -1
  35. package/dist/chunk-O5EK52TG.js.map +1 -0
  36. package/dist/{chunk-LXE7OR6N.js → chunk-PV3NVBGL.js} +3 -3
  37. package/dist/chunk-PV3NVBGL.js.map +1 -0
  38. package/dist/{chunk-G3DMDTZI.js → chunk-RO4Y4BFI.js} +1511 -1135
  39. package/dist/chunk-RO4Y4BFI.js.map +1 -0
  40. package/dist/{chunk-WTT2KJOM.js → chunk-SA22NK56.js} +14 -10
  41. package/dist/chunk-SA22NK56.js.map +1 -0
  42. package/dist/{chunk-GZ7W64QU.cjs → chunk-VQB5P5KJ.cjs} +4 -3
  43. package/dist/chunk-VQB5P5KJ.cjs.map +1 -0
  44. package/dist/{chunk-7FCNOKOR.js → chunk-W34QSJFF.js} +1 -1
  45. package/dist/chunk-W34QSJFF.js.map +1 -0
  46. package/dist/{chunk-VEHC4EJL.js → chunk-YDCD4AV4.js} +5 -5
  47. package/dist/chunk-YDCD4AV4.js.map +1 -0
  48. package/dist/{chunk-NNDVFNMJ.js → chunk-YNJDNP44.js} +1 -1
  49. package/dist/chunk-YNJDNP44.js.map +1 -0
  50. package/dist/{chunk-3TSTZGCJ.js → chunk-ZJ75YDW7.js} +4 -3
  51. package/dist/chunk-ZJ75YDW7.js.map +1 -0
  52. package/dist/{chunk-6GM2QLM3.cjs → chunk-ZMKTKJZN.cjs} +21 -21
  53. package/dist/chunk-ZMKTKJZN.cjs.map +1 -0
  54. package/dist/{chunk-LM7C6HI2.js → chunk-ZSL6MQVE.js} +1 -1
  55. package/dist/chunk-ZSL6MQVE.js.map +1 -0
  56. package/dist/{chunk-KTPZYXN2.js → chunk-ZYYTU7NY.js} +2 -2
  57. package/dist/chunk-ZYYTU7NY.js.map +1 -0
  58. package/dist/chunks-JEYQWXIZ.js +10 -0
  59. package/dist/cli.js +75 -39
  60. package/dist/cli.js.map +1 -1
  61. package/dist/{formula-SXC6LRRC.js → formula-7ZPC54JO.js} +2 -2
  62. package/dist/formula-7ZPC54JO.js.map +1 -0
  63. package/dist/{formula-KSNAEBZ2.js → formula-EQRHSMYW.js} +2 -2
  64. package/dist/formula-EQRHSMYW.js.map +1 -0
  65. package/dist/{formula-SB3NHDUM.cjs → formula-IWV5IJ3Y.cjs} +8 -8
  66. package/dist/formula-IWV5IJ3Y.cjs.map +1 -0
  67. package/dist/{image-ocr-PS7XCV6K.js → image-ocr-6MO4FXJK.js} +8 -6
  68. package/dist/image-ocr-6MO4FXJK.js.map +1 -0
  69. package/dist/{image-ocr-J5656BEE.cjs → image-ocr-MGVXT6XD.cjs} +13 -11
  70. package/dist/image-ocr-MGVXT6XD.cjs.map +1 -0
  71. package/dist/{image-ocr-KZTNJ62U.js → image-ocr-TWUU2RNG.js} +10 -8
  72. package/dist/image-ocr-TWUU2RNG.js.map +1 -0
  73. package/dist/index.cjs +1993 -1845
  74. package/dist/index.cjs.map +1 -1
  75. package/dist/index.d.cts +75 -4
  76. package/dist/index.d.ts +75 -4
  77. package/dist/index.js +2128 -1980
  78. package/dist/index.js.map +1 -1
  79. package/dist/mcp.js +123 -78
  80. package/dist/mcp.js.map +1 -1
  81. package/dist/{models-KA4GLQK5.js → models-6VMAVZR6.js} +3 -3
  82. package/dist/page-range-737B4EZW.js +0 -0
  83. package/dist/page-range-P7SDW6LR.cjs.map +1 -1
  84. package/dist/{parser-CSYU3CAW.cjs → parser-JFGLN4DA.cjs} +25 -25
  85. package/dist/parser-JFGLN4DA.cjs.map +1 -0
  86. package/dist/{parser-WNFWMBSL.js → parser-VR2BTUT7.js} +8 -8
  87. package/dist/parser-VR2BTUT7.js.map +1 -0
  88. package/dist/{parser-A34NAYKA.js → parser-YA7HSSPK.js} +6 -6
  89. package/dist/parser-YA7HSSPK.js.map +1 -0
  90. package/dist/pdf-ocr-4LISPZEM.cjs +13 -0
  91. package/dist/pdf-ocr-4LISPZEM.cjs.map +1 -0
  92. package/dist/pdf-ocr-AIWB2QCS.js +12 -0
  93. package/dist/pdf-ocr-QMOKUMC2.js +15 -0
  94. package/dist/{profile-io-43MF52H6.js → profile-io-QPIMGVO3.js} +3 -3
  95. package/dist/{rasterize-OG7KXBJS.js → rasterize-AFYGNKXG.js} +2 -2
  96. package/dist/rasterize-AFYGNKXG.js.map +1 -0
  97. package/dist/{redact-7ILEAUTS.js → redact-W5K3ILQH.js} +2 -2
  98. package/dist/render-PSPTMZIZ.js +10 -0
  99. package/dist/seal-4XBJ7ISE.js +10 -0
  100. package/dist/{setup-GUQT5ELJ.js → setup-I4I75QJU.js} +1 -1
  101. package/dist/setup-I4I75QJU.js.map +1 -0
  102. package/dist/{watch-SOHQMZN2.js → watch-D3JJEOAB.js} +12 -14
  103. package/dist/watch-D3JJEOAB.js.map +1 -0
  104. package/package.json +101 -98
  105. package/templates//352/260/204/354/235/264/352/270/260/354/225/210/353/254/270_/354/204/234/354/213/235.hwpx +0 -0
  106. package/templates//352/260/204/354/235/264/352/270/260/354/225/210/353/254/270_/354/230/210/354/213/234/352/260/222.json +9 -0
  107. package/templates//354/235/274/353/260/230/352/270/260/354/225/210/353/254/270_/354/204/234/354/213/235.hwpx +0 -0
  108. package/templates//354/235/274/353/260/230/352/270/260/354/225/210/353/254/270_/354/230/210/354/213/234/352/260/222.json +25 -0
  109. package/dist/chunk-27Z4PUZZ.js.map +0 -1
  110. package/dist/chunk-3PLF5UIS.js.map +0 -1
  111. package/dist/chunk-3TSTZGCJ.js.map +0 -1
  112. package/dist/chunk-4ZWKFDUC.cjs.map +0 -1
  113. package/dist/chunk-5RPYBC2Q.js.map +0 -1
  114. package/dist/chunk-6GM2QLM3.cjs.map +0 -1
  115. package/dist/chunk-7FCNOKOR.js.map +0 -1
  116. package/dist/chunk-7ZARSBQ3.js.map +0 -1
  117. package/dist/chunk-DZIXKL7E.js.map +0 -1
  118. package/dist/chunk-G3DMDTZI.js.map +0 -1
  119. package/dist/chunk-GEPINOL6.js.map +0 -1
  120. package/dist/chunk-GZ7W64QU.cjs.map +0 -1
  121. package/dist/chunk-H7OWEYH2.cjs.map +0 -1
  122. package/dist/chunk-J2RZ444H.js.map +0 -1
  123. package/dist/chunk-KTPZYXN2.js.map +0 -1
  124. package/dist/chunk-LM7C6HI2.js.map +0 -1
  125. package/dist/chunk-LXE7OR6N.js.map +0 -1
  126. package/dist/chunk-MHMVFI2N.js.map +0 -1
  127. package/dist/chunk-NIISBGSR.js.map +0 -1
  128. package/dist/chunk-NNDVFNMJ.js.map +0 -1
  129. package/dist/chunk-PATPZTH7.cjs.map +0 -1
  130. package/dist/chunk-STYN45MS.js.map +0 -1
  131. package/dist/chunk-TCJRGFYY.js +0 -280
  132. package/dist/chunk-TCJRGFYY.js.map +0 -1
  133. package/dist/chunk-VEHC4EJL.js.map +0 -1
  134. package/dist/chunk-WTT2KJOM.js.map +0 -1
  135. package/dist/chunk-XU4VWOKL.js.map +0 -1
  136. package/dist/chunk-YR4PF56B.js.map +0 -1
  137. package/dist/chunks-VFIMYK36.js +0 -10
  138. package/dist/detect-FX2ZHKZP.js +0 -26
  139. package/dist/formula-KSNAEBZ2.js.map +0 -1
  140. package/dist/formula-SB3NHDUM.cjs.map +0 -1
  141. package/dist/formula-SXC6LRRC.js.map +0 -1
  142. package/dist/image-ocr-J5656BEE.cjs.map +0 -1
  143. package/dist/image-ocr-KZTNJ62U.js.map +0 -1
  144. package/dist/image-ocr-PS7XCV6K.js.map +0 -1
  145. package/dist/parser-A34NAYKA.js.map +0 -1
  146. package/dist/parser-CSYU3CAW.cjs.map +0 -1
  147. package/dist/parser-WNFWMBSL.js.map +0 -1
  148. package/dist/pdf-ocr-KBMWPG2L.js +0 -12
  149. package/dist/pdf-ocr-NPZG67FN.js +0 -15
  150. package/dist/pdf-ocr-O44DAGTY.cjs +0 -13
  151. package/dist/pdf-ocr-O44DAGTY.cjs.map +0 -1
  152. package/dist/rasterize-OG7KXBJS.js.map +0 -1
  153. package/dist/render-336CJHAS.js +0 -10
  154. package/dist/seal-TM2XCE2R.js +0 -10
  155. package/dist/seal-TM2XCE2R.js.map +0 -1
  156. package/dist/setup-GUQT5ELJ.js.map +0 -1
  157. package/dist/watch-SOHQMZN2.js.map +0 -1
  158. /package/dist/{-ECUCTQTR.js.map → -RH6BEVTT.js.map} +0 -0
  159. /package/dist/{chunks-VFIMYK36.js.map → chunks-JEYQWXIZ.js.map} +0 -0
  160. /package/dist/{detect-FX2ZHKZP.js.map → models-6VMAVZR6.js.map} +0 -0
  161. /package/dist/{models-KA4GLQK5.js.map → pdf-ocr-AIWB2QCS.js.map} +0 -0
  162. /package/dist/{pdf-ocr-KBMWPG2L.js.map → pdf-ocr-QMOKUMC2.js.map} +0 -0
  163. /package/dist/{pdf-ocr-NPZG67FN.js.map → profile-io-QPIMGVO3.js.map} +0 -0
  164. /package/dist/{profile-io-43MF52H6.js.map → redact-W5K3ILQH.js.map} +0 -0
  165. /package/dist/{redact-7ILEAUTS.js.map → render-PSPTMZIZ.js.map} +0 -0
  166. /package/dist/{render-336CJHAS.js.map → seal-4XBJ7ISE.js.map} +0 -0
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../src/ocr/ruling-lines.ts","../src/ocr/engine.ts","../src/ocr/pdf-ocr.ts"],"sourcesContent":["/**\r\n * 래스터 괘선 감지 — 스캔/이미지 페이지의 픽셀에서 표 괘선(수평/수직 선)을\r\n * 찾아 선 기반 표 파이프라인(line-detector → table-grid)에 공급한다.\r\n *\r\n * PDF 그래픽 ops 가 없는 OCR 경로는 종전에 클러스터 감지기만 탔는데,\r\n * 병합 라벨 셀 + 다중줄 서술형 서식(정부 제출 서식류)은 라인 클러스터로\r\n * 행 경계를 잡을 수 없다. 괘선은 이미지에 실존하므로 이진화 + 런렝스로\r\n * 직접 감지한다 — ML 없음, 축 정렬 전제(스캔 스큐 보정은 스코프 밖).\r\n *\r\n * 오탐 방어 3겹 (실측: 우수사례 제출 서식 스캔 — 진짜 괘선은 길고 얇음):\r\n * - 최소 길이 20pt: 글리프 획(가로 ≤한 글자폭·세로 ≤글자높이)은 걸러짐\r\n * - 두께 상한 2.5pt: 굵은 제목 획(6~13px)·텍스트 줄 밴드·채운 색상바 탈락\r\n * - 양측 잉크 포위 제외: 색상바(붙임 박스 등) 안 흰 글자 틈새 슬리버 탈락\r\n */\r\n\r\nimport type { LineSegment } from \"../pdf/line-types.js\"\r\n\r\n/** 잉크 판정 휘도 상한 — 셀 배경 음영(회색 ~220)은 통과, 실선(진회~검정·남색)만 잉크 */\r\nconst INK_LUMA_MAX = 190\r\n/** 최소 선 길이 (pt) — 최소 셀 변(≈7mm)과 대형 글리프 획(≤1em) 사이 */\r\nconst MIN_LINE_LENGTH_PT = 20\r\n/** 선 두께 상한 (pt) — 진짜 괘선(0.12~0.7mm ≈ 0.3~2pt)과 굵은 글리프 획(2pt+) 분리 */\r\nconst MAX_LINE_WIDTH_PT = 2.5\r\n/** 런 내 허용 끊김 (px) — 스캔 노이즈/JPEG 아티팩트 */\r\nconst GAP_TOL_PX = 2\r\n/** 밴드 병합 시 요구하는 x(또는 y) 범위 겹침 비율 (짧은 쪽 기준) */\r\nconst BAND_OVERLAP_RATIO = 0.5\r\n/** 밴드 양측(±2px)이 이 비율 이상 잉크면 채운 영역 내부 슬리버로 판정 */\r\nconst SURROUND_INK_RATIO = 0.35\r\n\r\n/** 픽셀 공간(top-left origin) 선분 — 두께는 밴드 픽셀 수 */\r\nexport interface PxSegment {\r\n x1: number; y1: number\r\n x2: number; y2: number\r\n thicknessPx: number\r\n}\r\n\r\nexport interface RulingLines {\r\n horizontals: PxSegment[]\r\n verticals: PxSegment[]\r\n}\r\n\r\n/**\r\n * RGBA 래스터에서 수평/수직 괘선 감지.\r\n * @param scale 렌더 스케일 (px/pt) — 길이·두께 임계를 pt 기준으로 환산\r\n */\r\nexport function detectRulingLines(\r\n rgba: Uint8Array,\r\n width: number,\r\n height: number,\r\n scale: number,\r\n): RulingLines {\r\n const minLenPx = Math.round(MIN_LINE_LENGTH_PT * scale)\r\n const maxThickPx = Math.max(1, Math.floor(MAX_LINE_WIDTH_PT * scale))\r\n\r\n // 잉크 마스크 1회 구축 — 수평/수직 두 패스가 공유\r\n const ink = new Uint8Array(width * height)\r\n for (let p = 0, i = 0; p < ink.length; p++, i += 4) {\r\n // ITU-R BT.601 휘도 근사 (정수 시프트)\r\n const luma = (rgba[i] * 77 + rgba[i + 1] * 150 + rgba[i + 2] * 29) >> 8\r\n if (luma <= INK_LUMA_MAX && rgba[i + 3] >= 128) ink[p] = 1\r\n }\r\n\r\n const horizontals = detectBands(ink, width, height, minLenPx, maxThickPx, false)\r\n const verticals = detectBands(ink, width, height, minLenPx, maxThickPx, true)\r\n return { horizontals, verticals }\r\n}\r\n\r\n/** 열린 밴드 — 인접 스캔라인의 런을 누적 */\r\ninterface Band {\r\n lo: number // 런 진행축 시작 (h: x1, v: y1)\r\n hi: number // 런 진행축 끝\r\n first: number // 직교축 시작 스캔라인 (h: yTop, v: xLeft)\r\n last: number // 직교축 마지막 스캔라인\r\n}\r\n\r\n/**\r\n * 스캔라인 런렝스 → 인접 라인 밴드 병합 → 두께 필터 → 선분 방출.\r\n * transpose=false 면 수평선(행 스캔), true 면 수직선(열 스캔).\r\n */\r\nfunction detectBands(\r\n ink: Uint8Array,\r\n width: number,\r\n height: number,\r\n minLenPx: number,\r\n maxThickPx: number,\r\n transpose: boolean,\r\n): PxSegment[] {\r\n const lines = transpose ? width : height // 스캔라인 개수 (직교축)\r\n const span = transpose ? height : width // 런 진행축 길이\r\n const at = transpose\r\n ? (line: number, pos: number) => ink[pos * width + line]\r\n : (line: number, pos: number) => ink[line * width + pos]\r\n\r\n const out: PxSegment[] = []\r\n let open: Band[] = []\r\n\r\n /** 밴드 양측(±2px 스캔라인)의 잉크 비율 — 채운 색상바 안 흰 글자 틈새 슬리버 판별 */\r\n const surroundInkRatio = (b: Band, side: number): number => {\r\n if (side < 0 || side >= lines) return 0\r\n let dark = 0, total = 0\r\n for (let pos = b.lo; pos <= b.hi; pos += 3) {\r\n total++\r\n if (at(side, pos) === 1) dark++\r\n }\r\n return total > 0 ? dark / total : 0\r\n }\r\n\r\n const emit = (b: Band) => {\r\n const thick = b.last - b.first + 1\r\n if (thick > maxThickPx) return // 색상바·텍스트 밴드\r\n if (\r\n surroundInkRatio(b, b.first - 2) >= SURROUND_INK_RATIO &&\r\n surroundInkRatio(b, b.last + 2) >= SURROUND_INK_RATIO\r\n ) return // 채운 영역 내부 슬리버\r\n const center = (b.first + b.last) / 2\r\n out.push(\r\n transpose\r\n ? { x1: center, y1: b.lo, x2: center, y2: b.hi, thicknessPx: thick }\r\n : { x1: b.lo, y1: center, x2: b.hi, y2: center, thicknessPx: thick },\r\n )\r\n }\r\n\r\n for (let line = 0; line < lines; line++) {\r\n // 1) 이 스캔라인의 잉크 런 수집 (GAP_TOL_PX 이하 끊김 허용)\r\n const runs: Array<{ lo: number; hi: number }> = []\r\n let runStart = -1\r\n let gap = 0\r\n for (let pos = 0; pos <= span; pos++) {\r\n const on = pos < span && at(line, pos) === 1\r\n if (on) {\r\n if (runStart < 0) runStart = pos\r\n gap = 0\r\n } else if (runStart >= 0) {\r\n if (++gap > GAP_TOL_PX || pos >= span) {\r\n const hi = pos - gap\r\n if (hi - runStart + 1 >= minLenPx) runs.push({ lo: runStart, hi })\r\n runStart = -1\r\n gap = 0\r\n }\r\n }\r\n }\r\n\r\n // 2) 열린 밴드와 병합 — 직전 스캔라인까지 이어졌고 범위가 겹치면 확장\r\n const next: Band[] = []\r\n const used = new Set<number>()\r\n for (const band of open) {\r\n if (band.last !== line - 1) { emit(band); continue } // 연속 끊김 → 확정\r\n let merged = false\r\n for (let r = 0; r < runs.length; r++) {\r\n if (used.has(r)) continue\r\n const run = runs[r]\r\n const overlap = Math.min(band.hi, run.hi) - Math.max(band.lo, run.lo) + 1\r\n const shorter = Math.min(band.hi - band.lo, run.hi - run.lo) + 1\r\n if (overlap >= shorter * BAND_OVERLAP_RATIO) {\r\n band.lo = Math.min(band.lo, run.lo)\r\n band.hi = Math.max(band.hi, run.hi)\r\n band.last = line\r\n next.push(band)\r\n used.add(r)\r\n merged = true\r\n break\r\n }\r\n }\r\n if (!merged) emit(band)\r\n }\r\n for (let r = 0; r < runs.length; r++) {\r\n if (!used.has(r)) next.push({ lo: runs[r].lo, hi: runs[r].hi, first: line, last: line })\r\n }\r\n open = next\r\n }\r\n for (const band of open) emit(band)\r\n return out\r\n}\r\n\r\n/**\r\n * 픽셀 선분 → PDF pt LineSegment (bottom-up, classifyAndAdd 규약:\r\n * 수평선 y1==y2·x1<x2, 수직선 x1==x2·y1<y2).\r\n */\r\nexport function rulingToPdfLines(\r\n ruling: RulingLines,\r\n scale: number,\r\n pdfHeight: number,\r\n): { horizontals: LineSegment[]; verticals: LineSegment[] } {\r\n const horizontals: LineSegment[] = ruling.horizontals.map(s => {\r\n const y = pdfHeight - (s.y1 + s.y2) / 2 / scale\r\n return {\r\n x1: Math.min(s.x1, s.x2) / scale,\r\n y1: y,\r\n x2: Math.max(s.x1, s.x2) / scale,\r\n y2: y,\r\n lineWidth: s.thicknessPx / scale,\r\n }\r\n })\r\n const verticals: LineSegment[] = ruling.verticals.map(s => {\r\n const x = (s.x1 + s.x2) / 2 / scale\r\n const yA = pdfHeight - Math.max(s.y1, s.y2) / scale\r\n const yB = pdfHeight - Math.min(s.y1, s.y2) / scale\r\n return { x1: x, y1: yA, x2: x, y2: yB, lineWidth: s.thicknessPx / scale }\r\n })\r\n return { horizontals, verticals }\r\n}\r\n","/**\r\n * 내장 텍스트 OCR 엔진 — PP-OCRv5 korean (det DBNet + rec SVTR/CTC) ONNX 추론.\r\n *\r\n * 파이프라인: 페이지 RGBA → det(선 검출) → 라인 crop → rec(CTC 인식) → OcrItem[]\r\n * (좌표는 입력 픽셀 기준 top-left origin — 호출자가 PDF 좌표계로 변환).\r\n *\r\n * 전·후처리는 공식 inference.yml 스펙 그대로:\r\n * - det: BGR, 긴 변 960 리사이즈(32 배수), mean/std [0.485,0.456,0.406]/[0.229,0.224,0.225],\r\n * DBPostProcess thresh 0.3 / box_thresh 0.6 / unclip_ratio 1.5\r\n * - rec: BGR, 높이 48 고정 비율 리사이즈 + 우측 zero-pad, (x/255-0.5)/0.5,\r\n * CTC 디코드 (blank=0, 1..N=사전, N+1=공백), text_score 0.5\r\n * DB 후처리의 contour+minAreaRect 는 축정렬 connected-component bbox 로 근사\r\n * (공문서 스캔은 수평 텍스트가 지배적 — 회전 텍스트는 v1 범위 밖).\r\n *\r\n * 의존성(onnxruntime-node, sharp)은 optional — 미설치 시 create()가 명확한 에러.\r\n * 모델 미다운로드 시에도 즉시 실패 — 호출자가 ensureOcrModels() 먼저.\r\n */\r\n\r\nimport type { InferenceSession } from \"onnxruntime-node\"\r\nimport { readFile } from \"fs/promises\"\r\nimport { join } from \"path\"\r\nimport {\r\n OCR_DET_MODEL,\r\n OCR_REC_MODEL,\r\n OCR_REC_DICT,\r\n getOcrModelsDir,\r\n parseCharacterDict,\r\n} from \"./models.js\"\r\n\r\n/** OCR 인식 결과 한 줄 — 좌표는 입력 이미지 픽셀 (top-left origin, y down) */\r\nexport interface OcrItem {\r\n text: string\r\n x: number\r\n y: number\r\n w: number\r\n h: number\r\n /** CTC 평균 신뢰도 0~1 */\r\n confidence: number\r\n}\r\n\r\n/** det 입력 긴 변 (공식 resize_long) */\r\nconst DET_LONG_SIDE = 960\r\nconst DET_THRESH = 0.3\r\nconst DET_BOX_THRESH = 0.6\r\nconst DET_UNCLIP_RATIO = 1.5\r\nconst DET_MIN_SIZE = 3\r\nconst DET_MAX_BOXES = 1000\r\nconst REC_HEIGHT = 48\r\nconst REC_MIN_WIDTH = 320\r\nconst REC_MAX_WIDTH = 3200\r\nconst TEXT_SCORE = 0.5\r\n\r\n// det: BGR 채널 순서에 yml 기재 순서 그대로 적용 (mean[0]→B)\r\nconst DET_MEAN = [0.485, 0.456, 0.406]\r\nconst DET_STD = [0.229, 0.224, 0.225]\r\n\r\ntype SharpFactory = (\r\n input: Uint8Array | Buffer,\r\n options?: { raw?: { width: number; height: number; channels: number } },\r\n) => SharpChain\r\ninterface SharpChain {\r\n extract(region: { left: number; top: number; width: number; height: number }): SharpChain\r\n resize(w: number, h: number, opts?: { fit?: string }): SharpChain\r\n removeAlpha(): SharpChain\r\n raw(): { toBuffer(): Promise<Buffer> }\r\n}\r\n\r\nexport class OcrEngine {\r\n private det: InferenceSession\r\n private rec: InferenceSession\r\n private dict: string[]\r\n private ort: typeof import(\"onnxruntime-node\")\r\n private sharp: SharpFactory\r\n\r\n private constructor(parts: {\r\n det: InferenceSession\r\n rec: InferenceSession\r\n dict: string[]\r\n ort: typeof import(\"onnxruntime-node\")\r\n sharp: SharpFactory\r\n }) {\r\n this.det = parts.det\r\n this.rec = parts.rec\r\n this.dict = parts.dict\r\n this.ort = parts.ort\r\n this.sharp = parts.sharp\r\n }\r\n\r\n static async create(): Promise<OcrEngine> {\r\n const [ortMod, sharpModRaw] = await Promise.all([\r\n tryImport<typeof import(\"onnxruntime-node\")>(\"onnxruntime-node\", () => import(\"onnxruntime-node\")),\r\n tryImport<{ default?: SharpFactory } & SharpFactory>(\r\n \"sharp\",\r\n () => import(\"sharp\") as unknown as Promise<{ default?: SharpFactory } & SharpFactory>,\r\n ),\r\n ])\r\n const sharpAny = sharpModRaw as { default?: SharpFactory } | SharpFactory\r\n const sharpMod: SharpFactory =\r\n typeof sharpAny === \"function\" ? sharpAny : (sharpAny.default ?? (sharpAny as unknown as SharpFactory))\r\n\r\n const dir = getOcrModelsDir()\r\n const sessionOpts: import(\"onnxruntime-node\").InferenceSession.SessionOptions = {\r\n graphOptimizationLevel: \"all\",\r\n executionProviders: [\"cpu\"],\r\n logSeverityLevel: 3, // paddle2onnx 변환 잔여물 W 로그 폭주 억제\r\n }\r\n const [det, rec, dictYml] = await Promise.all([\r\n ortMod.InferenceSession.create(join(dir, OCR_DET_MODEL.filename), sessionOpts),\r\n ortMod.InferenceSession.create(join(dir, OCR_REC_MODEL.filename), sessionOpts),\r\n readFile(join(dir, OCR_REC_DICT.filename), \"utf-8\"),\r\n ])\r\n const dict = parseCharacterDict(dictYml)\r\n if (dict.length === 0) throw new Error(\"OCR 사전 파싱 실패 — 모델 캐시를 삭제 후 재다운로드하세요\")\r\n\r\n return new OcrEngine({ det, rec, dict, ort: ortMod, sharp: sharpMod })\r\n }\r\n\r\n /** onnxruntime-node 1.14+ InferenceSession.release() — 구버전은 무시 */\r\n async destroy(): Promise<void> {\r\n for (const s of [this.det, this.rec]) {\r\n const rel = (s as unknown as { release?: () => Promise<void> }).release\r\n if (typeof rel === \"function\") {\r\n try { await rel.call(s) } catch { /* ignore */ }\r\n }\r\n }\r\n }\r\n\r\n /**\r\n * 페이지 RGBA 픽셀 → 텍스트 라인 인식.\r\n * 반환 좌표는 입력 픽셀 기준. 라인은 위→아래, 좌→우 정렬.\r\n */\r\n async recognizePage(rgba: Uint8Array, width: number, height: number): Promise<OcrItem[]> {\r\n if (width < DET_MIN_SIZE || height < DET_MIN_SIZE) return []\r\n const boxes = await this.detect(rgba, width, height)\r\n const items: OcrItem[] = []\r\n for (const b of boxes) {\r\n const r = await this.recognizeLine(rgba, width, height, b)\r\n if (r && r.confidence >= TEXT_SCORE && r.text.trim()) items.push(r)\r\n }\r\n items.sort((a, b) => (a.y - b.y) || (a.x - b.x))\r\n return items\r\n }\r\n\r\n // ─── det ─────────────────────────────────────────────\r\n\r\n private async detect(\r\n rgba: Uint8Array,\r\n width: number,\r\n height: number,\r\n ): Promise<Array<{ x: number; y: number; w: number; h: number }>> {\r\n const ratio = DET_LONG_SIDE / Math.max(width, height)\r\n const dw = Math.max(32, Math.round((width * ratio) / 32) * 32)\r\n const dh = Math.max(32, Math.round((height * ratio) / 32) * 32)\r\n\r\n const rgb = await this.sharp(rgba, { raw: { width, height, channels: 4 } })\r\n .resize(dw, dh, { fit: \"fill\" })\r\n .removeAlpha()\r\n .raw()\r\n .toBuffer()\r\n\r\n // HWC RGB → CHW BGR float32 정규화\r\n const plane = dw * dh\r\n const input = new Float32Array(3 * plane)\r\n for (let i = 0; i < plane; i++) {\r\n const r = rgb[i * 3] / 255\r\n const g = rgb[i * 3 + 1] / 255\r\n const b = rgb[i * 3 + 2] / 255\r\n input[i] = (b - DET_MEAN[0]) / DET_STD[0]\r\n input[plane + i] = (g - DET_MEAN[1]) / DET_STD[1]\r\n input[2 * plane + i] = (r - DET_MEAN[2]) / DET_STD[2]\r\n }\r\n\r\n const tensor = new this.ort.Tensor(\"float32\", input, [1, 3, dh, dw])\r\n const out = await this.det.run({ [this.det.inputNames[0]]: tensor })\r\n const probMap = out[this.det.outputNames[0]].data as Float32Array\r\n\r\n const rawBoxes = componentBoxes(probMap, dw, dh)\r\n const sx = width / dw\r\n const sy = height / dh\r\n const boxes: Array<{ x: number; y: number; w: number; h: number }> = []\r\n for (const rb of rawBoxes.slice(0, DET_MAX_BOXES)) {\r\n // unclip: DB 는 학습 시 텍스트 영역을 수축시키므로 검출 박스를 되팽창\r\n const bw = rb.x2 - rb.x1 + 1\r\n const bh = rb.y2 - rb.y1 + 1\r\n const delta = (bw * bh * DET_UNCLIP_RATIO) / (2 * (bw + bh))\r\n const x1 = Math.max(0, Math.floor((rb.x1 - delta) * sx))\r\n const y1 = Math.max(0, Math.floor((rb.y1 - delta) * sy))\r\n const x2 = Math.min(width, Math.ceil((rb.x2 + 1 + delta) * sx))\r\n const y2 = Math.min(height, Math.ceil((rb.y2 + 1 + delta) * sy))\r\n if (x2 - x1 < DET_MIN_SIZE || y2 - y1 < DET_MIN_SIZE) continue\r\n boxes.push({ x: x1, y: y1, w: x2 - x1, h: y2 - y1 })\r\n }\r\n return boxes\r\n }\r\n\r\n // ─── rec ─────────────────────────────────────────────\r\n\r\n private async recognizeLine(\r\n rgba: Uint8Array,\r\n width: number,\r\n height: number,\r\n box: { x: number; y: number; w: number; h: number },\r\n ): Promise<OcrItem | null> {\r\n const rw = Math.min(REC_MAX_WIDTH, Math.max(16, Math.round((box.w * REC_HEIGHT) / box.h)))\r\n const padded = Math.max(REC_MIN_WIDTH, rw)\r\n\r\n const rgb = await this.sharp(rgba, { raw: { width, height, channels: 4 } })\r\n .extract({ left: box.x, top: box.y, width: box.w, height: box.h })\r\n .resize(rw, REC_HEIGHT, { fit: \"fill\" })\r\n .removeAlpha()\r\n .raw()\r\n .toBuffer()\r\n\r\n // HWC RGB → CHW BGR, (x/255-0.5)/0.5, 우측 zero-pad\r\n const plane = padded * REC_HEIGHT\r\n const input = new Float32Array(3 * plane) // pad 영역은 0 (=정규화 후 회색 아님 주의: PP 도 0 pad)\r\n for (let y = 0; y < REC_HEIGHT; y++) {\r\n for (let x = 0; x < rw; x++) {\r\n const src = (y * rw + x) * 3\r\n const dst = y * padded + x\r\n input[dst] = rgb[src + 2] / 255 / 0.5 - 1 // B\r\n input[plane + dst] = rgb[src + 1] / 255 / 0.5 - 1 // G\r\n input[2 * plane + dst] = rgb[src] / 255 / 0.5 - 1 // R\r\n }\r\n }\r\n\r\n const tensor = new this.ort.Tensor(\"float32\", input, [1, 3, REC_HEIGHT, padded])\r\n const out = await this.rec.run({ [this.rec.inputNames[0]]: tensor })\r\n const logits = out[this.rec.outputNames[0]]\r\n const [, T, C] = logits.dims as number[]\r\n const data = logits.data as Float32Array\r\n\r\n const decoded = ctcDecode(data, T, C, this.dict)\r\n if (!decoded) return null\r\n return { text: decoded.text, x: box.x, y: box.y, w: box.w, h: box.h, confidence: decoded.confidence }\r\n }\r\n}\r\n\r\n/** CTC greedy 디코드 — 연속 중복 붕괴 → blank(0) 제거 → 사전 매핑 (테스트용 export) */\r\nexport function ctcDecode(\r\n data: Float32Array,\r\n T: number,\r\n C: number,\r\n dict: string[],\r\n): { text: string; confidence: number } | null {\r\n let text = \"\"\r\n let confSum = 0\r\n let confCount = 0\r\n let prev = -1\r\n for (let t = 0; t < T; t++) {\r\n const off = t * C\r\n let best = 0\r\n let bestV = data[off]\r\n for (let c = 1; c < C; c++) {\r\n const v = data[off + c]\r\n if (v > bestV) { bestV = v; best = c }\r\n }\r\n const repeat = best === prev\r\n prev = best\r\n if (best === 0 || repeat) continue\r\n // 모델 출력이 softmax 확률이 아니면 (>1) 해당 스텝만 정규화\r\n let p = bestV\r\n if (p > 1.0001 || p < 0) {\r\n let denom = 0\r\n for (let c = 0; c < C; c++) denom += Math.exp(data[off + c] - bestV)\r\n p = 1 / denom\r\n }\r\n confSum += p\r\n confCount++\r\n if (best >= 1 && best <= dict.length) text += dict[best - 1]\r\n else if (best === dict.length + 1) text += \" \"\r\n }\r\n if (!text) return null\r\n return { text, confidence: confCount > 0 ? confSum / confCount : 0 }\r\n}\r\n\r\n/** 이진화 확률맵의 4-연결 성분 bbox (score = 성분 평균 확률, 테스트용 export) */\r\nexport function componentBoxes(\r\n prob: Float32Array,\r\n w: number,\r\n h: number,\r\n): Array<{ x1: number; y1: number; x2: number; y2: number }> {\r\n const visited = new Uint8Array(w * h)\r\n const boxes: Array<{ x1: number; y1: number; x2: number; y2: number; score: number }> = []\r\n const stack: number[] = []\r\n\r\n for (let start = 0; start < w * h; start++) {\r\n if (visited[start] || prob[start] <= DET_THRESH) continue\r\n let x1 = start % w, x2 = x1, y1 = (start / w) | 0, y2 = y1\r\n let sum = 0\r\n let count = 0\r\n stack.length = 0\r\n stack.push(start)\r\n visited[start] = 1\r\n while (stack.length) {\r\n const p = stack.pop()!\r\n const px = p % w\r\n const py = (p / w) | 0\r\n sum += prob[p]\r\n count++\r\n if (px < x1) x1 = px\r\n if (px > x2) x2 = px\r\n if (py < y1) y1 = py\r\n if (py > y2) y2 = py\r\n // 4-이웃\r\n if (px > 0 && !visited[p - 1] && prob[p - 1] > DET_THRESH) { visited[p - 1] = 1; stack.push(p - 1) }\r\n if (px < w - 1 && !visited[p + 1] && prob[p + 1] > DET_THRESH) { visited[p + 1] = 1; stack.push(p + 1) }\r\n if (py > 0 && !visited[p - w] && prob[p - w] > DET_THRESH) { visited[p - w] = 1; stack.push(p - w) }\r\n if (py < h - 1 && !visited[p + w] && prob[p + w] > DET_THRESH) { visited[p + w] = 1; stack.push(p + w) }\r\n }\r\n if (x2 - x1 + 1 < DET_MIN_SIZE && y2 - y1 + 1 < DET_MIN_SIZE) continue\r\n boxes.push({ x1, y1, x2, y2, score: sum / count })\r\n }\r\n\r\n return boxes\r\n .filter(b => b.score >= DET_BOX_THRESH)\r\n .sort((a, b) => (a.y1 - b.y1) || (a.x1 - b.x1))\r\n}\r\n\r\nasync function tryImport<T>(name: string, loader: () => Promise<T>): Promise<T> {\r\n try {\r\n return await loader()\r\n } catch (e) {\r\n throw new Error(\r\n `내장 OCR 을 사용하려면 optional dependency '${name}' 이 필요합니다. ` +\r\n `\\`npm install ${name}\\` 후 다시 실행하세요. 원인: ${(e as Error).message}`,\r\n )\r\n }\r\n}\r\n\r\n// ─── 엔진 싱글턴 (watch/서버 장기 실행에서 세션 재사용) ───\r\nlet enginePromise: Promise<OcrEngine> | null = null\r\n\r\nexport function getOcrEngine(): Promise<OcrEngine> {\r\n if (!enginePromise) {\r\n enginePromise = OcrEngine.create().catch(err => {\r\n enginePromise = null // 실패는 캐시하지 않음 — 모델 설치 후 재시도 가능\r\n throw err\r\n })\r\n }\r\n return enginePromise\r\n}\r\n","/**\r\n * PDF OCR 브릿지 — 대상 페이지를 pdfium 으로 래스터해 내장 엔진 또는\r\n * 사용자 OcrProvider 로 인식하고, 페이지별 IRBlock[] 을 돌려준다.\r\n *\r\n * - 내장 엔진: OcrItem 좌표를 PDF 포인트로 환산해 기존 블록 파이프라인\r\n * (extractPageBlocksWithLines — xy-cut 읽기순서·클러스터 표 감지)에 태운다.\r\n * 스캔 문서도 표 구조 복원이 가능한 이유.\r\n * - 사용자 프로바이더: 페이지 PNG → 텍스트 → 페이지당 paragraph 블록 (종전 계약).\r\n *\r\n * 에러 계약 (v4.1.0 리뷰 F6/F7):\r\n * - 환경 오류(의존성·모델 미설치, PDF 오픈 실패)는 throw — 호출자가 NEEDS_OCR 폴백.\r\n * - 페이지 단위 실패는 본문 오염 대신 warnings(OCR_FAILED) 로 기록하고 계속.\r\n */\r\n\r\nimport type { IRBlock, OcrProvider, ParseWarning } from \"../types.js\"\r\nimport type { NormItem } from \"../pdf/text-line.js\"\r\nimport type { LineSegment } from \"../pdf/line-types.js\"\r\nimport { extractPageBlocksWithLines } from \"../pdf/page-blocks.js\"\r\nimport { detectRulingLines, rulingToPdfLines } from \"./ruling-lines.js\"\r\nimport { getOcrEngine, type OcrItem } from \"./engine.js\"\r\nimport { ensureOcrModels } from \"./models.js\"\r\n\r\n/** OCR 렌더 스케일 (72dpi × 3 = 216dpi) — 10pt 본문이 rec 입력 높이(48px)에 근접 */\r\nconst OCR_RENDER_SCALE = 3\r\n/** 페이지 하나당 OCR 타임아웃 — det+rec 수십 라인 기준 넉넉히 */\r\nconst PAGE_TIMEOUT_MS = 120_000\r\n\r\nexport type OcrMode = \"builtin\" | OcrProvider\r\n\r\n/**\r\n * 대상 페이지들을 OCR 해 페이지별 블록 맵 반환.\r\n * @param buffer 원본 PDF (pdfjs 가 detach 하기 전에 clone 해 둔 것)\r\n * @param targets 1-based 페이지 번호 집합\r\n */\r\nexport async function runPdfOcr(\r\n buffer: ArrayBuffer,\r\n targets: Set<number>,\r\n mode: OcrMode,\r\n warnings: ParseWarning[],\r\n onProgress?: (current: number, total: number) => void,\r\n): Promise<Map<number, IRBlock[]>> {\r\n const result = new Map<number, IRBlock[]>()\r\n if (targets.size === 0) return result\r\n\r\n // 환경 준비 — 실패는 그대로 throw (호출자가 NEEDS_OCR 폴백)\r\n const pdfiumMod = await tryImport<typeof import(\"@hyzyla/pdfium\")>(\r\n \"@hyzyla/pdfium\",\r\n () => import(\"@hyzyla/pdfium\"),\r\n )\r\n if (mode === \"builtin\") {\r\n await ensureOcrModels(p => {\r\n if (p.phase === \"download\" && p.downloaded === 0) {\r\n process.stderr.write(`[kordoc-ocr] ${p.spec.name} 다운로드 중 (~${p.spec.sizeMb}MB)...\\n`)\r\n }\r\n })\r\n }\r\n const engine = mode === \"builtin\" ? await getOcrEngine() : null\r\n\r\n const pdfium = await pdfiumMod.PDFiumLibrary.init()\r\n const doc = await pdfium.loadDocument(new Uint8Array(buffer))\r\n try {\r\n let done = 0\r\n for (const page of doc.pages()) {\r\n // pdfium page.number 는 0-based pageIndex — 대외 계약(1-based)으로 환산\r\n const pageNo = page.number + 1\r\n if (!targets.has(pageNo)) continue\r\n onProgress?.(++done, targets.size)\r\n try {\r\n const blocks = await withTimeout(\r\n ocrOnePage(page, pageNo, mode, engine, warnings),\r\n PAGE_TIMEOUT_MS,\r\n `OCR 페이지 ${pageNo} 타임아웃 (${PAGE_TIMEOUT_MS / 1000}초)`,\r\n )\r\n result.set(pageNo, blocks)\r\n } catch (e) {\r\n warnings.push({\r\n page: pageNo,\r\n message: `페이지 ${pageNo} OCR 실패: ${e instanceof Error ? e.message : String(e)}`,\r\n code: \"OCR_FAILED\",\r\n })\r\n }\r\n }\r\n } finally {\r\n doc.destroy()\r\n pdfium.destroy()\r\n }\r\n return result\r\n}\r\n\r\nasync function ocrOnePage(\r\n page: import(\"@hyzyla/pdfium\").PDFiumPage,\r\n pageNo: number,\r\n mode: OcrMode,\r\n engine: Awaited<ReturnType<typeof getOcrEngine>> | null,\r\n warnings: ParseWarning[],\r\n): Promise<IRBlock[]> {\r\n const { originalWidth: pdfW, originalHeight: pdfH } = page.getOriginalSize()\r\n const rendered = await page.render({\r\n scale: OCR_RENDER_SCALE,\r\n render: async ({ data }) => data,\r\n })\r\n const { data: bgra, width: rw, height: rh } = rendered\r\n const rgba = bgraToRgba(bgra)\r\n\r\n if (mode === \"builtin\") {\r\n const items = await engine!.recognizePage(rgba, rw, rh)\r\n // 래스터에서 표 괘선 감지 — 스캔본 병합셀 서식도 선 기반 표 파이프라인을 탄다\r\n const scale = rh / pdfH\r\n const ruling = detectRulingLines(rgba, rw, rh, scale)\r\n const extraLines = rulingToPdfLines(ruling, scale, pdfH)\r\n return ocrItemsToBlocks(items, pageNo, pdfW, pdfH, scale, extraLines)\r\n }\r\n\r\n // 사용자 프로바이더 — PNG 인코딩 후 호출, 페이지당 paragraph (종전 계약)\r\n type SharpPngFactory = (\r\n input: Uint8Array,\r\n opts: { raw: { width: number; height: number; channels: number } },\r\n ) => { png(): { toBuffer(): Promise<Buffer> } }\r\n const sharpModRaw = await tryImport<{ default?: SharpPngFactory } & SharpPngFactory>(\r\n \"sharp\",\r\n () => import(\"sharp\") as unknown as Promise<{ default?: SharpPngFactory } & SharpPngFactory>,\r\n )\r\n const sharpAny = sharpModRaw as { default?: SharpPngFactory } | SharpPngFactory\r\n const sharp: SharpPngFactory =\r\n typeof sharpAny === \"function\" ? sharpAny : (sharpAny.default ?? (sharpAny as unknown as SharpPngFactory))\r\n const png = await sharp(rgba, { raw: { width: rw, height: rh, channels: 4 } })\r\n .png()\r\n .toBuffer()\r\n\r\n const text = await mode(new Uint8Array(png), pageNo, \"image/png\")\r\n if (!text.trim()) {\r\n warnings.push({ page: pageNo, message: `페이지 ${pageNo} OCR 결과 없음`, code: \"OCR_FAILED\" })\r\n return []\r\n }\r\n return [{ type: \"paragraph\", text: text.trim(), pageNumber: pageNo }]\r\n}\r\n\r\n/**\r\n * OcrItem(렌더 픽셀, top-left origin) → NormItem(PDF pt, bottom-up baseline)\r\n * 으로 환산해 기존 블록 파이프라인에 태운다. 괘선은 그래픽 ops 대신\r\n * 래스터 감지 결과(extraLines)로 공급 — 없으면 클러스터 감지기 몫이다.\r\n * (이미지 직접 입력 경로 image-ocr.ts 와 공유)\r\n */\r\nexport function ocrItemsToBlocks(\r\n items: OcrItem[],\r\n pageNumber: number,\r\n pdfW: number,\r\n pdfH: number,\r\n scale: number,\r\n extraLines?: { horizontals: LineSegment[]; verticals: LineSegment[] },\r\n): IRBlock[] {\r\n const norm: NormItem[] = items.map(it => {\r\n const h = it.h / scale\r\n return {\r\n text: it.text,\r\n x: Math.round(it.x / scale),\r\n // NormItem.y 는 pdfjs transform[5] = 베이스라인 (bottom-up) — 박스 하단으로 근사\r\n y: Math.round(pdfH - (it.y + it.h) / scale),\r\n w: Math.round(it.w / scale),\r\n h: Math.round(h),\r\n fontSize: Math.max(1, Math.round(h * 0.8)), // 박스높이는 어센더+디센더 포함 — 글자크기 근사\r\n fontName: \"ocr\",\r\n isHidden: false,\r\n }\r\n })\r\n return extractPageBlocksWithLines(norm, pageNumber, { fnArray: [], argsArray: [] }, pdfW, pdfH, extraLines)\r\n}\r\n\r\nasync function tryImport<T>(name: string, loader: () => Promise<T>): Promise<T> {\r\n try {\r\n return await loader()\r\n } catch (e) {\r\n throw new Error(\r\n `OCR 을 사용하려면 optional dependency '${name}' 이 필요합니다. ` +\r\n `\\`npm install ${name}\\` 후 다시 실행하세요. 원인: ${(e as Error).message}`,\r\n )\r\n }\r\n}\r\n\r\nasync function withTimeout<T>(promise: Promise<T>, ms: number, msg: string): Promise<T> {\r\n // 타임아웃 패배 후 원 promise 의 사후 reject 가 unhandled rejection 이 되지 않게 흡수\r\n promise.catch(() => {})\r\n let timer: NodeJS.Timeout | undefined\r\n try {\r\n return await Promise.race([\r\n promise,\r\n new Promise<never>((_, reject) => {\r\n timer = setTimeout(() => reject(new Error(msg)), ms)\r\n }),\r\n ])\r\n } finally {\r\n if (timer) clearTimeout(timer)\r\n }\r\n}\r\n\r\n/** pdfium BGRA → RGBA */\r\nfunction bgraToRgba(bgra: Uint8Array): Uint8Array {\r\n const out = new Uint8Array(bgra.length)\r\n for (let i = 0; i < bgra.length; i += 4) {\r\n out[i] = bgra[i + 2]\r\n out[i + 1] = bgra[i + 1]\r\n out[i + 2] = bgra[i]\r\n out[i + 3] = bgra[i + 3]\r\n }\r\n return out\r\n}\r\n"],"mappings":";;;;;;;;;;;;;;AAkBA,IAAM,eAAe;AAErB,IAAM,qBAAqB;AAE3B,IAAM,oBAAoB;AAE1B,IAAM,aAAa;AAEnB,IAAM,qBAAqB;AAE3B,IAAM,qBAAqB;AAkBpB,SAAS,kBACd,MACA,OACA,QACA,OACa;AACb,QAAM,WAAW,KAAK,MAAM,qBAAqB,KAAK;AACtD,QAAM,aAAa,KAAK,IAAI,GAAG,KAAK,MAAM,oBAAoB,KAAK,CAAC;AAGpE,QAAM,MAAM,IAAI,WAAW,QAAQ,MAAM;AACzC,WAAS,IAAI,GAAG,IAAI,GAAG,IAAI,IAAI,QAAQ,KAAK,KAAK,GAAG;AAElD,UAAM,OAAQ,KAAK,CAAC,IAAI,KAAK,KAAK,IAAI,CAAC,IAAI,MAAM,KAAK,IAAI,CAAC,IAAI,MAAO;AACtE,QAAI,QAAQ,gBAAgB,KAAK,IAAI,CAAC,KAAK,IAAK,KAAI,CAAC,IAAI;AAAA,EAC3D;AAEA,QAAM,cAAc,YAAY,KAAK,OAAO,QAAQ,UAAU,YAAY,KAAK;AAC/E,QAAM,YAAY,YAAY,KAAK,OAAO,QAAQ,UAAU,YAAY,IAAI;AAC5E,SAAO,EAAE,aAAa,UAAU;AAClC;AAcA,SAAS,YACP,KACA,OACA,QACA,UACA,YACA,WACa;AACb,QAAM,QAAQ,YAAY,QAAQ;AAClC,QAAM,OAAO,YAAY,SAAS;AAClC,QAAM,KAAK,YACP,CAAC,MAAc,QAAgB,IAAI,MAAM,QAAQ,IAAI,IACrD,CAAC,MAAc,QAAgB,IAAI,OAAO,QAAQ,GAAG;AAEzD,QAAM,MAAmB,CAAC;AAC1B,MAAI,OAAe,CAAC;AAGpB,QAAM,mBAAmB,CAAC,GAAS,SAAyB;AAC1D,QAAI,OAAO,KAAK,QAAQ,MAAO,QAAO;AACtC,QAAI,OAAO,GAAG,QAAQ;AACtB,aAAS,MAAM,EAAE,IAAI,OAAO,EAAE,IAAI,OAAO,GAAG;AAC1C;AACA,UAAI,GAAG,MAAM,GAAG,MAAM,EAAG;AAAA,IAC3B;AACA,WAAO,QAAQ,IAAI,OAAO,QAAQ;AAAA,EACpC;AAEA,QAAM,OAAO,CAAC,MAAY;AACxB,UAAM,QAAQ,EAAE,OAAO,EAAE,QAAQ;AACjC,QAAI,QAAQ,WAAY;AACxB,QACE,iBAAiB,GAAG,EAAE,QAAQ,CAAC,KAAK,sBACpC,iBAAiB,GAAG,EAAE,OAAO,CAAC,KAAK,mBACnC;AACF,UAAM,UAAU,EAAE,QAAQ,EAAE,QAAQ;AACpC,QAAI;AAAA,MACF,YACI,EAAE,IAAI,QAAQ,IAAI,EAAE,IAAI,IAAI,QAAQ,IAAI,EAAE,IAAI,aAAa,MAAM,IACjE,EAAE,IAAI,EAAE,IAAI,IAAI,QAAQ,IAAI,EAAE,IAAI,IAAI,QAAQ,aAAa,MAAM;AAAA,IACvE;AAAA,EACF;AAEA,WAAS,OAAO,GAAG,OAAO,OAAO,QAAQ;AAEvC,UAAM,OAA0C,CAAC;AACjD,QAAI,WAAW;AACf,QAAI,MAAM;AACV,aAAS,MAAM,GAAG,OAAO,MAAM,OAAO;AACpC,YAAM,KAAK,MAAM,QAAQ,GAAG,MAAM,GAAG,MAAM;AAC3C,UAAI,IAAI;AACN,YAAI,WAAW,EAAG,YAAW;AAC7B,cAAM;AAAA,MACR,WAAW,YAAY,GAAG;AACxB,YAAI,EAAE,MAAM,cAAc,OAAO,MAAM;AACrC,gBAAM,KAAK,MAAM;AACjB,cAAI,KAAK,WAAW,KAAK,SAAU,MAAK,KAAK,EAAE,IAAI,UAAU,GAAG,CAAC;AACjE,qBAAW;AACX,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAGA,UAAM,OAAe,CAAC;AACtB,UAAM,OAAO,oBAAI,IAAY;AAC7B,eAAW,QAAQ,MAAM;AACvB,UAAI,KAAK,SAAS,OAAO,GAAG;AAAE,aAAK,IAAI;AAAG;AAAA,MAAS;AACnD,UAAI,SAAS;AACb,eAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,YAAI,KAAK,IAAI,CAAC,EAAG;AACjB,cAAM,MAAM,KAAK,CAAC;AAClB,cAAM,UAAU,KAAK,IAAI,KAAK,IAAI,IAAI,EAAE,IAAI,KAAK,IAAI,KAAK,IAAI,IAAI,EAAE,IAAI;AACxE,cAAM,UAAU,KAAK,IAAI,KAAK,KAAK,KAAK,IAAI,IAAI,KAAK,IAAI,EAAE,IAAI;AAC/D,YAAI,WAAW,UAAU,oBAAoB;AAC3C,eAAK,KAAK,KAAK,IAAI,KAAK,IAAI,IAAI,EAAE;AAClC,eAAK,KAAK,KAAK,IAAI,KAAK,IAAI,IAAI,EAAE;AAClC,eAAK,OAAO;AACZ,eAAK,KAAK,IAAI;AACd,eAAK,IAAI,CAAC;AACV,mBAAS;AACT;AAAA,QACF;AAAA,MACF;AACA,UAAI,CAAC,OAAQ,MAAK,IAAI;AAAA,IACxB;AACA,aAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,UAAI,CAAC,KAAK,IAAI,CAAC,EAAG,MAAK,KAAK,EAAE,IAAI,KAAK,CAAC,EAAE,IAAI,IAAI,KAAK,CAAC,EAAE,IAAI,OAAO,MAAM,MAAM,KAAK,CAAC;AAAA,IACzF;AACA,WAAO;AAAA,EACT;AACA,aAAW,QAAQ,KAAM,MAAK,IAAI;AAClC,SAAO;AACT;AAMO,SAAS,iBACd,QACA,OACA,WAC0D;AAC1D,QAAM,cAA6B,OAAO,YAAY,IAAI,OAAK;AAC7D,UAAM,IAAI,aAAa,EAAE,KAAK,EAAE,MAAM,IAAI;AAC1C,WAAO;AAAA,MACL,IAAI,KAAK,IAAI,EAAE,IAAI,EAAE,EAAE,IAAI;AAAA,MAC3B,IAAI;AAAA,MACJ,IAAI,KAAK,IAAI,EAAE,IAAI,EAAE,EAAE,IAAI;AAAA,MAC3B,IAAI;AAAA,MACJ,WAAW,EAAE,cAAc;AAAA,IAC7B;AAAA,EACF,CAAC;AACD,QAAM,YAA2B,OAAO,UAAU,IAAI,OAAK;AACzD,UAAM,KAAK,EAAE,KAAK,EAAE,MAAM,IAAI;AAC9B,UAAM,KAAK,YAAY,KAAK,IAAI,EAAE,IAAI,EAAE,EAAE,IAAI;AAC9C,UAAM,KAAK,YAAY,KAAK,IAAI,EAAE,IAAI,EAAE,EAAE,IAAI;AAC9C,WAAO,EAAE,IAAI,GAAG,IAAI,IAAI,IAAI,GAAG,IAAI,IAAI,WAAW,EAAE,cAAc,MAAM;AAAA,EAC1E,CAAC;AACD,SAAO,EAAE,aAAa,UAAU;AAClC;;;ACtLA,SAAS,gBAAgB;AACzB,SAAS,YAAY;AAqBrB,IAAM,gBAAgB;AACtB,IAAM,aAAa;AACnB,IAAM,iBAAiB;AACvB,IAAM,mBAAmB;AACzB,IAAM,eAAe;AACrB,IAAM,gBAAgB;AACtB,IAAM,aAAa;AACnB,IAAM,gBAAgB;AACtB,IAAM,gBAAgB;AACtB,IAAM,aAAa;AAGnB,IAAM,WAAW,CAAC,OAAO,OAAO,KAAK;AACrC,IAAM,UAAU,CAAC,OAAO,OAAO,KAAK;AAa7B,IAAM,YAAN,MAAM,WAAU;AAAA,EACb;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA,YAAY,OAMjB;AACD,SAAK,MAAM,MAAM;AACjB,SAAK,MAAM,MAAM;AACjB,SAAK,OAAO,MAAM;AAClB,SAAK,MAAM,MAAM;AACjB,SAAK,QAAQ,MAAM;AAAA,EACrB;AAAA,EAEA,aAAa,SAA6B;AACxC,UAAM,CAAC,QAAQ,WAAW,IAAI,MAAM,QAAQ,IAAI;AAAA,MAC9C,UAA6C,oBAAoB,MAAM,OAAO,kBAAkB,CAAC;AAAA,MACjG;AAAA,QACE;AAAA,QACA,MAAM,OAAO,OAAO;AAAA,MACtB;AAAA,IACF,CAAC;AACD,UAAM,WAAW;AACjB,UAAM,WACJ,OAAO,aAAa,aAAa,WAAY,SAAS,WAAY;AAEpE,UAAM,MAAM,gBAAgB;AAC5B,UAAM,cAA0E;AAAA,MAC9E,wBAAwB;AAAA,MACxB,oBAAoB,CAAC,KAAK;AAAA,MAC1B,kBAAkB;AAAA;AAAA,IACpB;AACA,UAAM,CAAC,KAAK,KAAK,OAAO,IAAI,MAAM,QAAQ,IAAI;AAAA,MAC5C,OAAO,iBAAiB,OAAO,KAAK,KAAK,cAAc,QAAQ,GAAG,WAAW;AAAA,MAC7E,OAAO,iBAAiB,OAAO,KAAK,KAAK,cAAc,QAAQ,GAAG,WAAW;AAAA,MAC7E,SAAS,KAAK,KAAK,aAAa,QAAQ,GAAG,OAAO;AAAA,IACpD,CAAC;AACD,UAAM,OAAO,mBAAmB,OAAO;AACvC,QAAI,KAAK,WAAW,EAAG,OAAM,IAAI,MAAM,wJAAqC;AAE5E,WAAO,IAAI,WAAU,EAAE,KAAK,KAAK,MAAM,KAAK,QAAQ,OAAO,SAAS,CAAC;AAAA,EACvE;AAAA;AAAA,EAGA,MAAM,UAAyB;AAC7B,eAAW,KAAK,CAAC,KAAK,KAAK,KAAK,GAAG,GAAG;AACpC,YAAM,MAAO,EAAmD;AAChE,UAAI,OAAO,QAAQ,YAAY;AAC7B,YAAI;AAAE,gBAAM,IAAI,KAAK,CAAC;AAAA,QAAE,QAAQ;AAAA,QAAe;AAAA,MACjD;AAAA,IACF;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,MAAM,cAAc,MAAkB,OAAe,QAAoC;AACvF,QAAI,QAAQ,gBAAgB,SAAS,aAAc,QAAO,CAAC;AAC3D,UAAM,QAAQ,MAAM,KAAK,OAAO,MAAM,OAAO,MAAM;AACnD,UAAM,QAAmB,CAAC;AAC1B,eAAW,KAAK,OAAO;AACrB,YAAM,IAAI,MAAM,KAAK,cAAc,MAAM,OAAO,QAAQ,CAAC;AACzD,UAAI,KAAK,EAAE,cAAc,cAAc,EAAE,KAAK,KAAK,EAAG,OAAM,KAAK,CAAC;AAAA,IACpE;AACA,UAAM,KAAK,CAAC,GAAG,MAAO,EAAE,IAAI,EAAE,KAAO,EAAE,IAAI,EAAE,CAAE;AAC/C,WAAO;AAAA,EACT;AAAA;AAAA,EAIA,MAAc,OACZ,MACA,OACA,QACgE;AAChE,UAAM,QAAQ,gBAAgB,KAAK,IAAI,OAAO,MAAM;AACpD,UAAM,KAAK,KAAK,IAAI,IAAI,KAAK,MAAO,QAAQ,QAAS,EAAE,IAAI,EAAE;AAC7D,UAAM,KAAK,KAAK,IAAI,IAAI,KAAK,MAAO,SAAS,QAAS,EAAE,IAAI,EAAE;AAE9D,UAAM,MAAM,MAAM,KAAK,MAAM,MAAM,EAAE,KAAK,EAAE,OAAO,QAAQ,UAAU,EAAE,EAAE,CAAC,EACvE,OAAO,IAAI,IAAI,EAAE,KAAK,OAAO,CAAC,EAC9B,YAAY,EACZ,IAAI,EACJ,SAAS;AAGZ,UAAM,QAAQ,KAAK;AACnB,UAAM,QAAQ,IAAI,aAAa,IAAI,KAAK;AACxC,aAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,YAAM,IAAI,IAAI,IAAI,CAAC,IAAI;AACvB,YAAM,IAAI,IAAI,IAAI,IAAI,CAAC,IAAI;AAC3B,YAAM,IAAI,IAAI,IAAI,IAAI,CAAC,IAAI;AAC3B,YAAM,CAAC,KAAK,IAAI,SAAS,CAAC,KAAK,QAAQ,CAAC;AACxC,YAAM,QAAQ,CAAC,KAAK,IAAI,SAAS,CAAC,KAAK,QAAQ,CAAC;AAChD,YAAM,IAAI,QAAQ,CAAC,KAAK,IAAI,SAAS,CAAC,KAAK,QAAQ,CAAC;AAAA,IACtD;AAEA,UAAM,SAAS,IAAI,KAAK,IAAI,OAAO,WAAW,OAAO,CAAC,GAAG,GAAG,IAAI,EAAE,CAAC;AACnE,UAAM,MAAM,MAAM,KAAK,IAAI,IAAI,EAAE,CAAC,KAAK,IAAI,WAAW,CAAC,CAAC,GAAG,OAAO,CAAC;AACnE,UAAM,UAAU,IAAI,KAAK,IAAI,YAAY,CAAC,CAAC,EAAE;AAE7C,UAAM,WAAW,eAAe,SAAS,IAAI,EAAE;AAC/C,UAAM,KAAK,QAAQ;AACnB,UAAM,KAAK,SAAS;AACpB,UAAM,QAA+D,CAAC;AACtE,eAAW,MAAM,SAAS,MAAM,GAAG,aAAa,GAAG;AAEjD,YAAM,KAAK,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAM,KAAK,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAM,QAAS,KAAK,KAAK,oBAAqB,KAAK,KAAK;AACxD,YAAM,KAAK,KAAK,IAAI,GAAG,KAAK,OAAO,GAAG,KAAK,SAAS,EAAE,CAAC;AACvD,YAAM,KAAK,KAAK,IAAI,GAAG,KAAK,OAAO,GAAG,KAAK,SAAS,EAAE,CAAC;AACvD,YAAM,KAAK,KAAK,IAAI,OAAO,KAAK,MAAM,GAAG,KAAK,IAAI,SAAS,EAAE,CAAC;AAC9D,YAAM,KAAK,KAAK,IAAI,QAAQ,KAAK,MAAM,GAAG,KAAK,IAAI,SAAS,EAAE,CAAC;AAC/D,UAAI,KAAK,KAAK,gBAAgB,KAAK,KAAK,aAAc;AACtD,YAAM,KAAK,EAAE,GAAG,IAAI,GAAG,IAAI,GAAG,KAAK,IAAI,GAAG,KAAK,GAAG,CAAC;AAAA,IACrD;AACA,WAAO;AAAA,EACT;AAAA;AAAA,EAIA,MAAc,cACZ,MACA,OACA,QACA,KACyB;AACzB,UAAM,KAAK,KAAK,IAAI,eAAe,KAAK,IAAI,IAAI,KAAK,MAAO,IAAI,IAAI,aAAc,IAAI,CAAC,CAAC,CAAC;AACzF,UAAM,SAAS,KAAK,IAAI,eAAe,EAAE;AAEzC,UAAM,MAAM,MAAM,KAAK,MAAM,MAAM,EAAE,KAAK,EAAE,OAAO,QAAQ,UAAU,EAAE,EAAE,CAAC,EACvE,QAAQ,EAAE,MAAM,IAAI,GAAG,KAAK,IAAI,GAAG,OAAO,IAAI,GAAG,QAAQ,IAAI,EAAE,CAAC,EAChE,OAAO,IAAI,YAAY,EAAE,KAAK,OAAO,CAAC,EACtC,YAAY,EACZ,IAAI,EACJ,SAAS;AAGZ,UAAM,QAAQ,SAAS;AACvB,UAAM,QAAQ,IAAI,aAAa,IAAI,KAAK;AACxC,aAAS,IAAI,GAAG,IAAI,YAAY,KAAK;AACnC,eAAS,IAAI,GAAG,IAAI,IAAI,KAAK;AAC3B,cAAM,OAAO,IAAI,KAAK,KAAK;AAC3B,cAAM,MAAM,IAAI,SAAS;AACzB,cAAM,GAAG,IAAI,IAAI,MAAM,CAAC,IAAI,MAAM,MAAM;AACxC,cAAM,QAAQ,GAAG,IAAI,IAAI,MAAM,CAAC,IAAI,MAAM,MAAM;AAChD,cAAM,IAAI,QAAQ,GAAG,IAAI,IAAI,GAAG,IAAI,MAAM,MAAM;AAAA,MAClD;AAAA,IACF;AAEA,UAAM,SAAS,IAAI,KAAK,IAAI,OAAO,WAAW,OAAO,CAAC,GAAG,GAAG,YAAY,MAAM,CAAC;AAC/E,UAAM,MAAM,MAAM,KAAK,IAAI,IAAI,EAAE,CAAC,KAAK,IAAI,WAAW,CAAC,CAAC,GAAG,OAAO,CAAC;AACnE,UAAM,SAAS,IAAI,KAAK,IAAI,YAAY,CAAC,CAAC;AAC1C,UAAM,CAAC,EAAE,GAAG,CAAC,IAAI,OAAO;AACxB,UAAM,OAAO,OAAO;AAEpB,UAAM,UAAU,UAAU,MAAM,GAAG,GAAG,KAAK,IAAI;AAC/C,QAAI,CAAC,QAAS,QAAO;AACrB,WAAO,EAAE,MAAM,QAAQ,MAAM,GAAG,IAAI,GAAG,GAAG,IAAI,GAAG,GAAG,IAAI,GAAG,GAAG,IAAI,GAAG,YAAY,QAAQ,WAAW;AAAA,EACtG;AACF;AAGO,SAAS,UACd,MACA,GACA,GACA,MAC6C;AAC7C,MAAI,OAAO;AACX,MAAI,UAAU;AACd,MAAI,YAAY;AAChB,MAAI,OAAO;AACX,WAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,UAAM,MAAM,IAAI;AAChB,QAAI,OAAO;AACX,QAAI,QAAQ,KAAK,GAAG;AACpB,aAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,YAAM,IAAI,KAAK,MAAM,CAAC;AACtB,UAAI,IAAI,OAAO;AAAE,gBAAQ;AAAG,eAAO;AAAA,MAAE;AAAA,IACvC;AACA,UAAM,SAAS,SAAS;AACxB,WAAO;AACP,QAAI,SAAS,KAAK,OAAQ;AAE1B,QAAI,IAAI;AACR,QAAI,IAAI,UAAU,IAAI,GAAG;AACvB,UAAI,QAAQ;AACZ,eAAS,IAAI,GAAG,IAAI,GAAG,IAAK,UAAS,KAAK,IAAI,KAAK,MAAM,CAAC,IAAI,KAAK;AACnE,UAAI,IAAI;AAAA,IACV;AACA,eAAW;AACX;AACA,QAAI,QAAQ,KAAK,QAAQ,KAAK,OAAQ,SAAQ,KAAK,OAAO,CAAC;AAAA,aAClD,SAAS,KAAK,SAAS,EAAG,SAAQ;AAAA,EAC7C;AACA,MAAI,CAAC,KAAM,QAAO;AAClB,SAAO,EAAE,MAAM,YAAY,YAAY,IAAI,UAAU,YAAY,EAAE;AACrE;AAGO,SAAS,eACd,MACA,GACA,GAC2D;AAC3D,QAAM,UAAU,IAAI,WAAW,IAAI,CAAC;AACpC,QAAM,QAAkF,CAAC;AACzF,QAAM,QAAkB,CAAC;AAEzB,WAAS,QAAQ,GAAG,QAAQ,IAAI,GAAG,SAAS;AAC1C,QAAI,QAAQ,KAAK,KAAK,KAAK,KAAK,KAAK,WAAY;AACjD,QAAI,KAAK,QAAQ,GAAG,KAAK,IAAI,KAAM,QAAQ,IAAK,GAAG,KAAK;AACxD,QAAI,MAAM;AACV,QAAI,QAAQ;AACZ,UAAM,SAAS;AACf,UAAM,KAAK,KAAK;AAChB,YAAQ,KAAK,IAAI;AACjB,WAAO,MAAM,QAAQ;AACnB,YAAM,IAAI,MAAM,IAAI;AACpB,YAAM,KAAK,IAAI;AACf,YAAM,KAAM,IAAI,IAAK;AACrB,aAAO,KAAK,CAAC;AACb;AACA,UAAI,KAAK,GAAI,MAAK;AAClB,UAAI,KAAK,GAAI,MAAK;AAClB,UAAI,KAAK,GAAI,MAAK;AAClB,UAAI,KAAK,GAAI,MAAK;AAElB,UAAI,KAAK,KAAK,CAAC,QAAQ,IAAI,CAAC,KAAK,KAAK,IAAI,CAAC,IAAI,YAAY;AAAE,gBAAQ,IAAI,CAAC,IAAI;AAAG,cAAM,KAAK,IAAI,CAAC;AAAA,MAAE;AACnG,UAAI,KAAK,IAAI,KAAK,CAAC,QAAQ,IAAI,CAAC,KAAK,KAAK,IAAI,CAAC,IAAI,YAAY;AAAE,gBAAQ,IAAI,CAAC,IAAI;AAAG,cAAM,KAAK,IAAI,CAAC;AAAA,MAAE;AACvG,UAAI,KAAK,KAAK,CAAC,QAAQ,IAAI,CAAC,KAAK,KAAK,IAAI,CAAC,IAAI,YAAY;AAAE,gBAAQ,IAAI,CAAC,IAAI;AAAG,cAAM,KAAK,IAAI,CAAC;AAAA,MAAE;AACnG,UAAI,KAAK,IAAI,KAAK,CAAC,QAAQ,IAAI,CAAC,KAAK,KAAK,IAAI,CAAC,IAAI,YAAY;AAAE,gBAAQ,IAAI,CAAC,IAAI;AAAG,cAAM,KAAK,IAAI,CAAC;AAAA,MAAE;AAAA,IACzG;AACA,QAAI,KAAK,KAAK,IAAI,gBAAgB,KAAK,KAAK,IAAI,aAAc;AAC9D,UAAM,KAAK,EAAE,IAAI,IAAI,IAAI,IAAI,OAAO,MAAM,MAAM,CAAC;AAAA,EACnD;AAEA,SAAO,MACJ,OAAO,OAAK,EAAE,SAAS,cAAc,EACrC,KAAK,CAAC,GAAG,MAAO,EAAE,KAAK,EAAE,MAAQ,EAAE,KAAK,EAAE,EAAG;AAClD;AAEA,eAAe,UAAa,MAAc,QAAsC;AAC9E,MAAI;AACF,WAAO,MAAM,OAAO;AAAA,EACtB,SAAS,GAAG;AACV,UAAM,IAAI;AAAA,MACR,+EAAuC,IAAI,0DACxB,IAAI,wEAAuB,EAAY,OAAO;AAAA,IACnE;AAAA,EACF;AACF;AAGA,IAAI,gBAA2C;AAExC,SAAS,eAAmC;AACjD,MAAI,CAAC,eAAe;AAClB,oBAAgB,UAAU,OAAO,EAAE,MAAM,SAAO;AAC9C,sBAAgB;AAChB,YAAM;AAAA,IACR,CAAC;AAAA,EACH;AACA,SAAO;AACT;;;AC9TA,IAAM,mBAAmB;AAEzB,IAAM,kBAAkB;AASxB,eAAsB,UACpB,QACA,SACA,MACA,UACA,YACiC;AACjC,QAAM,SAAS,oBAAI,IAAuB;AAC1C,MAAI,QAAQ,SAAS,EAAG,QAAO;AAG/B,QAAM,YAAY,MAAMA;AAAA,IACtB;AAAA,IACA,MAAM,OAAO,gBAAgB;AAAA,EAC/B;AACA,MAAI,SAAS,WAAW;AACtB,UAAM,gBAAgB,OAAK;AACzB,UAAI,EAAE,UAAU,cAAc,EAAE,eAAe,GAAG;AAChD,gBAAQ,OAAO,MAAM,gBAAgB,EAAE,KAAK,IAAI,sCAAa,EAAE,KAAK,MAAM;AAAA,CAAU;AAAA,MACtF;AAAA,IACF,CAAC;AAAA,EACH;AACA,QAAM,SAAS,SAAS,YAAY,MAAM,aAAa,IAAI;AAE3D,QAAM,SAAS,MAAM,UAAU,cAAc,KAAK;AAClD,QAAM,MAAM,MAAM,OAAO,aAAa,IAAI,WAAW,MAAM,CAAC;AAC5D,MAAI;AACF,QAAI,OAAO;AACX,eAAW,QAAQ,IAAI,MAAM,GAAG;AAE9B,YAAM,SAAS,KAAK,SAAS;AAC7B,UAAI,CAAC,QAAQ,IAAI,MAAM,EAAG;AAC1B,mBAAa,EAAE,MAAM,QAAQ,IAAI;AACjC,UAAI;AACF,cAAM,SAAS,MAAM;AAAA,UACnB,WAAW,MAAM,QAAQ,MAAM,QAAQ,QAAQ;AAAA,UAC/C;AAAA,UACA,0BAAW,MAAM,8BAAU,kBAAkB,GAAI;AAAA,QACnD;AACA,eAAO,IAAI,QAAQ,MAAM;AAAA,MAC3B,SAAS,GAAG;AACV,iBAAS,KAAK;AAAA,UACZ,MAAM;AAAA,UACN,SAAS,sBAAO,MAAM,sBAAY,aAAa,QAAQ,EAAE,UAAU,OAAO,CAAC,CAAC;AAAA,UAC5E,MAAM;AAAA,QACR,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF,UAAE;AACA,QAAI,QAAQ;AACZ,WAAO,QAAQ;AAAA,EACjB;AACA,SAAO;AACT;AAEA,eAAe,WACb,MACA,QACA,MACA,QACA,UACoB;AACpB,QAAM,EAAE,eAAe,MAAM,gBAAgB,KAAK,IAAI,KAAK,gBAAgB;AAC3E,QAAM,WAAW,MAAM,KAAK,OAAO;AAAA,IACjC,OAAO;AAAA,IACP,QAAQ,OAAO,EAAE,KAAK,MAAM;AAAA,EAC9B,CAAC;AACD,QAAM,EAAE,MAAM,MAAM,OAAO,IAAI,QAAQ,GAAG,IAAI;AAC9C,QAAM,OAAO,WAAW,IAAI;AAE5B,MAAI,SAAS,WAAW;AACtB,UAAM,QAAQ,MAAM,OAAQ,cAAc,MAAM,IAAI,EAAE;AAEtD,UAAM,QAAQ,KAAK;AACnB,UAAM,SAAS,kBAAkB,MAAM,IAAI,IAAI,KAAK;AACpD,UAAM,aAAa,iBAAiB,QAAQ,OAAO,IAAI;AACvD,WAAO,iBAAiB,OAAO,QAAQ,MAAM,MAAM,OAAO,UAAU;AAAA,EACtE;AAOA,QAAM,cAAc,MAAMA;AAAA,IACxB;AAAA,IACA,MAAM,OAAO,OAAO;AAAA,EACtB;AACA,QAAM,WAAW;AACjB,QAAM,QACJ,OAAO,aAAa,aAAa,WAAY,SAAS,WAAY;AACpE,QAAM,MAAM,MAAM,MAAM,MAAM,EAAE,KAAK,EAAE,OAAO,IAAI,QAAQ,IAAI,UAAU,EAAE,EAAE,CAAC,EAC1E,IAAI,EACJ,SAAS;AAEZ,QAAM,OAAO,MAAM,KAAK,IAAI,WAAW,GAAG,GAAG,QAAQ,WAAW;AAChE,MAAI,CAAC,KAAK,KAAK,GAAG;AAChB,aAAS,KAAK,EAAE,MAAM,QAAQ,SAAS,sBAAO,MAAM,kCAAc,MAAM,aAAa,CAAC;AACtF,WAAO,CAAC;AAAA,EACV;AACA,SAAO,CAAC,EAAE,MAAM,aAAa,MAAM,KAAK,KAAK,GAAG,YAAY,OAAO,CAAC;AACtE;AAQO,SAAS,iBACd,OACA,YACA,MACA,MACA,OACA,YACW;AACX,QAAM,OAAmB,MAAM,IAAI,QAAM;AACvC,UAAM,IAAI,GAAG,IAAI;AACjB,WAAO;AAAA,MACL,MAAM,GAAG;AAAA,MACT,GAAG,KAAK,MAAM,GAAG,IAAI,KAAK;AAAA;AAAA,MAE1B,GAAG,KAAK,MAAM,QAAQ,GAAG,IAAI,GAAG,KAAK,KAAK;AAAA,MAC1C,GAAG,KAAK,MAAM,GAAG,IAAI,KAAK;AAAA,MAC1B,GAAG,KAAK,MAAM,CAAC;AAAA,MACf,UAAU,KAAK,IAAI,GAAG,KAAK,MAAM,IAAI,GAAG,CAAC;AAAA;AAAA,MACzC,UAAU;AAAA,MACV,UAAU;AAAA,IACZ;AAAA,EACF,CAAC;AACD,SAAO,2BAA2B,MAAM,YAAY,EAAE,SAAS,CAAC,GAAG,WAAW,CAAC,EAAE,GAAG,MAAM,MAAM,UAAU;AAC5G;AAEA,eAAeA,WAAa,MAAc,QAAsC;AAC9E,MAAI;AACF,WAAO,MAAM,OAAO;AAAA,EACtB,SAAS,GAAG;AACV,UAAM,IAAI;AAAA,MACR,kEAAoC,IAAI,0DACrB,IAAI,wEAAuB,EAAY,OAAO;AAAA,IACnE;AAAA,EACF;AACF;AAEA,eAAe,YAAe,SAAqB,IAAY,KAAyB;AAEtF,UAAQ,MAAM,MAAM;AAAA,EAAC,CAAC;AACtB,MAAI;AACJ,MAAI;AACF,WAAO,MAAM,QAAQ,KAAK;AAAA,MACxB;AAAA,MACA,IAAI,QAAe,CAAC,GAAG,WAAW;AAChC,gBAAQ,WAAW,MAAM,OAAO,IAAI,MAAM,GAAG,CAAC,GAAG,EAAE;AAAA,MACrD,CAAC;AAAA,IACH,CAAC;AAAA,EACH,UAAE;AACA,QAAI,MAAO,cAAa,KAAK;AAAA,EAC/B;AACF;AAGA,SAAS,WAAW,MAA8B;AAChD,QAAM,MAAM,IAAI,WAAW,KAAK,MAAM;AACtC,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK,GAAG;AACvC,QAAI,CAAC,IAAI,KAAK,IAAI,CAAC;AACnB,QAAI,IAAI,CAAC,IAAI,KAAK,IAAI,CAAC;AACvB,QAAI,IAAI,CAAC,IAAI,KAAK,CAAC;AACnB,QAAI,IAAI,CAAC,IAAI,KAAK,IAAI,CAAC;AAAA,EACzB;AACA,SAAO;AACT;","names":["tryImport"]}
@@ -243,4 +243,4 @@ export {
243
243
  ensureFormulaModels,
244
244
  ensureSingleModel
245
245
  };
246
- //# sourceMappingURL=chunk-NNDVFNMJ.js.map
246
+ //# sourceMappingURL=chunk-YNJDNP44.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../src/pdf/formula/models.ts"],"sourcesContent":["/**\r\n * 수식 OCR 모델 자동 다운로드 + SHA-256 검증\r\n *\r\n * 캐시 위치: `~/.cache/kordoc/models/pix2text/`\r\n * 총 ~155MB (MFD 44MB + MFR encoder 87MB + MFR decoder 30MB + tokenizer 40KB)\r\n *\r\n * HF URL + SHA 는 Docufinder archive/rust-formula-ocr-wip 에서 검증 완료됨.\r\n * 모델 버전 갱신 시 URL + SHA 를 함께 갱신해야 한다.\r\n */\r\n\r\nimport { createHash } from \"crypto\"\r\nimport { createReadStream } from \"fs\"\r\nimport { mkdir, stat, unlink, rename } from \"fs/promises\"\r\nimport { createWriteStream } from \"fs\"\r\nimport { homedir } from \"os\"\r\nimport { join, dirname } from \"path\"\r\nimport { pipeline } from \"stream/promises\"\r\nimport { Readable } from \"stream\"\r\n\r\nexport interface ModelSpec {\r\n name: string\r\n filename: string\r\n url: string\r\n sha256: string\r\n sizeMb: number\r\n}\r\n\r\n/** 모델 스펙 (SHA-256 은 실제 다운로드로 검증됨 — 변경 금지) */\r\nexport const MFD_MODEL: ModelSpec = {\r\n name: \"Pix2Text MFD\",\r\n filename: \"mfd.onnx\",\r\n url: \"https://huggingface.co/breezedeus/pix2text-mfd/resolve/main/mfd-v20240618.onnx\",\r\n sha256: \"51a8854743b17ae654729af8db82a630c1ccfa06debf4856c8b28055f87d02c1\",\r\n sizeMb: 42,\r\n}\r\n\r\nexport const MFR_ENCODER_MODEL: ModelSpec = {\r\n name: \"Pix2Text MFR encoder\",\r\n filename: \"encoder_model.onnx\",\r\n url: \"https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/encoder_model.onnx\",\r\n sha256: \"bd8d5c322792e9ec45793af5569e9748f82a3d728a9e00213dbfc56c1486f37d\",\r\n sizeMb: 87,\r\n}\r\n\r\nexport const MFR_DECODER_MODEL: ModelSpec = {\r\n name: \"Pix2Text MFR decoder\",\r\n filename: \"decoder_model.onnx\",\r\n url: \"https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/decoder_model.onnx\",\r\n sha256: \"fd0f92d7a012f3dae41e1ac79421aea0ea888b5a66cb3f9a004e424f82f3daed\",\r\n sizeMb: 30,\r\n}\r\n\r\nexport const MFR_TOKENIZER: ModelSpec = {\r\n name: \"Pix2Text MFR tokenizer\",\r\n filename: \"tokenizer.json\",\r\n url: \"https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/tokenizer.json\",\r\n sha256: \"3e2ab757277d22639bec28c9d7972e352d3d1dba223051fa674002dc5ab64df3\",\r\n sizeMb: 1,\r\n}\r\n\r\nexport const ALL_FORMULA_MODELS: ReadonlyArray<ModelSpec> = [\r\n MFD_MODEL,\r\n MFR_ENCODER_MODEL,\r\n MFR_DECODER_MODEL,\r\n MFR_TOKENIZER,\r\n]\r\n\r\n/**\r\n * 모델 캐시 하위 디렉토리 경로. 환경 변수 `KORDOC_MODEL_CACHE` 가 설정되면\r\n * 해당 경로를 우선 사용, 없으면 `~/.cache/kordoc/models/<subdir>/`.\r\n * (수식 OCR 외 텍스트 OCR 등 다른 모델군도 공용 — subdir 로 격리)\r\n */\r\nexport function getModelsDir(subdir: string): string {\r\n const override = process.env.KORDOC_MODEL_CACHE\r\n if (override && override.trim()) {\r\n return join(override, subdir)\r\n }\r\n return join(homedir(), \".cache\", \"kordoc\", \"models\", subdir)\r\n}\r\n\r\nexport function getFormulaModelsDir(): string {\r\n return getModelsDir(\"pix2text\")\r\n}\r\n\r\nexport interface ModelStatus {\r\n spec: ModelSpec\r\n localPath: string\r\n exists: boolean\r\n /** SHA 검증까지 끝났는가 (exists && valid) */\r\n verified: boolean\r\n /** 검증 실패 시 이유 */\r\n invalidReason?: string\r\n}\r\n\r\n/** 지정 모델군의 존재/유효성 상태 반환 (다운로드 없이 확인만) */\r\nexport async function getModelStatusIn(dir: string, specs: ReadonlyArray<ModelSpec>): Promise<ModelStatus[]> {\r\n const result: ModelStatus[] = []\r\n for (const spec of specs) {\r\n const localPath = join(dir, spec.filename)\r\n let exists = false\r\n try {\r\n const s = await stat(localPath)\r\n exists = s.isFile() && s.size > 0\r\n } catch {\r\n exists = false\r\n }\r\n if (!exists) {\r\n result.push({ spec, localPath, exists: false, verified: false })\r\n continue\r\n }\r\n try {\r\n const actual = await sha256OfFile(localPath)\r\n if (actual === spec.sha256) {\r\n result.push({ spec, localPath, exists: true, verified: true })\r\n } else {\r\n result.push({\r\n spec,\r\n localPath,\r\n exists: true,\r\n verified: false,\r\n invalidReason: `SHA256 mismatch: expected ${spec.sha256}, got ${actual}`,\r\n })\r\n }\r\n } catch (e) {\r\n result.push({\r\n spec,\r\n localPath,\r\n exists: true,\r\n verified: false,\r\n invalidReason: `SHA compute failed: ${(e as Error).message}`,\r\n })\r\n }\r\n }\r\n return result\r\n}\r\n\r\n/** 모든 수식 모델의 존재/유효성 상태 반환 (다운로드 없이 확인만) */\r\nexport async function getFormulaModelStatus(): Promise<ModelStatus[]> {\r\n return getModelStatusIn(getFormulaModelsDir(), ALL_FORMULA_MODELS)\r\n}\r\n\r\nexport interface DownloadProgress {\r\n spec: ModelSpec\r\n /** 현재까지 다운로드한 바이트 */\r\n downloaded: number\r\n /** 전체 바이트 (알 수 있으면) */\r\n total: number | null\r\n /** \"download\" | \"verify\" | \"done\" | \"skip\" */\r\n phase: \"download\" | \"verify\" | \"done\" | \"skip\" | \"error\"\r\n message?: string\r\n}\r\n\r\nexport type ProgressHandler = (p: DownloadProgress) => void\r\n\r\n/**\r\n * 지정 모델군을 다운로드/검증한다.\r\n * - 이미 있고 SHA 일치 → skip\r\n * - 없음 → 다운로드 후 SHA 검증. 실패 시 파일 삭제.\r\n */\r\nexport async function ensureModelsIn(dir: string, specs: ReadonlyArray<ModelSpec>, onProgress?: ProgressHandler): Promise<void> {\r\n await mkdir(dir, { recursive: true })\r\n\r\n for (const spec of specs) {\r\n const localPath = join(dir, spec.filename)\r\n\r\n if (await isExistingValid(localPath, spec.sha256)) {\r\n onProgress?.({\r\n spec,\r\n downloaded: 0,\r\n total: null,\r\n phase: \"skip\",\r\n message: \"이미 존재 + SHA 일치\",\r\n })\r\n continue\r\n }\r\n\r\n // 기존 파일 있지만 SHA 불일치 → 삭제\r\n try {\r\n await unlink(localPath)\r\n } catch {\r\n // 없을 수 있음\r\n }\r\n\r\n await downloadToFile(spec, localPath, onProgress)\r\n }\r\n}\r\n\r\n/** 필요한 모든 수식 모델을 다운로드/검증한다. */\r\nexport async function ensureFormulaModels(onProgress?: ProgressHandler): Promise<void> {\r\n return ensureModelsIn(getFormulaModelsDir(), ALL_FORMULA_MODELS, onProgress)\r\n}\r\n\r\n/** 단일 모델만 확인/다운로드 (진단 UI 에서 개별 상태 새로고침용) */\r\nexport async function ensureSingleModel(spec: ModelSpec, onProgress?: ProgressHandler): Promise<void> {\r\n const dir = getFormulaModelsDir()\r\n await mkdir(dir, { recursive: true })\r\n const localPath = join(dir, spec.filename)\r\n if (await isExistingValid(localPath, spec.sha256)) {\r\n onProgress?.({ spec, downloaded: 0, total: null, phase: \"skip\" })\r\n return\r\n }\r\n try {\r\n await unlink(localPath)\r\n } catch {\r\n // 재다운로드 전 손상/잔존 파일 정리 best-effort — 애초에 없으면 실패해도 무방\r\n }\r\n await downloadToFile(spec, localPath, onProgress)\r\n}\r\n\r\nasync function isExistingValid(localPath: string, sha256Expected: string): Promise<boolean> {\r\n try {\r\n const s = await stat(localPath)\r\n if (!s.isFile() || s.size === 0) return false\r\n } catch {\r\n return false\r\n }\r\n try {\r\n const actual = await sha256OfFile(localPath)\r\n return actual === sha256Expected\r\n } catch {\r\n return false\r\n }\r\n}\r\n\r\nasync function downloadToFile(\r\n spec: ModelSpec,\r\n localPath: string,\r\n onProgress?: ProgressHandler,\r\n): Promise<void> {\r\n // 먼저 .part 로 받고 검증 후 rename — 중단된 다운로드가 \"정상 파일\"로 오인되는 걸 방지\r\n const partPath = `${localPath}.part`\r\n await mkdir(dirname(localPath), { recursive: true })\r\n\r\n const resp = await fetch(spec.url, {\r\n headers: {\r\n // HF CDN 은 UA 없으면 가끔 403 을 뱉는다\r\n \"User-Agent\": \"kordoc-formula-ocr/1.0 (+https://github.com/chrisryugj/kordoc)\",\r\n },\r\n })\r\n if (!resp.ok || !resp.body) {\r\n throw new Error(\r\n `${spec.name} 다운로드 실패: HTTP ${resp.status} ${resp.statusText} (${spec.url})`,\r\n )\r\n }\r\n\r\n const lenHeader = resp.headers.get(\"content-length\")\r\n const total = lenHeader ? Number.parseInt(lenHeader, 10) : null\r\n let downloaded = 0\r\n\r\n const ws = createWriteStream(partPath)\r\n try {\r\n const reader = Readable.fromWeb(resp.body as unknown as import(\"stream/web\").ReadableStream)\r\n reader.on(\"data\", (chunk: Buffer | Uint8Array) => {\r\n downloaded += chunk.length\r\n onProgress?.({\r\n spec,\r\n downloaded,\r\n total,\r\n phase: \"download\",\r\n })\r\n })\r\n await pipeline(reader, ws)\r\n } catch (e) {\r\n try {\r\n await unlink(partPath)\r\n } catch {\r\n // .part 임시파일 정리 best-effort — 본 에러(스트리밍 실패)를 가리지 않도록 무시\r\n }\r\n throw new Error(`${spec.name} 스트리밍 실패: ${(e as Error).message}`)\r\n }\r\n\r\n onProgress?.({\r\n spec,\r\n downloaded,\r\n total,\r\n phase: \"verify\",\r\n })\r\n\r\n // SHA 검증\r\n let actual: string\r\n try {\r\n actual = await sha256OfFile(partPath)\r\n } catch (e) {\r\n try {\r\n await unlink(partPath)\r\n } catch {\r\n // .part 임시파일 정리 best-effort — 본 에러(SHA 계산 실패)를 가리지 않도록 무시\r\n }\r\n throw new Error(`${spec.name} SHA 계산 실패: ${(e as Error).message}`)\r\n }\r\n\r\n if (actual !== spec.sha256) {\r\n try {\r\n await unlink(partPath)\r\n } catch {\r\n // 오염된 .part 정리 best-effort — 본 에러(SHA mismatch)를 가리지 않도록 무시\r\n }\r\n throw new Error(\r\n `${spec.name} SHA256 mismatch: expected ${spec.sha256}, got ${actual} — 모델 URL 이 오염되었거나 전송 중 손상되었습니다.`,\r\n )\r\n }\r\n\r\n await rename(partPath, localPath)\r\n onProgress?.({\r\n spec,\r\n downloaded,\r\n total,\r\n phase: \"done\",\r\n })\r\n}\r\n\r\nasync function sha256OfFile(p: string): Promise<string> {\r\n const h = createHash(\"sha256\")\r\n const stream = createReadStream(p)\r\n await pipeline(stream, async function* (src) {\r\n for await (const chunk of src) {\r\n h.update(chunk)\r\n // yield 안 함 (소비만)\r\n }\r\n })\r\n return h.digest(\"hex\")\r\n}\r\n"],"mappings":";;;AAUA,SAAS,kBAAkB;AAC3B,SAAS,wBAAwB;AACjC,SAAS,OAAO,MAAM,QAAQ,cAAc;AAC5C,SAAS,yBAAyB;AAClC,SAAS,eAAe;AACxB,SAAS,MAAM,eAAe;AAC9B,SAAS,gBAAgB;AACzB,SAAS,gBAAgB;AAWlB,IAAM,YAAuB;AAAA,EAClC,MAAM;AAAA,EACN,UAAU;AAAA,EACV,KAAK;AAAA,EACL,QAAQ;AAAA,EACR,QAAQ;AACV;AAEO,IAAM,oBAA+B;AAAA,EAC1C,MAAM;AAAA,EACN,UAAU;AAAA,EACV,KAAK;AAAA,EACL,QAAQ;AAAA,EACR,QAAQ;AACV;AAEO,IAAM,oBAA+B;AAAA,EAC1C,MAAM;AAAA,EACN,UAAU;AAAA,EACV,KAAK;AAAA,EACL,QAAQ;AAAA,EACR,QAAQ;AACV;AAEO,IAAM,gBAA2B;AAAA,EACtC,MAAM;AAAA,EACN,UAAU;AAAA,EACV,KAAK;AAAA,EACL,QAAQ;AAAA,EACR,QAAQ;AACV;AAEO,IAAM,qBAA+C;AAAA,EAC1D;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAOO,SAAS,aAAa,QAAwB;AACnD,QAAM,WAAW,QAAQ,IAAI;AAC7B,MAAI,YAAY,SAAS,KAAK,GAAG;AAC/B,WAAO,KAAK,UAAU,MAAM;AAAA,EAC9B;AACA,SAAO,KAAK,QAAQ,GAAG,UAAU,UAAU,UAAU,MAAM;AAC7D;AAEO,SAAS,sBAA8B;AAC5C,SAAO,aAAa,UAAU;AAChC;AAaA,eAAsB,iBAAiB,KAAa,OAAyD;AAC3G,QAAM,SAAwB,CAAC;AAC/B,aAAW,QAAQ,OAAO;AACxB,UAAM,YAAY,KAAK,KAAK,KAAK,QAAQ;AACzC,QAAI,SAAS;AACb,QAAI;AACF,YAAM,IAAI,MAAM,KAAK,SAAS;AAC9B,eAAS,EAAE,OAAO,KAAK,EAAE,OAAO;AAAA,IAClC,QAAQ;AACN,eAAS;AAAA,IACX;AACA,QAAI,CAAC,QAAQ;AACX,aAAO,KAAK,EAAE,MAAM,WAAW,QAAQ,OAAO,UAAU,MAAM,CAAC;AAC/D;AAAA,IACF;AACA,QAAI;AACF,YAAM,SAAS,MAAM,aAAa,SAAS;AAC3C,UAAI,WAAW,KAAK,QAAQ;AAC1B,eAAO,KAAK,EAAE,MAAM,WAAW,QAAQ,MAAM,UAAU,KAAK,CAAC;AAAA,MAC/D,OAAO;AACL,eAAO,KAAK;AAAA,UACV;AAAA,UACA;AAAA,UACA,QAAQ;AAAA,UACR,UAAU;AAAA,UACV,eAAe,6BAA6B,KAAK,MAAM,SAAS,MAAM;AAAA,QACxE,CAAC;AAAA,MACH;AAAA,IACF,SAAS,GAAG;AACV,aAAO,KAAK;AAAA,QACV;AAAA,QACA;AAAA,QACA,QAAQ;AAAA,QACR,UAAU;AAAA,QACV,eAAe,uBAAwB,EAAY,OAAO;AAAA,MAC5D,CAAC;AAAA,IACH;AAAA,EACF;AACA,SAAO;AACT;AAGA,eAAsB,wBAAgD;AACpE,SAAO,iBAAiB,oBAAoB,GAAG,kBAAkB;AACnE;AAoBA,eAAsB,eAAe,KAAa,OAAiC,YAA6C;AAC9H,QAAM,MAAM,KAAK,EAAE,WAAW,KAAK,CAAC;AAEpC,aAAW,QAAQ,OAAO;AACxB,UAAM,YAAY,KAAK,KAAK,KAAK,QAAQ;AAEzC,QAAI,MAAM,gBAAgB,WAAW,KAAK,MAAM,GAAG;AACjD,mBAAa;AAAA,QACX;AAAA,QACA,YAAY;AAAA,QACZ,OAAO;AAAA,QACP,OAAO;AAAA,QACP,SAAS;AAAA,MACX,CAAC;AACD;AAAA,IACF;AAGA,QAAI;AACF,YAAM,OAAO,SAAS;AAAA,IACxB,QAAQ;AAAA,IAER;AAEA,UAAM,eAAe,MAAM,WAAW,UAAU;AAAA,EAClD;AACF;AAGA,eAAsB,oBAAoB,YAA6C;AACrF,SAAO,eAAe,oBAAoB,GAAG,oBAAoB,UAAU;AAC7E;AAGA,eAAsB,kBAAkB,MAAiB,YAA6C;AACpG,QAAM,MAAM,oBAAoB;AAChC,QAAM,MAAM,KAAK,EAAE,WAAW,KAAK,CAAC;AACpC,QAAM,YAAY,KAAK,KAAK,KAAK,QAAQ;AACzC,MAAI,MAAM,gBAAgB,WAAW,KAAK,MAAM,GAAG;AACjD,iBAAa,EAAE,MAAM,YAAY,GAAG,OAAO,MAAM,OAAO,OAAO,CAAC;AAChE;AAAA,EACF;AACA,MAAI;AACF,UAAM,OAAO,SAAS;AAAA,EACxB,QAAQ;AAAA,EAER;AACA,QAAM,eAAe,MAAM,WAAW,UAAU;AAClD;AAEA,eAAe,gBAAgB,WAAmB,gBAA0C;AAC1F,MAAI;AACF,UAAM,IAAI,MAAM,KAAK,SAAS;AAC9B,QAAI,CAAC,EAAE,OAAO,KAAK,EAAE,SAAS,EAAG,QAAO;AAAA,EAC1C,QAAQ;AACN,WAAO;AAAA,EACT;AACA,MAAI;AACF,UAAM,SAAS,MAAM,aAAa,SAAS;AAC3C,WAAO,WAAW;AAAA,EACpB,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAEA,eAAe,eACb,MACA,WACA,YACe;AAEf,QAAM,WAAW,GAAG,SAAS;AAC7B,QAAM,MAAM,QAAQ,SAAS,GAAG,EAAE,WAAW,KAAK,CAAC;AAEnD,QAAM,OAAO,MAAM,MAAM,KAAK,KAAK;AAAA,IACjC,SAAS;AAAA;AAAA,MAEP,cAAc;AAAA,IAChB;AAAA,EACF,CAAC;AACD,MAAI,CAAC,KAAK,MAAM,CAAC,KAAK,MAAM;AAC1B,UAAM,IAAI;AAAA,MACR,GAAG,KAAK,IAAI,gDAAkB,KAAK,MAAM,IAAI,KAAK,UAAU,KAAK,KAAK,GAAG;AAAA,IAC3E;AAAA,EACF;AAEA,QAAM,YAAY,KAAK,QAAQ,IAAI,gBAAgB;AACnD,QAAM,QAAQ,YAAY,OAAO,SAAS,WAAW,EAAE,IAAI;AAC3D,MAAI,aAAa;AAEjB,QAAM,KAAK,kBAAkB,QAAQ;AACrC,MAAI;AACF,UAAM,SAAS,SAAS,QAAQ,KAAK,IAAsD;AAC3F,WAAO,GAAG,QAAQ,CAAC,UAA+B;AAChD,oBAAc,MAAM;AACpB,mBAAa;AAAA,QACX;AAAA,QACA;AAAA,QACA;AAAA,QACA,OAAO;AAAA,MACT,CAAC;AAAA,IACH,CAAC;AACD,UAAM,SAAS,QAAQ,EAAE;AAAA,EAC3B,SAAS,GAAG;AACV,QAAI;AACF,YAAM,OAAO,QAAQ;AAAA,IACvB,QAAQ;AAAA,IAER;AACA,UAAM,IAAI,MAAM,GAAG,KAAK,IAAI,2CAAc,EAAY,OAAO,EAAE;AAAA,EACjE;AAEA,eAAa;AAAA,IACX;AAAA,IACA;AAAA,IACA;AAAA,IACA,OAAO;AAAA,EACT,CAAC;AAGD,MAAI;AACJ,MAAI;AACF,aAAS,MAAM,aAAa,QAAQ;AAAA,EACtC,SAAS,GAAG;AACV,QAAI;AACF,YAAM,OAAO,QAAQ;AAAA,IACvB,QAAQ;AAAA,IAER;AACA,UAAM,IAAI,MAAM,GAAG,KAAK,IAAI,mCAAgB,EAAY,OAAO,EAAE;AAAA,EACnE;AAEA,MAAI,WAAW,KAAK,QAAQ;AAC1B,QAAI;AACF,YAAM,OAAO,QAAQ;AAAA,IACvB,QAAQ;AAAA,IAER;AACA,UAAM,IAAI;AAAA,MACR,GAAG,KAAK,IAAI,8BAA8B,KAAK,MAAM,SAAS,MAAM;AAAA,IACtE;AAAA,EACF;AAEA,QAAM,OAAO,UAAU,SAAS;AAChC,eAAa;AAAA,IACX;AAAA,IACA;AAAA,IACA;AAAA,IACA,OAAO;AAAA,EACT,CAAC;AACH;AAEA,eAAe,aAAa,GAA4B;AACtD,QAAM,IAAI,WAAW,QAAQ;AAC7B,QAAM,SAAS,iBAAiB,CAAC;AACjC,QAAM,SAAS,QAAQ,iBAAiB,KAAK;AAC3C,qBAAiB,SAAS,KAAK;AAC7B,QAAE,OAAO,KAAK;AAAA,IAEhB;AAAA,EACF,CAAC;AACD,SAAO,EAAE,OAAO,KAAK;AACvB;","names":[]}
@@ -1,7 +1,7 @@
1
1
  #!/usr/bin/env node
2
2
 
3
3
  // src/utils.ts
4
- var VERSION = true ? "4.2.9" : "0.0.0-dev";
4
+ var VERSION = true ? "4.3.1" : "0.0.0-dev";
5
5
  function toArrayBuffer(buf) {
6
6
  if (buf.byteOffset === 0 && buf.byteLength === buf.buffer.byteLength) {
7
7
  return buf.buffer;
@@ -96,8 +96,9 @@ function safeMax(arr) {
96
96
  function classifyError(err) {
97
97
  if (!(err instanceof Error)) return "PARSE_ERROR";
98
98
  const msg = err.message;
99
- if (msg.includes("\uC554\uD638\uD654")) return "ENCRYPTED";
100
99
  if (msg.includes("DRM")) return "DRM_PROTECTED";
100
+ if (msg.includes("\uC554\uD638\uD654") || msg.includes("\uC554\uD638\uB85C \uBCF4\uD638")) return "ENCRYPTED";
101
+ if (msg.includes("optional dependency")) return "MISSING_DEPENDENCY";
101
102
  if (msg.includes("ZIP bomb") || msg.includes("ZIP \uBE44\uC555\uCD95 \uD06C\uAE30 \uCD08\uACFC") || msg.includes("ZIP \uC5D4\uD2B8\uB9AC \uC218 \uCD08\uACFC")) return "ZIP_BOMB";
102
103
  if (msg.includes("bomb") || msg.includes("\uD06C\uAE30 \uCD08\uACFC") || msg.includes("\uC555\uCD95 \uD574\uC81C")) return "DECOMPRESSION_BOMB";
103
104
  if (msg.includes("\uC774\uBBF8\uC9C0 \uAE30\uBC18")) return "IMAGE_BASED_PDF";
@@ -121,4 +122,4 @@ export {
121
122
  safeMax,
122
123
  classifyError
123
124
  };
124
- //# sourceMappingURL=chunk-3TSTZGCJ.js.map
125
+ //# sourceMappingURL=chunk-ZJ75YDW7.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../src/utils.ts"],"sourcesContent":["/** kordoc 공용 유틸리티 */\r\n\r\n/** 빌드 타임에 tsup define으로 주입되는 버전 */\r\ndeclare const __KORDOC_VERSION__: string\r\nexport const VERSION: string = typeof __KORDOC_VERSION__ !== \"undefined\" ? __KORDOC_VERSION__ : \"0.0.0-dev\"\r\n\r\n/**\r\n * Node.js Buffer → ArrayBuffer 변환\r\n * pool Buffer의 공유 ArrayBuffer 문제를 안전하게 처리.\r\n * offset=0이고 전체 ArrayBuffer를 차지하면 복사 없이 직접 반환.\r\n */\r\nexport function toArrayBuffer(buf: Buffer): ArrayBuffer {\r\n if (buf.byteOffset === 0 && buf.byteLength === buf.buffer.byteLength) {\r\n return buf.buffer as ArrayBuffer\r\n }\r\n return buf.buffer.slice(buf.byteOffset, buf.byteOffset + buf.byteLength) as ArrayBuffer\r\n}\r\n\r\n/**\r\n * kordoc 내부 에러 클래스 — 사용자에게 노출해도 안전한 메시지만 포함.\r\n * MCP 에러 정제에서 instanceof로 판별하여 allowlist 패턴 매칭 없이 안전하게 통과.\r\n */\r\nexport class KordocError extends Error {\r\n constructor(message: string) {\r\n super(message)\r\n this.name = \"KordocError\"\r\n }\r\n}\r\n\r\n/**\r\n * 에러 메시지 정제 — KordocError는 그대로, 나머지는 일반 메시지로 대체.\r\n * 파일시스템 경로, 스택 트레이스 등 내부 정보 노출 방지.\r\n */\r\nexport function sanitizeError(err: unknown): string {\r\n if (err instanceof KordocError) return err.message\r\n return \"문서 처리 중 오류가 발생했습니다\"\r\n}\r\n\r\n/**\r\n * ZIP 엔트리 경로의 경로 순회 여부 판별.\r\n * 백슬래시 정규화, .., 절대경로, Windows 드라이브 문자 모두 차단.\r\n */\r\nexport function isPathTraversal(name: string): boolean {\r\n if (name.includes(\"\\x00\")) return true\r\n const normalized = name.replace(/\\\\/g, \"/\")\r\n const segments = normalized.split(\"/\")\r\n return segments.some(s => s === \"..\") || normalized.startsWith(\"/\") || /^[A-Za-z]:/.test(normalized)\r\n}\r\n\r\n// ─── HWPX 섹션 href 해석 (parser/roundtrip 공용) ────────────────────\r\n\r\n/**\r\n * manifest href를 본문 섹션 경로(`Contents/sectionN.xml`)로 정규화.\r\n * 비본문 XML(header/script/settings)·path traversal·드라이브 경로는 null로 거른다.\r\n * 백슬래시 구분자를 슬래시로 통일해 parser와 roundtrip이 동일한 목록을 만든다.\r\n */\r\nexport function normalizeSectionHref(href: string): string | null {\r\n if (!href) return null\r\n let normalized = href.replace(/\\\\/g, \"/\").replace(/^\\/+/, \"\")\r\n if (isPathTraversal(normalized)) return null\r\n if (/^[Ss]ection\\d+\\.xml$/.test(normalized)) normalized = \"Contents/\" + normalized\r\n return /(?:^|\\/)[Ss]ection\\d+\\.xml$/.test(normalized) ? normalized : null\r\n}\r\n\r\n/** sectionN.xml을 N 숫자 순서로 정렬 (section10 > section2). */\r\nexport function compareSectionPaths(a: string, b: string): number {\r\n const ai = Number(a.match(/[Ss]ection(\\d+)\\.xml$/)?.[1] ?? Number.MAX_SAFE_INTEGER)\r\n const bi = Number(b.match(/[Ss]ection(\\d+)\\.xml$/)?.[1] ?? Number.MAX_SAFE_INTEGER)\r\n return ai === bi ? a.localeCompare(b) : ai - bi\r\n}\r\n\r\n// ─── ZIP 안전 로딩 (ZIP bomb 방지) ────────────────────\r\n\r\n/**\r\n * ZIP bomb 사전 검사 — Central Directory에서 비압축 합계와 엔트리 수 확인.\r\n * HWPX/XLSX/DOCX 등 모든 ZIP 기반 포맷에서 공통 사용.\r\n */\r\nexport function precheckZipSize(\r\n buffer: ArrayBuffer,\r\n maxUncompressedSize = 256 * 1024 * 1024, // parser-shared MAX_DECOMPRESS_SIZE 와 동기\r\n maxEntries = 500,\r\n): { totalUncompressed: number; entryCount: number } {\r\n try {\r\n const data = new DataView(buffer)\r\n const len = buffer.byteLength\r\n // EOCD 시그니처 역방향 스캔\r\n let eocdOffset = -1\r\n for (let i = len - 22; i >= Math.max(0, len - 65557); i--) {\r\n if (data.getUint32(i, true) === 0x06054b50) { eocdOffset = i; break }\r\n }\r\n if (eocdOffset < 0) return { totalUncompressed: 0, entryCount: 0 }\r\n\r\n const entryCount = data.getUint16(eocdOffset + 10, true)\r\n if (entryCount > maxEntries) {\r\n throw new KordocError(`ZIP 엔트리 수 초과: ${entryCount} (최대 ${maxEntries})`)\r\n }\r\n\r\n const cdSize = data.getUint32(eocdOffset + 12, true)\r\n const cdOffset = data.getUint32(eocdOffset + 16, true)\r\n if (cdOffset + cdSize > len) return { totalUncompressed: 0, entryCount }\r\n\r\n let totalUncompressed = 0\r\n let pos = cdOffset\r\n for (let i = 0; i < entryCount && pos + 46 <= cdOffset + cdSize; i++) {\r\n if (data.getUint32(pos, true) !== 0x02014b50) break\r\n totalUncompressed += data.getUint32(pos + 24, true)\r\n const nameLen = data.getUint16(pos + 28, true)\r\n const extraLen = data.getUint16(pos + 30, true)\r\n const commentLen = data.getUint16(pos + 32, true)\r\n pos += 46 + nameLen + extraLen + commentLen\r\n }\r\n\r\n if (totalUncompressed > maxUncompressedSize) {\r\n throw new KordocError(`ZIP 비압축 크기 초과: ${(totalUncompressed / 1024 / 1024).toFixed(1)}MB (최대 ${maxUncompressedSize / 1024 / 1024}MB)`)\r\n }\r\n\r\n return { totalUncompressed, entryCount }\r\n } catch (err) {\r\n if (err instanceof KordocError) throw err\r\n return { totalUncompressed: 0, entryCount: 0 }\r\n }\r\n}\r\n\r\n/** XXE/Billion Laughs 방지 — DOCTYPE 제거 (내부 DTD 서브셋 포함) */\r\nexport function stripDtd(xml: string): string {\r\n return xml.replace(/<!DOCTYPE\\s[^[>]*(\\[[\\s\\S]*?\\])?\\s*>/gi, \"\")\r\n}\r\n\r\n/** 하이퍼링크 URL 살균 — javascript: 등 XSS 위험 스킴 차단 */\r\nconst SAFE_HREF_RE = /^(?:https?:|mailto:|tel:|#)/i\r\nexport function sanitizeHref(href: string): string | null {\r\n const trimmed = href.trim()\r\n if (!trimmed || !SAFE_HREF_RE.test(trimmed)) return null\r\n // 괄호 percent-encoding — 마크다운 링크 목적지에서 불균형 ')'는 링크를 조기 종료시켜\r\n // 문법이 깨진다 (원문 하이퍼링크에 ')'가 박힌 실문서 존재). %28/%29는 의미 동일.\r\n return trimmed.replace(/\\(/g, \"%28\").replace(/\\)/g, \"%29\")\r\n}\r\n\r\n// ─── 안전한 min/max (스택 오버플로 방지) ─────────────\r\n\r\n/** Math.min(...arr) 대체 — 대형 배열에서 스택 오버플로 방지 */\r\nexport function safeMin(arr: number[]): number {\r\n let min = Infinity\r\n for (let i = 0; i < arr.length; i++) if (arr[i] < min) min = arr[i]\r\n return min\r\n}\r\n\r\n/** Math.max(...arr) 대체 — 대형 배열에서 스택 오버플로 방지 */\r\nexport function safeMax(arr: number[]): number {\r\n let max = -Infinity\r\n for (let i = 0; i < arr.length; i++) if (arr[i] > max) max = arr[i]\r\n return max\r\n}\r\n\r\n// ─── 에러 분류 ──────────────────────────────────────\r\n\r\nimport type { ErrorCode } from \"./types.js\"\r\n\r\n/** 에러를 구조화된 ErrorCode로 분류 — KordocError 메시지 패턴 매칭 */\r\nexport function classifyError(err: unknown): ErrorCode {\r\n if (!(err instanceof Error)) return \"PARSE_ERROR\"\r\n const msg = err.message\r\n // DRM 을 암호화보다 먼저 — \"DRM 암호화된 HWPX…\" 류 메시지가 ENCRYPTED 로 떨어지지 않도록\r\n if (msg.includes(\"DRM\")) return \"DRM_PROTECTED\"\r\n if (msg.includes(\"암호화\") || msg.includes(\"암호로 보호\")) return \"ENCRYPTED\"\r\n if (msg.includes(\"optional dependency\")) return \"MISSING_DEPENDENCY\"\r\n if (msg.includes(\"ZIP bomb\") || msg.includes(\"ZIP 비압축 크기 초과\") || msg.includes(\"ZIP 엔트리 수 초과\")) return \"ZIP_BOMB\"\r\n if (msg.includes(\"bomb\") || msg.includes(\"크기 초과\") || msg.includes(\"압축 해제\")) return \"DECOMPRESSION_BOMB\"\r\n if (msg.includes(\"이미지 기반\")) return \"IMAGE_BASED_PDF\"\r\n if (msg.includes(\"섹션\") && (msg.includes(\"찾을 수 없\") || msg.includes(\"없음\"))) return \"NO_SECTIONS\"\r\n if (msg.includes(\"시그니처\") || msg.includes(\"복구할 수 없\")) return \"CORRUPTED\"\r\n return \"PARSE_ERROR\"\r\n}\r\n"],"mappings":";;;AAIO,IAAM,UAAkB,OAA4C,UAAqB;AAOzF,SAAS,cAAc,KAA0B;AACtD,MAAI,IAAI,eAAe,KAAK,IAAI,eAAe,IAAI,OAAO,YAAY;AACpE,WAAO,IAAI;AAAA,EACb;AACA,SAAO,IAAI,OAAO,MAAM,IAAI,YAAY,IAAI,aAAa,IAAI,UAAU;AACzE;AAMO,IAAM,cAAN,cAA0B,MAAM;AAAA,EACrC,YAAY,SAAiB;AAC3B,UAAM,OAAO;AACb,SAAK,OAAO;AAAA,EACd;AACF;AAMO,SAAS,cAAc,KAAsB;AAClD,MAAI,eAAe,YAAa,QAAO,IAAI;AAC3C,SAAO;AACT;AAMO,SAAS,gBAAgB,MAAuB;AACrD,MAAI,KAAK,SAAS,IAAM,EAAG,QAAO;AAClC,QAAM,aAAa,KAAK,QAAQ,OAAO,GAAG;AAC1C,QAAM,WAAW,WAAW,MAAM,GAAG;AACrC,SAAO,SAAS,KAAK,OAAK,MAAM,IAAI,KAAK,WAAW,WAAW,GAAG,KAAK,aAAa,KAAK,UAAU;AACrG;AASO,SAAS,qBAAqB,MAA6B;AAChE,MAAI,CAAC,KAAM,QAAO;AAClB,MAAI,aAAa,KAAK,QAAQ,OAAO,GAAG,EAAE,QAAQ,QAAQ,EAAE;AAC5D,MAAI,gBAAgB,UAAU,EAAG,QAAO;AACxC,MAAI,uBAAuB,KAAK,UAAU,EAAG,cAAa,cAAc;AACxE,SAAO,8BAA8B,KAAK,UAAU,IAAI,aAAa;AACvE;AAGO,SAAS,oBAAoB,GAAW,GAAmB;AAChE,QAAM,KAAK,OAAO,EAAE,MAAM,uBAAuB,IAAI,CAAC,KAAK,OAAO,gBAAgB;AAClF,QAAM,KAAK,OAAO,EAAE,MAAM,uBAAuB,IAAI,CAAC,KAAK,OAAO,gBAAgB;AAClF,SAAO,OAAO,KAAK,EAAE,cAAc,CAAC,IAAI,KAAK;AAC/C;AAQO,SAAS,gBACd,QACA,sBAAsB,MAAM,OAAO,MACnC,aAAa,KACsC;AACnD,MAAI;AACF,UAAM,OAAO,IAAI,SAAS,MAAM;AAChC,UAAM,MAAM,OAAO;AAEnB,QAAI,aAAa;AACjB,aAAS,IAAI,MAAM,IAAI,KAAK,KAAK,IAAI,GAAG,MAAM,KAAK,GAAG,KAAK;AACzD,UAAI,KAAK,UAAU,GAAG,IAAI,MAAM,WAAY;AAAE,qBAAa;AAAG;AAAA,MAAM;AAAA,IACtE;AACA,QAAI,aAAa,EAAG,QAAO,EAAE,mBAAmB,GAAG,YAAY,EAAE;AAEjE,UAAM,aAAa,KAAK,UAAU,aAAa,IAAI,IAAI;AACvD,QAAI,aAAa,YAAY;AAC3B,YAAM,IAAI,YAAY,+CAAiB,UAAU,kBAAQ,UAAU,GAAG;AAAA,IACxE;AAEA,UAAM,SAAS,KAAK,UAAU,aAAa,IAAI,IAAI;AACnD,UAAM,WAAW,KAAK,UAAU,aAAa,IAAI,IAAI;AACrD,QAAI,WAAW,SAAS,IAAK,QAAO,EAAE,mBAAmB,GAAG,WAAW;AAEvE,QAAI,oBAAoB;AACxB,QAAI,MAAM;AACV,aAAS,IAAI,GAAG,IAAI,cAAc,MAAM,MAAM,WAAW,QAAQ,KAAK;AACpE,UAAI,KAAK,UAAU,KAAK,IAAI,MAAM,SAAY;AAC9C,2BAAqB,KAAK,UAAU,MAAM,IAAI,IAAI;AAClD,YAAM,UAAU,KAAK,UAAU,MAAM,IAAI,IAAI;AAC7C,YAAM,WAAW,KAAK,UAAU,MAAM,IAAI,IAAI;AAC9C,YAAM,aAAa,KAAK,UAAU,MAAM,IAAI,IAAI;AAChD,aAAO,KAAK,UAAU,WAAW;AAAA,IACnC;AAEA,QAAI,oBAAoB,qBAAqB;AAC3C,YAAM,IAAI,YAAY,sDAAmB,oBAAoB,OAAO,MAAM,QAAQ,CAAC,CAAC,oBAAU,sBAAsB,OAAO,IAAI,KAAK;AAAA,IACtI;AAEA,WAAO,EAAE,mBAAmB,WAAW;AAAA,EACzC,SAAS,KAAK;AACZ,QAAI,eAAe,YAAa,OAAM;AACtC,WAAO,EAAE,mBAAmB,GAAG,YAAY,EAAE;AAAA,EAC/C;AACF;AAGO,SAAS,SAAS,KAAqB;AAC5C,SAAO,IAAI,QAAQ,0CAA0C,EAAE;AACjE;AAGA,IAAM,eAAe;AACd,SAAS,aAAa,MAA6B;AACxD,QAAM,UAAU,KAAK,KAAK;AAC1B,MAAI,CAAC,WAAW,CAAC,aAAa,KAAK,OAAO,EAAG,QAAO;AAGpD,SAAO,QAAQ,QAAQ,OAAO,KAAK,EAAE,QAAQ,OAAO,KAAK;AAC3D;AAKO,SAAS,QAAQ,KAAuB;AAC7C,MAAI,MAAM;AACV,WAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,IAAK,KAAI,IAAI,CAAC,IAAI,IAAK,OAAM,IAAI,CAAC;AAClE,SAAO;AACT;AAGO,SAAS,QAAQ,KAAuB;AAC7C,MAAI,MAAM;AACV,WAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,IAAK,KAAI,IAAI,CAAC,IAAI,IAAK,OAAM,IAAI,CAAC;AAClE,SAAO;AACT;AAOO,SAAS,cAAc,KAAyB;AACrD,MAAI,EAAE,eAAe,OAAQ,QAAO;AACpC,QAAM,MAAM,IAAI;AAEhB,MAAI,IAAI,SAAS,KAAK,EAAG,QAAO;AAChC,MAAI,IAAI,SAAS,oBAAK,KAAK,IAAI,SAAS,iCAAQ,EAAG,QAAO;AAC1D,MAAI,IAAI,SAAS,qBAAqB,EAAG,QAAO;AAChD,MAAI,IAAI,SAAS,UAAU,KAAK,IAAI,SAAS,kDAAe,KAAK,IAAI,SAAS,4CAAc,EAAG,QAAO;AACtG,MAAI,IAAI,SAAS,MAAM,KAAK,IAAI,SAAS,2BAAO,KAAK,IAAI,SAAS,2BAAO,EAAG,QAAO;AACnF,MAAI,IAAI,SAAS,iCAAQ,EAAG,QAAO;AACnC,MAAI,IAAI,SAAS,cAAI,MAAM,IAAI,SAAS,4BAAQ,KAAK,IAAI,SAAS,cAAI,GAAI,QAAO;AACjF,MAAI,IAAI,SAAS,0BAAM,KAAK,IAAI,SAAS,kCAAS,EAAG,QAAO;AAC5D,SAAO;AACT;","names":[]}
@@ -1,6 +1,6 @@
1
1
  "use strict";Object.defineProperty(exports, "__esModule", {value: true}); function _nullishCoalesce(lhs, rhsFn) { if (lhs != null) { return lhs; } else { return rhsFn(); } } function _optionalChain(ops) { let lastAccessLHS = undefined; let value = ops[0]; let i = 1; while (i < ops.length) { const op = ops[i]; const fn = ops[i + 1]; i += 2; if ((op === 'optionalAccess' || op === 'optionalCall') && value == null) { return undefined; } if (op === 'access' || op === 'optionalAccess') { lastAccessLHS = value; value = fn(value); } else if (op === 'call' || op === 'optionalCall') { value = fn((...args) => value.call(lastAccessLHS, ...args)); lastAccessLHS = undefined; } } return value; }
2
2
 
3
- var _chunkGZ7W64QUcjs = require('./chunk-GZ7W64QU.cjs');
3
+ var _chunkVQB5P5KJcjs = require('./chunk-VQB5P5KJ.cjs');
4
4
 
5
5
  // src/shared/pua.ts
6
6
  var BMP_SYMBOL_MAP = {
@@ -183,7 +183,7 @@ function buildTable(rows, options) {
183
183
  const cell = rows[rowIdx][cellIdx];
184
184
  anchorCols.add(colIdx);
185
185
  grid[rowIdx][colIdx] = {
186
- text: cell.text.trim(),
186
+ text: _optionalChain([options, 'optionalAccess', _2 => _2.keepEmptyParagraphs]) ? cell.text : cell.text.trim(),
187
187
  colSpan: cell.colSpan,
188
188
  rowSpan: cell.rowSpan
189
189
  };
@@ -219,7 +219,7 @@ function buildTableDirect(rows, numRows, options) {
219
219
  const c = _nullishCoalesce(cell.colAddr, () => ( 0));
220
220
  if (r >= numRows || c >= maxCols || r < 0 || c < 0) continue;
221
221
  anchorCols.add(c);
222
- grid[r][c] = { text: cell.text.trim(), colSpan: cell.colSpan, rowSpan: cell.rowSpan };
222
+ grid[r][c] = { text: _optionalChain([options, 'optionalAccess', _3 => _3.keepEmptyParagraphs]) ? cell.text : cell.text.trim(), colSpan: cell.colSpan, rowSpan: cell.rowSpan };
223
223
  for (let dr = 0; dr < cell.rowSpan; dr++) {
224
224
  for (let dc = 0; dc < cell.colSpan; dc++) {
225
225
  if (dr === 0 && dc === 0) continue;
@@ -235,9 +235,9 @@ function buildTableDirect(rows, numRows, options) {
235
235
  function trimAndReturn(grid, numRows, maxCols, anchorCols, options) {
236
236
  let effectiveCols = maxCols;
237
237
  while (effectiveCols > 0) {
238
- const colEmpty = grid.every((row) => !_optionalChain([row, 'access', _2 => _2[effectiveCols - 1], 'optionalAccess', _3 => _3.text, 'optionalAccess', _4 => _4.trim, 'call', _5 => _5()]));
238
+ const colEmpty = grid.every((row) => !_optionalChain([row, 'access', _4 => _4[effectiveCols - 1], 'optionalAccess', _5 => _5.text, 'optionalAccess', _6 => _6.trim, 'call', _7 => _7()]));
239
239
  if (!colEmpty) break;
240
- if (_optionalChain([options, 'optionalAccess', _6 => _6.keepAnchoredEmptyCols]) && anchorCols.has(effectiveCols - 1)) break;
240
+ if (_optionalChain([options, 'optionalAccess', _8 => _8.keepAnchoredEmptyCols]) && anchorCols.has(effectiveCols - 1)) break;
241
241
  effectiveCols--;
242
242
  }
243
243
  if (effectiveCols < maxCols && effectiveCols > 0) {
@@ -290,7 +290,7 @@ function flattenLayoutTables(blocks) {
290
290
  let totalTextLen = 0;
291
291
  for (let r = 0; r < numRows; r++) {
292
292
  for (let c = 0; c < numCols; c++) {
293
- const t = _optionalChain([cells, 'access', _7 => _7[r], 'optionalAccess', _8 => _8[c], 'optionalAccess', _9 => _9.text]) || "";
293
+ const t = _optionalChain([cells, 'access', _9 => _9[r], 'optionalAccess', _10 => _10[c], 'optionalAccess', _11 => _11.text]) || "";
294
294
  totalNewlines += (t.match(/\n/g) || []).length;
295
295
  totalTextLen += t.length;
296
296
  }
@@ -298,13 +298,13 @@ function flattenLayoutTables(blocks) {
298
298
  if (numCols < 4 && (totalNewlines > 5 || numRows <= 2 && totalTextLen > 300)) {
299
299
  for (let r = 0; r < numRows; r++) {
300
300
  for (let c = 0; c < numCols; c++) {
301
- const cell = _optionalChain([cells, 'access', _10 => _10[r], 'optionalAccess', _11 => _11[c]]);
301
+ const cell = _optionalChain([cells, 'access', _12 => _12[r], 'optionalAccess', _13 => _13[c]]);
302
302
  if (!cell) continue;
303
- if (_optionalChain([cell, 'access', _12 => _12.blocks, 'optionalAccess', _13 => _13.length])) {
303
+ if (_optionalChain([cell, 'access', _14 => _14.blocks, 'optionalAccess', _15 => _15.length])) {
304
304
  result.push(...flattenLayoutTables(cell.blocks));
305
305
  continue;
306
306
  }
307
- const cellText = _optionalChain([cell, 'access', _14 => _14.text, 'optionalAccess', _15 => _15.trim, 'call', _16 => _16()]);
307
+ const cellText = _optionalChain([cell, 'access', _16 => _16.text, 'optionalAccess', _17 => _17.trim, 'call', _18 => _18()]);
308
308
  if (!cellText) continue;
309
309
  for (const line of cellText.split("\n")) {
310
310
  const trimmed = line.trim();
@@ -324,7 +324,7 @@ var RUNNING_HEADER_MAX_LEN = 40;
324
324
  var RUNNING_HEADER_MIN_FREQ = 3;
325
325
  function isRunningHeaderCandidate(block) {
326
326
  if (block.type !== "paragraph" && block.type !== "heading") return false;
327
- const text = _optionalChain([block, 'access', _17 => _17.text, 'optionalAccess', _18 => _18.trim, 'call', _19 => _19()]);
327
+ const text = _optionalChain([block, 'access', _19 => _19.text, 'optionalAccess', _20 => _20.trim, 'call', _21 => _21()]);
328
328
  if (!text || text.length > RUNNING_HEADER_MAX_LEN) return false;
329
329
  return /^\d+\.\s/.test(text);
330
330
  }
@@ -407,7 +407,7 @@ function blocksToMarkdown(blocks) {
407
407
  if (!text) continue;
408
408
  if (/^\[별표\s*\d+/.test(text)) {
409
409
  const nextBlock = blocks[i + 1];
410
- if (_optionalChain([nextBlock, 'optionalAccess', _20 => _20.type]) === "paragraph" && nextBlock.text && /관련\)?$/.test(nextBlock.text)) {
410
+ if (_optionalChain([nextBlock, 'optionalAccess', _22 => _22.type]) === "paragraph" && nextBlock.text && /관련\)?$/.test(nextBlock.text)) {
411
411
  lines.push("", `## ${escapeGfm(text)} ${escapeGfm(nextBlock.text)}`, "");
412
412
  i++;
413
413
  } else {
@@ -420,10 +420,10 @@ function blocksToMarkdown(blocks) {
420
420
  continue;
421
421
  }
422
422
  const listIndent = block.listDepth ? " ".repeat(block.listDepth) : "";
423
- if (_optionalChain([block, 'access', _21 => _21.spans, 'optionalAccess', _22 => _22.length])) {
423
+ if (_optionalChain([block, 'access', _23 => _23.spans, 'optionalAccess', _24 => _24.length])) {
424
424
  let rendered = spansToMarkdown(block.spans);
425
425
  if (block.href) {
426
- const href = _chunkGZ7W64QUcjs.sanitizeHref.call(void 0, block.href);
426
+ const href = _chunkVQB5P5KJcjs.sanitizeHref.call(void 0, block.href);
427
427
  if (href) rendered = `[${rendered}](${href})`;
428
428
  }
429
429
  if (block.footnoteText) rendered += ` (\uC8FC: ${block.footnoteText})`;
@@ -431,7 +431,7 @@ function blocksToMarkdown(blocks) {
431
431
  continue;
432
432
  }
433
433
  if (block.href) {
434
- const href = _chunkGZ7W64QUcjs.sanitizeHref.call(void 0, block.href);
434
+ const href = _chunkVQB5P5KJcjs.sanitizeHref.call(void 0, block.href);
435
435
  if (href) text = `[${text}](${href})`;
436
436
  }
437
437
  if (block.footnoteText) {
@@ -466,13 +466,13 @@ function hasMergedCells(table) {
466
466
  function hasNestedTables(table) {
467
467
  for (const row of table.cells) {
468
468
  for (const cell of row) {
469
- if (_optionalChain([cell, 'access', _23 => _23.blocks, 'optionalAccess', _24 => _24.some, 'call', _25 => _25((b) => b.type === "table" && b.table)])) return true;
469
+ if (_optionalChain([cell, 'access', _25 => _25.blocks, 'optionalAccess', _26 => _26.some, 'call', _27 => _27((b) => b.type === "table" && b.table)])) return true;
470
470
  }
471
471
  }
472
472
  return false;
473
473
  }
474
474
  function cellInnerHtml(cell) {
475
- if (_optionalChain([cell, 'access', _26 => _26.blocks, 'optionalAccess', _27 => _27.length])) {
475
+ if (_optionalChain([cell, 'access', _28 => _28.blocks, 'optionalAccess', _29 => _29.length])) {
476
476
  return cell.blocks.map((b) => {
477
477
  if (b.type === "table" && b.table) {
478
478
  const cap = b.table.caption ? sanitizeText(b.table.caption) : "";
@@ -505,7 +505,7 @@ function tableToHtml(table) {
505
505
  const rowHtml = [];
506
506
  for (let c = 0; c < numCols; c++) {
507
507
  if (skip.has(`${r},${c}`)) continue;
508
- const cell = _optionalChain([cells, 'access', _28 => _28[r], 'optionalAccess', _29 => _29[c]]);
508
+ const cell = _optionalChain([cells, 'access', _30 => _30[r], 'optionalAccess', _31 => _31[c]]);
509
509
  if (!cell) continue;
510
510
  for (let dr = 0; dr < cell.rowSpan; dr++) {
511
511
  for (let dc = 0; dc < cell.colSpan; dc++) {
@@ -550,9 +550,9 @@ function tableToMarkdown(table) {
550
550
  for (let r = 0; r < numRows; r++) {
551
551
  for (let c = 0; c < numCols; c++) {
552
552
  if (skip.has(`${r},${c}`)) continue;
553
- const cell = _optionalChain([cells, 'access', _30 => _30[r], 'optionalAccess', _31 => _31[c]]);
553
+ const cell = _optionalChain([cells, 'access', _32 => _32[r], 'optionalAccess', _33 => _33[c]]);
554
554
  if (!cell) continue;
555
- display[r][c] = (_optionalChain([cell, 'access', _32 => _32.blocks, 'optionalAccess', _33 => _33.some, 'call', _34 => _34((b) => b.spans)]) ? cell.blocks.map((b) => b.type === "image" && b.text ? `![image](${b.text})` : b.spans ? spansToMarkdown(b.spans) : escapeGfm(sanitizeText(_nullishCoalesce(b.text, () => ( ""))))).filter(Boolean).join("<br>") : escapeGfm(sanitizeText(cell.text)).replace(/\n/g, "<br>")).replace(/\|/g, "\\|");
555
+ display[r][c] = (_optionalChain([cell, 'access', _34 => _34.blocks, 'optionalAccess', _35 => _35.some, 'call', _36 => _36((b) => b.spans)]) ? cell.blocks.map((b) => b.type === "image" && b.text ? `![image](${b.text})` : b.spans ? spansToMarkdown(b.spans) : escapeGfm(sanitizeText(_nullishCoalesce(b.text, () => ( ""))))).filter(Boolean).join("<br>") : escapeGfm(sanitizeText(cell.text)).replace(/\n/g, "<br>")).replace(/\|/g, "\\|");
556
556
  for (let dr = 0; dr < cell.rowSpan; dr++) {
557
557
  for (let dc = 0; dc < cell.colSpan; dc++) {
558
558
  if (dr === 0 && dc === 0) continue;
@@ -713,7 +713,7 @@ function encodePng(width, height, rgba) {
713
713
  return out;
714
714
  }
715
715
  function inlineImagesIntoMarkdown(markdown, images, opts) {
716
- const compress = _optionalChain([opts, 'optionalAccess', _35 => _35.compress]) !== false;
716
+ const compress = _optionalChain([opts, 'optionalAccess', _37 => _37.compress]) !== false;
717
717
  const uris = /* @__PURE__ */ new Map();
718
718
  for (const img of images) {
719
719
  let bytes = img.data;
@@ -751,4 +751,4 @@ function inlineImagesIntoMarkdown(markdown, images, opts) {
751
751
 
752
752
 
753
753
  exports.mapPuaText = mapPuaText; exports.MAX_COLS = MAX_COLS; exports.MAX_ROWS = MAX_ROWS; exports.buildTable = buildTable; exports.convertTableToText = convertTableToText; exports.flattenLayoutTables = flattenLayoutTables; exports.dedupeRunningHeaders = dedupeRunningHeaders; exports.blocksToMarkdown = blocksToMarkdown; exports.encodePng = encodePng; exports.inlineImagesIntoMarkdown = inlineImagesIntoMarkdown;
754
- //# sourceMappingURL=chunk-6GM2QLM3.cjs.map
754
+ //# sourceMappingURL=chunk-ZMKTKJZN.cjs.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["c:\\github_project\\kordoc\\dist\\chunk-ZMKTKJZN.cjs"],"names":[],"mappings":"AAAA;AACE;AACF,wDAA6B;AAC7B;AACA;AACA,IAAI,eAAe,EAAE;AACrB;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA;AACA,EAAE,GAAG,EAAE,MAAM;AACb;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,MAAM;AACb;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA;AACA,EAAE,EAAE,EAAE,QAAQ;AACd;AACA,EAAE,EAAE,EAAE,QAAQ;AACd;AACA,EAAE,EAAE,EAAE,QAAQ;AACd;AACA,EAAE,EAAE,EAAE,QAAQ;AACd;AACA;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA,EAAE,GAAG,EAAE,QAAQ;AACf;AACA;AACA,EAAE,EAAE,EAAE,QAAQ;AACd;AACA,EAAE,EAAE,EAAE,QAAQ;AACd;AACA,EAAE,EAAE,EAAE,QAAQ;AACd;AACA,EAAE,EAAE,EAAE,QAAQ;AACd;AACA,EAAE,EAAE,EAAE,QAAQ;AACd;AACA,EAAE,EAAE,EAAE,QAAQ;AACd;AACA,EAAE,EAAE,EAAE,QAAQ;AACd;AACA,EAAE,EAAE,EAAE;AACN;AACA,CAAC;AACD,IAAI,kBAAkB,EAAE;AACxB,EAAE,MAAM,EAAE,QAAQ;AAClB;AACA,EAAE,MAAM,EAAE,MAAM;AAChB;AACA,EAAE,MAAM,EAAE,QAAQ;AAClB;AACA,EAAE,MAAM,EAAE,QAAQ;AAClB;AACA,EAAE,MAAM,EAAE,QAAQ;AAClB;AACA,EAAE,MAAM,EAAE,QAAQ;AAClB;AACA,EAAE,MAAM,EAAE,QAAQ;AAClB;AACA,EAAE,MAAM,EAAE;AACV;AACA,CAAC;AACD,SAAS,UAAU,CAAC,IAAI,EAAE;AAC1B,EAAE,GAAG,CAAC,KAAK,GAAG,MAAM,GAAG,KAAK,GAAG,KAAK,EAAE;AACtC,IAAI,OAAO,cAAc,CAAC,KAAK,EAAE,KAAK,CAAC;AACvC,EAAE;AACF,EAAE,GAAG,CAAC,KAAK,GAAG,OAAO,GAAG,KAAK,GAAG,MAAM,EAAE;AACxC,IAAI,OAAO,iBAAiB,CAAC,IAAI,CAAC;AAClC,EAAE;AACF,EAAE,OAAO,KAAK,CAAC;AACf;AACA,SAAS,UAAU,CAAC,IAAI,EAAE;AAC1B,EAAE,IAAI,IAAI,EAAE,EAAE;AACd,EAAE,IAAI,CAAC,MAAM,GAAG,GAAG,IAAI,EAAE;AACzB,IAAI,MAAM,KAAK,EAAE,EAAE,CAAC,WAAW,CAAC,CAAC,CAAC;AAClC,IAAI,IAAI,oBAAG,UAAU,CAAC,IAAI,CAAE,UAAG,IAAE;AACjC,EAAE;AACF,EAAE,OAAO,GAAG;AACZ;AACA;AACA;AACA,IAAI,SAAS,EAAE,GAAG;AAClB,IAAI,SAAS,EAAE,GAAG;AAClB,SAAS,UAAU,CAAC,IAAI,EAAE,OAAO,EAAE;AACnC,EAAE,GAAG,CAAC,IAAI,CAAC,OAAO,EAAE,QAAQ,EAAE,KAAK,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,QAAQ,CAAC;AAC5D,EAAE,MAAM,QAAQ,EAAE,IAAI,CAAC,MAAM;AAC7B,EAAE,MAAM,QAAQ,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC,GAAG,EAAE,GAAG,GAAG,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC,QAAQ,IAAI,KAAK,EAAE,GAAG,CAAC,CAAC,QAAQ,IAAI,KAAK,CAAC,CAAC,CAAC;AACnG,EAAE,GAAG,CAAC,OAAO,EAAE,OAAO,gBAAgB,CAAC,IAAI,EAAE,OAAO,EAAE,OAAO,CAAC;AAC9D,EAAE,IAAI,QAAQ,EAAE,CAAC;AACjB,EAAE,MAAM,aAAa,EAAE,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,QAAQ,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;AAChE,EAAE,IAAI,CAAC,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,EAAE,OAAO,EAAE,MAAM,EAAE,EAAE;AACnD,IAAI,IAAI,OAAO,EAAE,CAAC;AAClB,IAAI,IAAI,CAAC,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,EAAE;AACrC,MAAM,MAAM,CAAC,OAAO,EAAE,SAAS,GAAG,YAAY,CAAC,MAAM,CAAC,CAAC,MAAM,CAAC,EAAE,MAAM,EAAE;AACxE,MAAM,GAAG,CAAC,OAAO,GAAG,QAAQ,EAAE,KAAK;AACnC,MAAM,IAAI,CAAC,IAAI,EAAE,EAAE,MAAM,EAAE,EAAE,EAAE,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,IAAI,CAAC,OAAO,EAAE,OAAO,CAAC,EAAE,CAAC,EAAE,EAAE;AAC9E,QAAQ,IAAI,CAAC,IAAI,EAAE,EAAE,MAAM,EAAE,EAAE,EAAE,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,IAAI,CAAC,OAAO,EAAE,QAAQ,CAAC,EAAE,CAAC,EAAE,EAAE;AACjF,UAAU,YAAY,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE,IAAI;AACnC,QAAQ;AACR,MAAM;AACN,MAAM,OAAO,GAAG,IAAI,CAAC,OAAO;AAC5B,MAAM,GAAG,CAAC,OAAO,EAAE,OAAO,EAAE,QAAQ,EAAE,MAAM;AAC5C,IAAI;AACJ,EAAE;AACF,EAAE,GAAG,CAAC,QAAQ,IAAI,CAAC,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,EAAE,IAAI,EAAE,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC,EAAE,SAAS,EAAE,MAAM,CAAC;AAC7E,EAAE,MAAM,KAAK,EAAE,KAAK,CAAC,IAAI;AACzB,IAAI,EAAE,MAAM,EAAE,QAAQ,CAAC;AACvB,IAAI,CAAC,EAAE,GAAG,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,QAAQ,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,EAAE,IAAI,EAAE,EAAE,EAAE,OAAO,EAAE,CAAC,EAAE,OAAO,EAAE,EAAE,CAAC,CAAC;AACtF,EAAE,CAAC;AACH,EAAE,MAAM,SAAS,EAAE,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,QAAQ,CAAC,EAAE,CAAC,EAAE,GAAG,KAAK,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;AACpF,EAAE,MAAM,WAAW,kBAAkB,IAAI,GAAG,CAAC,CAAC;AAC9C,EAAE,IAAI,CAAC,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,EAAE,OAAO,EAAE,MAAM,EAAE,EAAE;AACnD,IAAI,IAAI,OAAO,EAAE,CAAC;AAClB,IAAI,IAAI,QAAQ,EAAE,CAAC;AACnB,IAAI,MAAM,CAAC,OAAO,EAAE,QAAQ,GAAG,QAAQ,EAAE,IAAI,CAAC,MAAM,CAAC,CAAC,MAAM,EAAE;AAC9D,MAAM,MAAM,CAAC,OAAO,EAAE,QAAQ,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,MAAM,CAAC,EAAE,MAAM,EAAE;AACnE,MAAM,GAAG,CAAC,OAAO,GAAG,OAAO,EAAE,KAAK;AAClC,MAAM,MAAM,KAAK,EAAE,IAAI,CAAC,MAAM,CAAC,CAAC,OAAO,CAAC;AACxC,MAAM,UAAU,CAAC,GAAG,CAAC,MAAM,CAAC;AAC5B,MAAM,IAAI,CAAC,MAAM,CAAC,CAAC,MAAM,EAAE,EAAE;AAC7B,QAAQ,IAAI,kBAAE,OAAO,6BAAE,sBAAoB,EAAE,IAAI,CAAC,KAAK,EAAE,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;AACzE,QAAQ,OAAO,EAAE,IAAI,CAAC,OAAO;AAC7B,QAAQ,OAAO,EAAE,IAAI,CAAC;AACtB,MAAM,CAAC;AACP,MAAM,IAAI,CAAC,IAAI,EAAE,EAAE,MAAM,EAAE,EAAE,EAAE,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,IAAI,CAAC,OAAO,EAAE,OAAO,CAAC,EAAE,CAAC,EAAE,EAAE;AAC9E,QAAQ,IAAI,CAAC,IAAI,EAAE,EAAE,MAAM,EAAE,EAAE,EAAE,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,IAAI,CAAC,OAAO,EAAE,OAAO,CAAC,EAAE,CAAC,EAAE,EAAE;AAChF,UAAU,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE,IAAI;AAC/B,QAAQ;AACR,MAAM;AACN,MAAM,OAAO,GAAG,IAAI,CAAC,OAAO;AAC5B,MAAM,OAAO,EAAE;AACf,IAAI;AACJ,EAAE;AACF,EAAE,OAAO,aAAa,CAAC,IAAI,EAAE,OAAO,EAAE,OAAO,EAAE,UAAU,EAAE,OAAO,CAAC;AACnE;AACA,SAAS,gBAAgB,CAAC,IAAI,EAAE,OAAO,EAAE,OAAO,EAAE;AAClD,EAAE,IAAI,QAAQ,EAAE,CAAC;AACjB,EAAE,IAAI,CAAC,MAAM,IAAI,GAAG,IAAI,EAAE;AAC1B,IAAI,IAAI,CAAC,MAAM,KAAK,GAAG,GAAG,EAAE;AAC5B,MAAM,MAAM,IAAI,EAAE,kBAAC,IAAI,CAAC,OAAQ,UAAG,GAAC,EAAE,EAAE,IAAI,CAAC,OAAO;AACpD,MAAM,GAAG,CAAC,IAAI,EAAE,OAAO,EAAE,QAAQ,EAAE,GAAG;AACtC,IAAI;AACJ,EAAE;AACF,EAAE,GAAG,CAAC,QAAQ,EAAE,QAAQ,EAAE,QAAQ,EAAE,QAAQ;AAC5C,EAAE,GAAG,CAAC,QAAQ,IAAI,CAAC,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,EAAE,IAAI,EAAE,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC,EAAE,SAAS,EAAE,MAAM,CAAC;AAC7E,EAAE,MAAM,KAAK,EAAE,KAAK,CAAC,IAAI;AACzB,IAAI,EAAE,MAAM,EAAE,QAAQ,CAAC;AACvB,IAAI,CAAC,EAAE,GAAG,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,QAAQ,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,EAAE,IAAI,EAAE,EAAE,EAAE,OAAO,EAAE,CAAC,EAAE,OAAO,EAAE,EAAE,CAAC,CAAC;AACtF,EAAE,CAAC;AACH,EAAE,MAAM,WAAW,kBAAkB,IAAI,GAAG,CAAC,CAAC;AAC9C,EAAE,IAAI,CAAC,MAAM,IAAI,GAAG,IAAI,EAAE;AAC1B,IAAI,IAAI,CAAC,MAAM,KAAK,GAAG,GAAG,EAAE;AAC5B,MAAM,MAAM,EAAE,mBAAE,IAAI,CAAC,OAAQ,UAAG,GAAC;AACjC,MAAM,MAAM,EAAE,mBAAE,IAAI,CAAC,OAAQ,UAAG,GAAC;AACjC,MAAM,GAAG,CAAC,EAAE,GAAG,QAAQ,GAAG,EAAE,GAAG,QAAQ,GAAG,EAAE,EAAE,EAAE,GAAG,EAAE,EAAE,CAAC,EAAE,QAAQ;AAClE,MAAM,UAAU,CAAC,GAAG,CAAC,CAAC,CAAC;AACvB,MAAM,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE,EAAE,IAAI,kBAAE,OAAO,6BAAE,sBAAoB,EAAE,IAAI,CAAC,KAAK,EAAE,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,EAAE,OAAO,EAAE,IAAI,CAAC,OAAO,EAAE,OAAO,EAAE,IAAI,CAAC,QAAQ,CAAC;AACtI,MAAM,IAAI,CAAC,IAAI,GAAG,EAAE,CAAC,EAAE,GAAG,EAAE,IAAI,CAAC,OAAO,EAAE,EAAE,EAAE,EAAE;AAChD,QAAQ,IAAI,CAAC,IAAI,GAAG,EAAE,CAAC,EAAE,GAAG,EAAE,IAAI,CAAC,OAAO,EAAE,EAAE,EAAE,EAAE;AAClD,UAAU,GAAG,CAAC,GAAG,IAAI,EAAE,GAAG,GAAG,IAAI,CAAC,EAAE,QAAQ;AAC5C,UAAU,GAAG,CAAC,EAAE,EAAE,GAAG,EAAE,QAAQ,GAAG,EAAE,EAAE,GAAG,EAAE,OAAO,EAAE;AACpD,YAAY,IAAI,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,IAAI,EAAE,EAAE,EAAE,OAAO,EAAE,CAAC,EAAE,OAAO,EAAE,EAAE,CAAC;AACvE,UAAU;AACV,QAAQ;AACR,MAAM;AACN,IAAI;AACJ,EAAE;AACF,EAAE,OAAO,aAAa,CAAC,IAAI,EAAE,OAAO,EAAE,OAAO,EAAE,UAAU,EAAE,OAAO,CAAC;AACnE;AACA,SAAS,aAAa,CAAC,IAAI,EAAE,OAAO,EAAE,OAAO,EAAE,UAAU,EAAE,OAAO,EAAE;AACpE,EAAE,IAAI,cAAc,EAAE,OAAO;AAC7B,EAAE,MAAM,CAAC,cAAc,EAAE,CAAC,EAAE;AAC5B,IAAI,MAAM,SAAS,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,GAAG,EAAE,GAAG,iBAAC,GAAG,qBAAC,cAAc,EAAE,CAAC,CAAC,6BAAE,IAAI,6BAAE,IAAI,mBAAC,GAAC,CAAC;AAC/E,IAAI,GAAG,CAAC,CAAC,QAAQ,EAAE,KAAK;AACxB,IAAI,GAAG,iBAAC,OAAO,6BAAE,wBAAsB,GAAG,UAAU,CAAC,GAAG,CAAC,cAAc,EAAE,CAAC,CAAC,EAAE,KAAK;AAClF,IAAI,aAAa,EAAE;AACnB,EAAE;AACF,EAAE,GAAG,CAAC,cAAc,EAAE,QAAQ,GAAG,cAAc,EAAE,CAAC,EAAE;AACpD,IAAI,MAAM,QAAQ,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,GAAG,CAAC,KAAK,CAAC,CAAC,EAAE,aAAa,CAAC,CAAC;AAClE,IAAI,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,IAAI,EAAE,aAAa,EAAE,KAAK,EAAE,OAAO,EAAE,SAAS,EAAE,QAAQ,EAAE,EAAE,CAAC;AACzF,EAAE;AACF,EAAE,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,KAAK,EAAE,IAAI,EAAE,SAAS,EAAE,QAAQ,EAAE,EAAE,CAAC;AAC9E;AACA,SAAS,kBAAkB,CAAC,IAAI,EAAE;AAClC,EAAE,OAAO,IAAI,CAAC,GAAG;AACjB,IAAI,CAAC,GAAG,EAAE,GAAG,GAAG,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,OAAO,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC,OAAO,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,KAAK;AAC/G,EAAE,CAAC,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC;AAC9B;AACA,SAAS,SAAS,CAAC,IAAI,EAAE;AACzB,EAAE,MAAM,IAAI,EAAE,MAAM,CAAC,YAAY,CAAC,CAAC,CAAC;AACpC,EAAE,MAAM,MAAM,EAAE,CAAC,CAAC;AAClB,EAAE,MAAM,OAAO,EAAE,IAAI,CAAC,OAAO,CAAC,4FAA4F,EAAE,CAAC,CAAC,EAAE,GAAG;AACnI,IAAI,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC;AACjB,IAAI,OAAO,IAAI,EAAE,CAAC,KAAK,CAAC,OAAO,EAAE,CAAC,EAAE,EAAE,GAAG;AACzC,EAAE,CAAC,CAAC;AACJ,EAAE,MAAM,QAAQ,EAAE,MAAM,CAAC,OAAO,CAAC,WAAW,EAAE,MAAM,CAAC;AACrD,EAAE,OAAO,OAAO,CAAC,OAAO,CAAC,IAAI,MAAM,CAAC,IAAI,EAAE,SAAS,EAAE,GAAG,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE,GAAG,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC;AAC3F;AACA,IAAI,sBAAsB,EAAE,sOAAsO;AAClQ,SAAS,YAAY,CAAC,IAAI,EAAE;AAC5B,EAAE,IAAI,OAAO,EAAE,UAAU,CAAC,IAAI,CAAC,CAAC,OAAO,CAAC,yBAAyB,EAAE,EAAE,CAAC,CAAC,OAAO,CAAC,qBAAqB,EAAE,EAAE,CAAC,CAAC,OAAO,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC;AACrI,EAAE,GAAG,CAAC,MAAM,CAAC,OAAO,GAAG,GAAG,GAAG,MAAM,CAAC,QAAQ,CAAC,GAAG,CAAC,EAAE;AACnD,IAAI,MAAM,OAAO,EAAE,MAAM,CAAC,KAAK,CAAC,GAAG,CAAC;AACpC,IAAI,MAAM,sBAAsB,EAAE,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC,OAAO,IAAI,EAAE,GAAG,8BAA8B,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM;AACvH,IAAI,GAAG,CAAC,MAAM,CAAC,OAAO,GAAG,EAAE,GAAG,sBAAsB,EAAE,MAAM,CAAC,OAAO,GAAG,GAAG,EAAE;AAC5E,MAAM,OAAO,EAAE,MAAM,CAAC,IAAI,CAAC,EAAE,CAAC;AAC9B,IAAI;AACJ,EAAE;AACF,EAAE,OAAO,MAAM;AACf;AACA,SAAS,mBAAmB,CAAC,MAAM,EAAE;AACrC,EAAE,MAAM,OAAO,EAAE,CAAC,CAAC;AACnB,EAAE,IAAI,CAAC,MAAM,MAAM,GAAG,MAAM,EAAE;AAC9B,IAAI,GAAG,CAAC,KAAK,CAAC,KAAK,IAAI,QAAQ,GAAG,CAAC,KAAK,CAAC,KAAK,EAAE;AAChD,MAAM,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC;AACxB,MAAM,QAAQ;AACd,IAAI;AACJ,IAAI,MAAM,EAAE,IAAI,EAAE,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,MAAM,EAAE,EAAE,KAAK,CAAC,KAAK;AAC/D,IAAI,GAAG,CAAC,QAAQ,IAAI,EAAE,GAAG,QAAQ,IAAI,CAAC,EAAE;AACxC,MAAM,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC;AACxB,MAAM,QAAQ;AACd,IAAI;AACJ,IAAI,GAAG,CAAC,QAAQ,GAAG,CAAC,EAAE;AACtB,MAAM,IAAI,cAAc,EAAE,CAAC;AAC3B,MAAM,IAAI,aAAa,EAAE,CAAC;AAC1B,MAAM,IAAI,CAAC,IAAI,EAAE,EAAE,CAAC,EAAE,EAAE,EAAE,OAAO,EAAE,CAAC,EAAE,EAAE;AACxC,QAAQ,IAAI,CAAC,IAAI,EAAE,EAAE,CAAC,EAAE,EAAE,EAAE,OAAO,EAAE,CAAC,EAAE,EAAE;AAC1C,UAAU,MAAM,EAAE,kBAAE,KAAK,qBAAC,CAAC,CAAC,8BAAE,CAAC,CAAC,CAAC,+BAAE,OAAK,GAAG,EAAE;AAC7C,UAAU,cAAc,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC,MAAM;AACxD,UAAU,aAAa,GAAG,CAAC,CAAC,MAAM;AAClC,QAAQ;AACR,MAAM;AACN,MAAM,GAAG,CAAC,QAAQ,EAAE,EAAE,GAAG,CAAC,cAAc,EAAE,EAAE,GAAG,QAAQ,GAAG,EAAE,GAAG,aAAa,EAAE,GAAG,CAAC,EAAE;AACpF,QAAQ,IAAI,CAAC,IAAI,EAAE,EAAE,CAAC,EAAE,EAAE,EAAE,OAAO,EAAE,CAAC,EAAE,EAAE;AAC1C,UAAU,IAAI,CAAC,IAAI,EAAE,EAAE,CAAC,EAAE,EAAE,EAAE,OAAO,EAAE,CAAC,EAAE,EAAE;AAC5C,YAAY,MAAM,KAAK,kBAAE,KAAK,uBAAC,CAAC,CAAC,8BAAE,CAAC,CAAC,GAAC;AACtC,YAAY,GAAG,CAAC,CAAC,IAAI,EAAE,QAAQ;AAC/B,YAAY,GAAG,iBAAC,IAAI,uBAAC,MAAM,+BAAE,QAAM,EAAE;AACrC,cAAc,MAAM,CAAC,IAAI,CAAC,GAAG,mBAAmB,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;AAC9D,cAAc,QAAQ;AACtB,YAAY;AACZ,YAAY,MAAM,SAAS,kBAAE,IAAI,uBAAC,IAAI,+BAAE,IAAI,qBAAC,GAAC;AAC9C,YAAY,GAAG,CAAC,CAAC,QAAQ,EAAE,QAAQ;AACnC,YAAY,IAAI,CAAC,MAAM,KAAK,GAAG,QAAQ,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE;AACrD,cAAc,MAAM,QAAQ,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC;AACzC,cAAc,GAAG,CAAC,CAAC,OAAO,EAAE,QAAQ;AACpC,cAAc,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,WAAW,EAAE,IAAI,EAAE,OAAO,EAAE,UAAU,EAAE,KAAK,CAAC,WAAW,CAAC,CAAC;AAC7F,YAAY;AACZ,UAAU;AACV,QAAQ;AACR,QAAQ,QAAQ;AAChB,MAAM;AACN,IAAI;AACJ,IAAI,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC;AACtB,EAAE;AACF,EAAE,OAAO,MAAM;AACf;AACA,IAAI,uBAAuB,EAAE,EAAE;AAC/B,IAAI,wBAAwB,EAAE,CAAC;AAC/B,SAAS,wBAAwB,CAAC,KAAK,EAAE;AACzC,EAAE,GAAG,CAAC,KAAK,CAAC,KAAK,IAAI,YAAY,GAAG,KAAK,CAAC,KAAK,IAAI,SAAS,EAAE,OAAO,KAAK;AAC1E,EAAE,MAAM,KAAK,kBAAE,KAAK,uBAAC,IAAI,+BAAE,IAAI,qBAAC,GAAC;AACjC,EAAE,GAAG,CAAC,CAAC,KAAK,GAAG,IAAI,CAAC,OAAO,EAAE,sBAAsB,EAAE,OAAO,KAAK;AACjE,EAAE,OAAO,UAAU,CAAC,IAAI,CAAC,IAAI,CAAC;AAC9B;AACA,SAAS,oBAAoB,CAAC,MAAM,EAAE;AACtC,EAAE,MAAM,KAAK,kBAAkB,IAAI,GAAG,CAAC,CAAC;AACxC,EAAE,IAAI,CAAC,MAAM,MAAM,GAAG,MAAM,EAAE;AAC9B,IAAI,GAAG,CAAC,CAAC,wBAAwB,CAAC,KAAK,CAAC,EAAE,QAAQ;AAClD,IAAI,MAAM,KAAK,EAAE,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;AAClC,IAAI,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,kBAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAE,UAAG,GAAC,EAAE,EAAE,CAAC,CAAC;AAC7C,EAAE;AACF,EAAE,MAAM,KAAK,kBAAkB,IAAI,GAAG,CAAC,CAAC;AACxC,EAAE,MAAM,OAAO,EAAE,CAAC,CAAC;AACnB,EAAE,IAAI,CAAC,MAAM,MAAM,GAAG,MAAM,EAAE;AAC9B,IAAI,GAAG,CAAC,wBAAwB,CAAC,KAAK,CAAC,EAAE;AACzC,MAAM,MAAM,KAAK,EAAE,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;AACpC,MAAM,GAAG,CAAC,kBAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAE,UAAG,GAAC,EAAE,GAAG,uBAAuB,EAAE;AAC5D,QAAQ,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,QAAQ;AACpC,QAAQ,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC;AACtB,MAAM;AACN,IAAI;AACJ,IAAI,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC;AACtB,EAAE;AACF,EAAE,OAAO,MAAM;AACf;AACA,SAAS,eAAe,CAAC,KAAK,EAAE;AAChC,EAAE,IAAI,IAAI,EAAE,EAAE;AACd,EAAE,IAAI,CAAC,MAAM,EAAE,GAAG,KAAK,EAAE;AACzB,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,QAAQ;AACzB,IAAI,IAAI,OAAO,EAAE,CAAC,CAAC,KAAK,EAAE,IAAI,EAAE,CAAC,CAAC,KAAK,GAAG,CAAC,CAAC,OAAO,EAAE,MAAM,EAAE,CAAC,CAAC,KAAK,EAAE,KAAK,EAAE,CAAC,CAAC,OAAO,EAAE,IAAI,EAAE,EAAE;AACh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file":"C:\\github_project\\kordoc\\dist\\chunk-ZMKTKJZN.cjs","sourcesContent":[null]}
@@ -10,4 +10,4 @@ export {
10
10
  HEADING_RATIO_H2,
11
11
  HEADING_RATIO_H3
12
12
  };
13
- //# sourceMappingURL=chunk-LM7C6HI2.js.map
13
+ //# sourceMappingURL=chunk-ZSL6MQVE.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../src/types.ts"],"sourcesContent":["/** kordoc 공통 타입 정의 */\r\n\r\n// ─── 중간 표현 (Intermediate Representation) ─────────\r\n\r\nexport interface CellContext {\r\n text: string\r\n colSpan: number\r\n rowSpan: number\r\n /** HWP5 셀 열 주소 (0-based) — 병합 테이블 배치용 */\r\n colAddr?: number\r\n /** HWP5 셀 행 주소 (0-based) — 병합 테이블 배치용 */\r\n rowAddr?: number\r\n}\r\n\r\n/** 블록 타입 — v2.0에서 heading, list, image, separator 추가 */\r\nexport type IRBlockType = \"paragraph\" | \"table\" | \"heading\" | \"list\" | \"image\" | \"separator\"\r\n\r\n/** 인라인 강조 run-span — 문단 텍스트를 서식 단위로 쪼갠 조각 (텍스트 연결 = block.text) */\r\nexport interface IRSpan {\r\n text: string\r\n bold?: boolean\r\n italic?: boolean\r\n strike?: boolean\r\n code?: boolean\r\n}\r\n\r\nexport interface IRBlock {\r\n type: IRBlockType\r\n text?: string\r\n table?: IRTable\r\n /** 헤딩 레벨 (1-6), type=\"heading\"일 때 사용 */\r\n level?: number\r\n /** 원본 페이지 번호 (1-based) */\r\n pageNumber?: number\r\n /** 바운딩 박스 — PDF에서만 제공 */\r\n bbox?: BoundingBox\r\n /** 텍스트 스타일 정보 (선택) */\r\n style?: InlineStyle\r\n /** 리스트 타입, type=\"list\"일 때 사용 */\r\n listType?: \"ordered\" | \"unordered\"\r\n /** 중첩 리스트 아이템 */\r\n children?: IRBlock[]\r\n /** 하이퍼링크 URL */\r\n href?: string\r\n /** 각주/미주 텍스트 (인라인 삽입용) */\r\n footnoteText?: string\r\n /** 이미지 데이터 (type=\"image\"일 때) */\r\n imageData?: ImageData\r\n /** 인라인 강조 run-span (선택 — kordoc 생성 hwpx 왕복 채널 + 외래 실속성 볼드/이탤릭).\r\n * 존재하면 마크다운 변환이 **·*·` 마커를 재방출한다 */\r\n spans?: IRSpan[]\r\n /** 인용문 문단 (선택 — kordoc 생성 hwpx 왕복 채널) — 마크다운 변환이 \"> \" 접두 재방출 */\r\n quote?: boolean\r\n /**\r\n * 문단 들여쓰기(HWPUNIT, 선택) — HWPX paraPr `<hh:margin>` 자식요소형 hc:left\r\n * (+양수 hc:intent 첫줄분). 마크다운 방출엔 쓰지 않는 관찰 슬롯 — gongmun 리스트\r\n * depth 재유도·양식 분석 등 소비자 몫 (v4.0.4)\r\n */\r\n indent?: number\r\n /**\r\n * gongmun 리스트 단계(1~7, 선택) — indent를 levelIndent 단위로 역산한 소비 결과\r\n * (v4.0.5). md 리스트 문법과 충돌하는 부호('- '·'1) ') 문단에만 채워지며,\r\n * 마크다운 변환이 2칸/단계 선행 공백을 방출해 재생성 시 depth가 복원된다\r\n */\r\n listDepth?: number\r\n}\r\n\r\n/** 추출된 이미지 바이너리 데이터 */\r\nexport interface ImageData {\r\n /** 이미지 바이너리 */\r\n data: Uint8Array\r\n /** MIME 타입 (image/png, image/jpeg, image/gif, image/bmp, image/wmf, image/emf) */\r\n mimeType: string\r\n /** 원본 파일명 (있는 경우) */\r\n filename?: string\r\n}\r\n\r\n/** 바운딩 박스 — PDF 포인트 단위 (72pt = 1인치) */\r\nexport interface BoundingBox {\r\n page: number\r\n x: number\r\n y: number\r\n width: number\r\n height: number\r\n}\r\n\r\n/** 인라인 텍스트 스타일 */\r\nexport interface InlineStyle {\r\n bold?: boolean\r\n italic?: boolean\r\n /** 취소선 — 법령 개정문 등의 삭제 표시. 판정은 취소선 모양 whitelist (비트만 믿으면 오탐) */\r\n strike?: boolean\r\n fontSize?: number\r\n fontName?: string\r\n}\r\n\r\nexport interface IRTable {\r\n rows: number\r\n cols: number\r\n cells: IRCell[][]\r\n /** 첫 행을 헤더로 렌더링할지 여부 (현재: rows > 1이면 true — 의미적 감지가 아닌 레이아웃 힌트) */\r\n hasHeader: boolean\r\n /** 표 캡션 (예: \"표 1. 부서별 예산\") — v3.0 */\r\n caption?: string\r\n /**\r\n * 캡션 내부 블록 콘텐츠 — v4.2.8 (#55).\r\n * 캡션 안 중첩표·문단을 구조 그대로, 원문 순서대로 보존한다 (IRCell.blocks와 같은 계약).\r\n * 표가 있을 때만 채워지며, caption은 하위 호환용 평탄화 문자열로 계속 제공된다.\r\n */\r\n captionBlocks?: IRBlock[]\r\n}\r\n\r\nexport interface IRCell {\r\n text: string\r\n colSpan: number\r\n rowSpan: number\r\n /**\r\n * 셀 내부 블록 콘텐츠 — v3.0.\r\n * 중첩 표·이미지 등 구조 콘텐츠(또는 왕복 채널 span 문단)가 있는 셀에만 채워지며,\r\n * 그 안의 문단·표·이미지를 문서(원문) 순서대로 보존한다. 다중 문단뿐인 평문 셀은\r\n * blocks 없이 text 평탄화(문단을 `\\n`으로 결합)로만 제공된다 (blocks 무게 억제, #57).\r\n * 표와 텍스트가 한 줄에 번갈아 놓인 셀도 배치 순서를 따른다 (v4.2.3, #49).\r\n * blocks가 있으면 text는 blocks의 평탄화 텍스트(하위 호환용)다.\r\n */\r\n blocks?: IRBlock[]\r\n /** 제목 셀 여부 (HWP5 width_ref bit2 / HWPX header 속성) — v3.0 */\r\n isHeader?: boolean\r\n}\r\n\r\n// ─── 메타데이터 ─────────────────────────────────────\r\n\r\n/** 문서 메타데이터 — 각 포맷에서 추출 가능한 필드만 채워짐 */\r\nexport interface DocumentMetadata {\r\n /** 문서 제목 */\r\n title?: string\r\n /** 작성자 */\r\n author?: string\r\n /** 작성 프로그램 (예: \"한글 2020\", \"Adobe Acrobat\") */\r\n creator?: string\r\n /** 생성일시 (ISO 8601) */\r\n createdAt?: string\r\n /** 수정일시 (ISO 8601) */\r\n modifiedAt?: string\r\n /** 페이지/섹션 수 */\r\n pageCount?: number\r\n /** 문서 포맷 버전 (예: HWP \"5.1.0.1\") */\r\n version?: string\r\n /** 설명 */\r\n description?: string\r\n /** 키워드 */\r\n keywords?: string[]\r\n}\r\n\r\n// ─── 파싱 옵션 ──────────────────────────────────────\r\n\r\n/** 파싱 옵션 — parse() 함수에 전달 */\r\nexport interface ParseOptions {\r\n /**\r\n * 파싱할 페이지/섹션 범위 (1-based).\r\n * - 배열: [1, 2, 3]\r\n * - 문자열: \"1-3\", \"1,3,5-7\"\r\n *\r\n * PDF: 정확한 페이지 단위. HWP/HWPX: 섹션 단위 근사치.\r\n */\r\n pages?: number[] | string\r\n /** 이미지 기반 PDF OCR (선택).\r\n * - `true`: 내장 엔진(PP-OCRv5 korean, ~18MB 자동 다운로드)으로 OCR 필요 판정\r\n * 페이지만 인식 (스캔 페이지·글꼴 매핑 깨진 페이지). 정상 페이지는 파싱 결과 유지.\r\n * - `\"force\"`: 전 페이지를 내장 엔진으로 강제 OCR.\r\n * - 함수: 사용자 제공 OcrProvider (Claude Vision·Tesseract 등) — 판정은 `true`와 동일. */\r\n ocr?: boolean | \"force\" | OcrProvider\r\n /** 진행률 콜백 — current: 현재 페이지/섹션, total: 전체 수 */\r\n onProgress?: (current: number, total: number) => void\r\n /** PDF 머리글/바닥글 자동 제거 */\r\n removeHeaderFooter?: boolean\r\n /** 표 오른쪽 끝의 빈 열(서식 문서의 입력란) 보존 (#47).\r\n * 기본 false: 마크다운 가독성을 위해 후행 빈 열을 트림.\r\n * 양식 인식 경로(parse_form·fill)는 내부적으로 항상 켠다. */\r\n keepTrailingEmptyCols?: boolean\r\n /** 빈 문단(텍스트 없는 hp:p) 보존 (#57). 기본 false: 종전대로 빈 문단 제거.\r\n * 켜면 본문은 `text: \"\"` paragraph 블록으로, 표 셀은 빈 줄로 순서대로 보존해\r\n * \"원문 문단 수 = 줄 수\" 대응을 유지한다 (행 줄맞춤 서식 문서용).\r\n * 개체(표·이미지·글상자)만 있는 문단은 개체 출력이 따로 있어 대상이 아니다.\r\n * 현재 HWPX 경로 적용. */\r\n keepEmptyParagraphs?: boolean\r\n /** 원본 파일 경로 (DRM COM fallback에 필요, 내부 전용) */\r\n filePath?: string\r\n /**\r\n * PDF 수식 OCR 활성화 (기본 false).\r\n *\r\n * 활성화 시 각 PDF 페이지를 이미지로 렌더링 → YOLOv8 기반 수식 영역 검출 →\r\n * TrOCR 기반 LaTeX 인식. 감지된 수식은 `$...$` (inline) / `$$...$$` (display) 로\r\n * 블록 텍스트에 삽입된다.\r\n *\r\n * 필수 optional 의존성: `onnxruntime-node`, `@huggingface/transformers`,\r\n * `@hyzyla/pdfium`, `sharp`. 미설치 시 parse 에 실패하지 않고 **경고만** 남기고\r\n * 수식 인식은 skip 한다 (일반 텍스트 추출은 정상 동작).\r\n *\r\n * 모델(~155MB) 은 첫 사용 시 HuggingFace 에서 자동 다운로드 되어\r\n * `~/.cache/kordoc/models/pix2text/` 에 SHA-256 검증과 함께 저장된다.\r\n */\r\n formulaOcr?: boolean\r\n /**\r\n * 레이아웃 표 페이지 반복 헤더(러닝 헤더) 정리 휴리스틱 활성화 (기본 false).\r\n *\r\n * 구형 HWP5 문서가 페이지마다 재삽입한 짧은 번호매김 러닝 헤더\r\n * (\"2. 과제 구축 내용\")를 최초 1회만 남기고 이후 중복을 제거한다.\r\n * 다만 페이지 위치 정보가 없는 HWP5 특성상 이 휴리스틱은 정당하게 반복되는\r\n * 번호매김 문단(예: 붙임/별지별 재번호 \"1. 목적\")도 삭제할 수 있어 opt-in 으로\r\n * 둔다. 활성 시 제거가 발생하면 HIDDEN_TEXT_FILTERED 경고를 남긴다.\r\n */\r\n dedupeRunningHeaders?: boolean\r\n /**\r\n * 추출된 이미지를 마크다운에 base64 data URI 로 인라인 (기본 false).\r\n *\r\n * 활성화 시 `![image](image_001.bmp)` 참조를 `![image](data:image/png;base64,...)` 로\r\n * 치환한다. 임베드된 BMP 는 PNG 로 무손실 압축 후 인라인하여 용량을 크게 줄인다.\r\n * 별도 이미지 파일 없이 자체 완결형 마크다운이 되어 MCP/AI 에이전트 소비에 적합하다.\r\n * (현재 HWP5 경로 지원)\r\n */\r\n inlineImages?: boolean\r\n}\r\n\r\n// ─── 파싱 경고 ──────────────────────────────────────\r\n\r\n/** 파싱 중 스킵/실패한 요소 보고 */\r\nexport interface ParseWarning {\r\n /** 관련 페이지 번호 (알 수 있는 경우) */\r\n page?: number\r\n /** 경고 메시지 */\r\n message: string\r\n /** 구조화된 경고 코드 */\r\n code: WarningCode\r\n}\r\n\r\nexport type WarningCode =\r\n | \"SKIPPED_IMAGE\"\r\n | \"SKIPPED_OLE\"\r\n | \"TRUNCATED_TABLE\"\r\n | \"OCR_FALLBACK\"\r\n | \"UNSUPPORTED_ELEMENT\"\r\n | \"BROKEN_ZIP_RECOVERY\"\r\n | \"HIDDEN_TEXT_FILTERED\"\r\n | \"MALFORMED_XML\"\r\n | \"PARTIAL_PARSE\"\r\n | \"LENIENT_CFB_RECOVERY\"\r\n | \"NEEDS_OCR\"\r\n | \"OCR_FAILED\"\r\n | \"OCR_APPLIED\"\r\n | \"COM_EMPTY\"\r\n | \"DRM_COM_FALLBACK\"\r\n\r\n/** 문서 구조 (헤딩 트리) */\r\nexport interface OutlineItem {\r\n level: number\r\n text: string\r\n pageNumber?: number\r\n}\r\n\r\n// ─── 에러 코드 ──────────────────────────────────────\r\n\r\n/** 구조화된 에러 코드 — 프로그래밍적 에러 핸들링용 */\r\nexport type ErrorCode =\r\n | \"EMPTY_INPUT\"\r\n | \"UNSUPPORTED_FORMAT\"\r\n | \"ENCRYPTED\"\r\n | \"DRM_PROTECTED\"\r\n | \"CORRUPTED\"\r\n | \"DECOMPRESSION_BOMB\"\r\n | \"ZIP_BOMB\"\r\n | \"IMAGE_BASED_PDF\"\r\n | \"NO_SECTIONS\"\r\n | \"PARSE_ERROR\"\r\n | \"MISSING_DEPENDENCY\"\r\n\r\n// ─── 파싱 결과 (discriminated union) ────────────────\r\n\r\nexport type FileType = \"hwpx\" | \"hwp\" | \"hwp3\" | \"hwpml\" | \"pdf\" | \"xlsx\" | \"xls\" | \"docx\" | \"image\" | \"unknown\"\r\n\r\ninterface ParseResultBase {\r\n fileType: FileType\r\n /** 페이지/섹션 수 — PDF: 실제 페이지 수, HWP/HWPX: 섹션 수, XLSX: 시트 수 */\r\n pageCount?: number\r\n /** 이미지 기반 PDF 여부 (텍스트 추출 불가) */\r\n isImageBased?: boolean\r\n}\r\n\r\nexport interface ParseSuccess extends ParseResultBase {\r\n success: true\r\n /** 추출된 마크다운 텍스트 */\r\n markdown: string\r\n /**\r\n * 중간 표현 블록 (구조화된 데이터 접근용).\r\n * 블록은 문서(원문) 읽기 순서를 따른다 — 한 문단 안에 글자취급(treatAsChar)\r\n * 표와 텍스트가 섞여 있어도 배치 순서대로 방출된다 (v4.2.3, #50).\r\n */\r\n blocks: IRBlock[]\r\n /** 문서 메타데이터 */\r\n metadata?: DocumentMetadata\r\n /** 문서 구조 (헤딩 트리) — v2.0 */\r\n outline?: OutlineItem[]\r\n /** 파싱 중 발생한 경고 — v2.0 */\r\n warnings?: ParseWarning[]\r\n /** 추출된 이미지 목록 — 마크다운에서 파일명으로 참조됨 */\r\n images?: ExtractedImage[]\r\n /** 페이지별 텍스트 품질 신호 — PDF에서만 제공 */\r\n pageQuality?: PageQuality[]\r\n /** 문서 단위 품질 요약 — PDF에서만 제공 */\r\n qualitySummary?: DocumentQualitySummary\r\n}\r\n\r\n/** 페이지별 텍스트 품질 신호 (PDF 전용). 자세한 설명은 src/pdf/quality.ts */\r\nexport interface PageQuality {\r\n page: number\r\n textChars: number\r\n hangulRatio: number\r\n controlCharRatio: number\r\n replacementCharRatio: number\r\n puaRatio: number\r\n needsOcr: boolean\r\n ocrReason?: \"low_text\" | \"high_pua\" | \"high_control\" | \"high_replacement\" | \"garbled_hangul\"\r\n}\r\n\r\n/** 문서 단위 품질 요약 (PDF 전용). */\r\nexport interface DocumentQualitySummary {\r\n totalPages: number\r\n totalTextChars: number\r\n avgHangulRatio: number\r\n avgControlCharRatio: number\r\n avgReplacementCharRatio: number\r\n avgPuaRatio: number\r\n lowTextPageCount: number\r\n highPuaPageCount: number\r\n needsOcr: boolean\r\n ocrCandidatePages: number[]\r\n}\r\n\r\n/** 추출된 이미지 — ParseSuccess.images에 포함 */\r\nexport interface ExtractedImage {\r\n /** 마크다운에서 참조되는 파일명 (예: image_001.png) */\r\n filename: string\r\n /** 이미지 바이너리 */\r\n data: Uint8Array\r\n /** MIME 타입 */\r\n mimeType: string\r\n}\r\n\r\nexport interface ParseFailure extends ParseResultBase {\r\n success: false\r\n /** 오류 메시지 */\r\n error: string\r\n /** 구조화된 에러 코드 */\r\n code?: ErrorCode\r\n}\r\n\r\nexport type ParseResult = ParseSuccess | ParseFailure\r\n\r\n// ─── 문서 비교 (Diff) ───────────────────────────────\r\n\r\nexport type DiffChangeType = \"added\" | \"removed\" | \"modified\" | \"unchanged\"\r\n\r\nexport interface BlockDiff {\r\n type: DiffChangeType\r\n /** 원본 블록 (added이면 undefined) */\r\n before?: IRBlock\r\n /** 변경 후 블록 (removed이면 undefined) */\r\n after?: IRBlock\r\n /** modified 테이블의 셀 단위 diff */\r\n cellDiffs?: CellDiff[][]\r\n /** 유사도 (0-1) */\r\n similarity?: number\r\n}\r\n\r\nexport interface CellDiff {\r\n type: DiffChangeType\r\n before?: string\r\n after?: string\r\n}\r\n\r\nexport interface DiffResult {\r\n stats: { added: number; removed: number; modified: number; unchanged: number }\r\n diffs: BlockDiff[]\r\n}\r\n\r\n// ─── 라운드트립 패치 (v3.0) ─────────────────────────\r\n\r\n/** 패치 중 매핑 실패/미지원으로 건너뛴 항목 — silent 실패 금지 */\r\nexport interface PatchSkip {\r\n /** 건너뛴 사유 */\r\n reason: string\r\n /** 원본 쪽 내용 요약 (최대 80자) */\r\n before?: string\r\n /** 편집 쪽 내용 요약 (최대 80자) */\r\n after?: string\r\n /**\r\n * 부분 적용 표시 — 변경이 적용은 됐지만(applied 계상) 편집 원형 그대로는\r\n * 아님 (예: 셀 내 줄 추가를 마지막 문단에 병합). 완전 미적용 skip과 구분.\r\n */\r\n partial?: boolean\r\n}\r\n\r\n/** patchHwpx 옵션 */\r\nexport interface PatchOptions {\r\n /** 패치 후 재파싱 자동 검증 (기본 true) */\r\n verify?: boolean\r\n}\r\n\r\n/** patchHwpx / patchBlocks 결과 */\r\nexport interface PatchResult {\r\n success: boolean\r\n /** 패치된 HWPX (success=true) */\r\n data?: Uint8Array\r\n /** 적용된 변경 수 */\r\n applied: number\r\n /** 매핑 실패 항목 (이유 포함) */\r\n skipped: PatchSkip[]\r\n /**\r\n * 무손실 검증 (patchHwpx/patchHwp 전용): 패치본 재파싱 vs 편집 마크다운의\r\n * 잔차 diff — modified/added/removed가 0이어야 의도가 전부 반영된 것.\r\n * session.patchBlocks는 이 필드를 채우지 않는다 (changes 참조).\r\n */\r\n verification?: DiffResult\r\n /**\r\n * 변경 가시화 (session.patchBlocks 전용): 패치 전 → 후 문서 diff —\r\n * 적용된 편집 수만큼 modified가 나오는 것이 정상. verification과 의미가\r\n * 정반대이므로 혼용 금지.\r\n */\r\n changes?: DiffResult\r\n /** 실패 사유 (success=false) */\r\n error?: string\r\n}\r\n\r\n// ─── 양식 인식 ──────────────────────────────────────\r\n\r\nexport interface FormField {\r\n label: string\r\n value: string\r\n /** 0-based 소스 행 */\r\n row: number\r\n /** 0-based 소스 열 */\r\n col: number\r\n /** 매칭된 입력 키(정규화) — 모호 라벨 거부 가드(fillWithUniqueGuard)의 집계 기준 */\r\n key?: string\r\n}\r\n\r\nexport interface FormResult {\r\n fields: FormField[]\r\n /** 양식 확신도 (0-1) */\r\n confidence: number\r\n}\r\n\r\n// ─── OCR 프로바이더 ─────────────────────────────────\r\n\r\n/** 사용자 제공 OCR 함수 — 페이지 이미지를 받아 텍스트 반환.\r\n * PDF 경로는 항상 \"image/png\", 이미지 직접 입력 경로는 원본 mime 그대로 전달. */\r\nexport type OcrProvider = (\r\n pageImage: Uint8Array,\r\n pageNumber: number,\r\n mimeType: \"image/png\" | \"image/jpeg\" | \"image/webp\"\r\n) => Promise<string>\r\n\r\n// ─── Watch 모드 ─────────────────────────────────────\r\n\r\nexport interface WatchOptions {\r\n dir: string\r\n outDir?: string\r\n webhook?: string\r\n format?: \"markdown\" | \"json\"\r\n pages?: string\r\n silent?: boolean\r\n}\r\n\r\n// ─── 헤딩 감지 공통 임계값 ──────────────────────────\r\n\r\n/** 폰트 크기 비율 → heading level (전 파서 공통) */\r\nexport const HEADING_RATIO_H1 = 1.5\r\nexport const HEADING_RATIO_H2 = 1.3\r\nexport const HEADING_RATIO_H3 = 1.15\r\n\r\n// ─── 내부 파서 반환 타입 ─────────────────────────────\r\n\r\n/** 내부 파서가 index.ts에 반환하는 공통 타입 (HWP5/HWPX/PDF/XLSX/DOCX) */\r\nexport interface InternalParseResult {\r\n markdown: string\r\n blocks: IRBlock[]\r\n metadata?: DocumentMetadata\r\n outline?: OutlineItem[]\r\n warnings?: ParseWarning[]\r\n images?: ExtractedImage[]\r\n /** PDF 전용: 이미지 기반 PDF 여부 */\r\n isImageBased?: boolean\r\n /** PDF 전용: 페이지별 품질 신호 */\r\n pageQuality?: PageQuality[]\r\n /** PDF 전용: 문서 단위 품질 요약 */\r\n qualitySummary?: DocumentQualitySummary\r\n}\r\n"],"mappings":";;;AA2dO,IAAM,mBAAmB;AACzB,IAAM,mBAAmB;AACzB,IAAM,mBAAmB;","names":[]}
@@ -1,7 +1,7 @@
1
1
  #!/usr/bin/env node
2
2
  import {
3
3
  blocksToMarkdown
4
- } from "./chunk-MHMVFI2N.js";
4
+ } from "./chunk-BPJSTJBN.js";
5
5
 
6
6
  // src/chunks.ts
7
7
  var LIST_MARKER_RE = /^(?:[□■◇◆○◎●◦ㅇ•▪▸▶※-]|\d{1,3}[.)]|[가나다라마바사아자차카타파하][.)]|\([가나다라마바사아자차카타파하0-9]{1,3}\)|[①-⑳㉮-㉻㈎-㈛])\s/;
@@ -79,4 +79,4 @@ function blocksToChunks(blocks, options) {
79
79
  export {
80
80
  blocksToChunks
81
81
  };
82
- //# sourceMappingURL=chunk-KTPZYXN2.js.map
82
+ //# sourceMappingURL=chunk-ZYYTU7NY.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../src/chunks.ts"],"sourcesContent":["/**\r\n * RAG용 구조 보존 청킹 — IRBlock[] → DocChunk[].\r\n *\r\n * 문서의 위계(마크다운 헤딩 스택 + 개조식 리스트 깊이)를 breadcrumb으로 보존한\r\n * 청크 목록을 만든다. 토큰 상한·오버랩 등 자르기 정책은 의도적으로 없다 —\r\n * \"구조 트리 제공\"까지가 이 모듈의 몫이고 실제 분할은 소비자(RAG 파이프라인)가 한다.\r\n *\r\n * breadcrumb 규칙:\r\n * - 헤딩(level 1~6)은 스택 구조 — 같거나 깊은 level을 팝하고 push, 리스트 위계 리셋\r\n * - 리스트 항목은 깊이 d 등장 시 depth>=d를 팝하고 push — 하위 항목이 뒤따르면\r\n * 그 breadcrumb으로 승격되고, 자식이 없으면 다음 형제/헤딩이 팝해 자연 소멸\r\n * - 깊이는 IR의 listDepth 그대로. listDepth가 없는 개조식 선두 부호 문단\r\n * (□·○·- / 1.·가.·1) 등)은 depth 0 항목으로 취급 — IR에 없는 깊이는 지어내지 않는다\r\n * - 청크의 breadcrumb은 자기 자신을 제외한 상위 경로다\r\n */\r\n\r\nimport type { IRBlock } from \"./types.js\"\r\nimport { blocksToMarkdown } from \"./table/builder.js\"\r\n\r\nexport interface DocChunk {\r\n /** \"c0001\" 순번 id — 같은 입력이면 같은 출력 (결정적) */\r\n id: string\r\n type: \"text\" | \"table\" | \"heading\"\r\n /** 상위 헤딩 + 리스트 위계 경로 (예: [\"1. 개요\", \"가. 목적\"]) */\r\n breadcrumb: string[]\r\n /** 청크 본문 마크다운 — 표는 GFM/HTML 그대로 (blocksToMarkdown 재사용) */\r\n text: string\r\n /** 원본 페이지/섹션 번호 (1-based) — IR에 없으면 생략 */\r\n page?: number\r\n /** 원본 IRBlock 인덱스 범위 [start, end] (양끝 포함) — 출처 앵커 */\r\n blockRange: [number, number]\r\n /** 표 청크 전용 구조 요약 — cells는 includeTableCells 옵션 시에만 */\r\n table?: { rows: number; cols: number; cells?: string[][] }\r\n}\r\n\r\nexport interface ChunkOptions {\r\n /** 표 청크에 셀 텍스트 행렬 포함 여부 (기본 false) */\r\n includeTableCells?: boolean\r\n /**\r\n * \"section\"(기본): 같은 breadcrumb 아래 연속 텍스트 블록을 하나로 병합,\r\n * \"block\": IRBlock 1개 = 청크 1개\r\n */\r\n granularity?: \"block\" | \"section\"\r\n}\r\n\r\n/**\r\n * 개조식/항목부호 선두 마커 — listDepth 미기재 리스트 항목의 depth 0 식별용.\r\n * gongmun 8단계 부호(1. 가. 1) 가) (1) (가) ① ㉮)와 개조식 글머리(□ ○ - ※ 등).\r\n * 파서는 md 문법과 충돌하는 부호('- '·'1) ')에만 listDepth(1~)를 채우므로,\r\n * depth 0 항목은 부호 텍스트로만 식별 가능하다 (승격이 동작하기 위한 최소 휴리스틱)\r\n */\r\nconst LIST_MARKER_RE =\r\n /^(?:[□■◇◆○◎●◦ㅇ•▪▸▶※-]|\\d{1,3}[.)]|[가나다라마바사아자차카타파하][.)]|\\([가나다라마바사아자차카타파하0-9]{1,3}\\)|[①-⑳㉮-㉻㈎-㈛])\\s/\r\n\r\n/** 블록의 리스트 깊이 — 리스트 항목이 아니면 undefined */\r\nfunction listDepthOf(block: IRBlock): number | undefined {\r\n if (block.listDepth !== undefined) return block.listDepth\r\n if (block.type === \"list\") return 0\r\n if (block.type === \"paragraph\" && block.text && LIST_MARKER_RE.test(block.text.trim())) return 0\r\n return undefined\r\n}\r\n\r\nfunction sameBreadcrumb(a: string[], b: string[]): boolean {\r\n return a.length === b.length && a.every((v, i) => v === b[i])\r\n}\r\n\r\nexport function blocksToChunks(blocks: IRBlock[], options?: ChunkOptions): DocChunk[] {\r\n const granularity = options?.granularity ?? \"section\"\r\n const includeCells = options?.includeTableCells ?? false\r\n\r\n const chunks: DocChunk[] = []\r\n const headingStack: { level: number; text: string }[] = []\r\n const listStack: { depth: number; text: string }[] = []\r\n const crumb = () => [...headingStack.map(h => h.text), ...listStack.map(l => l.text)]\r\n\r\n const push = (\r\n type: DocChunk[\"type\"],\r\n breadcrumb: string[],\r\n text: string,\r\n blockRange: [number, number],\r\n page?: number,\r\n table?: DocChunk[\"table\"]\r\n ) => {\r\n const chunk: DocChunk = { id: \"c\" + String(chunks.length + 1).padStart(4, \"0\"), type, breadcrumb, text, blockRange }\r\n if (page !== undefined) chunk.page = page\r\n if (table) chunk.table = table\r\n chunks.push(chunk)\r\n }\r\n\r\n // section 병합 누적 상태 — 같은 breadcrumb의 연속 텍스트 블록\r\n let run: { breadcrumb: string[]; blocks: IRBlock[]; start: number; end: number; page?: number } | null = null\r\n const flushRun = () => {\r\n if (!run) return\r\n push(\"text\", run.breadcrumb, blocksToMarkdown(run.blocks), [run.start, run.end], run.page)\r\n run = null\r\n }\r\n\r\n for (let i = 0; i < blocks.length; i++) {\r\n const block = blocks[i]\r\n const md = blocksToMarkdown([block])\r\n if (!md) continue // 빈 블록 스킵 — 스택에도 불참\r\n\r\n if (block.type === \"heading\") {\r\n flushRun()\r\n const level = Math.min(block.level || 2, 6) // blocksToMarkdown과 동일한 기본값\r\n while (headingStack.length && headingStack[headingStack.length - 1].level >= level) headingStack.pop()\r\n listStack.length = 0 // 새 헤딩 = 리스트 위계 리셋\r\n push(\"heading\", crumb(), md, [i, i], block.pageNumber)\r\n headingStack.push({ level, text: (block.text ?? \"\").trim() })\r\n continue\r\n }\r\n\r\n if (block.type === \"table\" && block.table) {\r\n // 표는 항상 독립 청크 — 병합 금지\r\n flushRun()\r\n const summary: NonNullable<DocChunk[\"table\"]> = { rows: block.table.rows, cols: block.table.cols }\r\n if (includeCells) summary.cells = block.table.cells.map(row => row.map(c => c.text))\r\n push(\"table\", crumb(), md, [i, i], block.pageNumber, summary)\r\n continue\r\n }\r\n\r\n // 텍스트 계열 (paragraph·list·image·separator)\r\n const depth = listDepthOf(block)\r\n if (depth !== undefined) {\r\n while (listStack.length && listStack[listStack.length - 1].depth >= depth) listStack.pop()\r\n }\r\n const breadcrumb = crumb()\r\n if (depth !== undefined) listStack.push({ depth, text: (block.text ?? \"\").trim() })\r\n\r\n if (granularity === \"block\") {\r\n push(\"text\", breadcrumb, md, [i, i], block.pageNumber)\r\n continue\r\n }\r\n if (run && sameBreadcrumb(run.breadcrumb, breadcrumb)) {\r\n run.blocks.push(block)\r\n run.end = i\r\n if (run.page === undefined) run.page = block.pageNumber\r\n } else {\r\n flushRun()\r\n run = { breadcrumb, blocks: [block], start: i, end: i, page: block.pageNumber }\r\n }\r\n }\r\n flushRun()\r\n\r\n return chunks\r\n}\r\n"],"mappings":";;;;;;AAmDA,IAAM,iBACJ;AAGF,SAAS,YAAY,OAAoC;AACvD,MAAI,MAAM,cAAc,OAAW,QAAO,MAAM;AAChD,MAAI,MAAM,SAAS,OAAQ,QAAO;AAClC,MAAI,MAAM,SAAS,eAAe,MAAM,QAAQ,eAAe,KAAK,MAAM,KAAK,KAAK,CAAC,EAAG,QAAO;AAC/F,SAAO;AACT;AAEA,SAAS,eAAe,GAAa,GAAsB;AACzD,SAAO,EAAE,WAAW,EAAE,UAAU,EAAE,MAAM,CAAC,GAAG,MAAM,MAAM,EAAE,CAAC,CAAC;AAC9D;AAEO,SAAS,eAAe,QAAmB,SAAoC;AACpF,QAAM,cAAc,SAAS,eAAe;AAC5C,QAAM,eAAe,SAAS,qBAAqB;AAEnD,QAAM,SAAqB,CAAC;AAC5B,QAAM,eAAkD,CAAC;AACzD,QAAM,YAA+C,CAAC;AACtD,QAAM,QAAQ,MAAM,CAAC,GAAG,aAAa,IAAI,OAAK,EAAE,IAAI,GAAG,GAAG,UAAU,IAAI,OAAK,EAAE,IAAI,CAAC;AAEpF,QAAM,OAAO,CACX,MACA,YACA,MACA,YACA,MACA,UACG;AACH,UAAM,QAAkB,EAAE,IAAI,MAAM,OAAO,OAAO,SAAS,CAAC,EAAE,SAAS,GAAG,GAAG,GAAG,MAAM,YAAY,MAAM,WAAW;AACnH,QAAI,SAAS,OAAW,OAAM,OAAO;AACrC,QAAI,MAAO,OAAM,QAAQ;AACzB,WAAO,KAAK,KAAK;AAAA,EACnB;AAGA,MAAI,MAAqG;AACzG,QAAM,WAAW,MAAM;AACrB,QAAI,CAAC,IAAK;AACV,SAAK,QAAQ,IAAI,YAAY,iBAAiB,IAAI,MAAM,GAAG,CAAC,IAAI,OAAO,IAAI,GAAG,GAAG,IAAI,IAAI;AACzF,UAAM;AAAA,EACR;AAEA,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,UAAM,QAAQ,OAAO,CAAC;AACtB,UAAM,KAAK,iBAAiB,CAAC,KAAK,CAAC;AACnC,QAAI,CAAC,GAAI;AAET,QAAI,MAAM,SAAS,WAAW;AAC5B,eAAS;AACT,YAAM,QAAQ,KAAK,IAAI,MAAM,SAAS,GAAG,CAAC;AAC1C,aAAO,aAAa,UAAU,aAAa,aAAa,SAAS,CAAC,EAAE,SAAS,MAAO,cAAa,IAAI;AACrG,gBAAU,SAAS;AACnB,WAAK,WAAW,MAAM,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,MAAM,UAAU;AACrD,mBAAa,KAAK,EAAE,OAAO,OAAO,MAAM,QAAQ,IAAI,KAAK,EAAE,CAAC;AAC5D;AAAA,IACF;AAEA,QAAI,MAAM,SAAS,WAAW,MAAM,OAAO;AAEzC,eAAS;AACT,YAAM,UAA0C,EAAE,MAAM,MAAM,MAAM,MAAM,MAAM,MAAM,MAAM,KAAK;AACjG,UAAI,aAAc,SAAQ,QAAQ,MAAM,MAAM,MAAM,IAAI,SAAO,IAAI,IAAI,OAAK,EAAE,IAAI,CAAC;AACnF,WAAK,SAAS,MAAM,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,MAAM,YAAY,OAAO;AAC5D;AAAA,IACF;AAGA,UAAM,QAAQ,YAAY,KAAK;AAC/B,QAAI,UAAU,QAAW;AACvB,aAAO,UAAU,UAAU,UAAU,UAAU,SAAS,CAAC,EAAE,SAAS,MAAO,WAAU,IAAI;AAAA,IAC3F;AACA,UAAM,aAAa,MAAM;AACzB,QAAI,UAAU,OAAW,WAAU,KAAK,EAAE,OAAO,OAAO,MAAM,QAAQ,IAAI,KAAK,EAAE,CAAC;AAElF,QAAI,gBAAgB,SAAS;AAC3B,WAAK,QAAQ,YAAY,IAAI,CAAC,GAAG,CAAC,GAAG,MAAM,UAAU;AACrD;AAAA,IACF;AACA,QAAI,OAAO,eAAe,IAAI,YAAY,UAAU,GAAG;AACrD,UAAI,OAAO,KAAK,KAAK;AACrB,UAAI,MAAM;AACV,UAAI,IAAI,SAAS,OAAW,KAAI,OAAO,MAAM;AAAA,IAC/C,OAAO;AACL,eAAS;AACT,YAAM,EAAE,YAAY,QAAQ,CAAC,KAAK,GAAG,OAAO,GAAG,KAAK,GAAG,MAAM,MAAM,WAAW;AAAA,IAChF;AAAA,EACF;AACA,WAAS;AAET,SAAO;AACT;","names":[]}
@@ -0,0 +1,10 @@
1
+ #!/usr/bin/env node
2
+ import {
3
+ blocksToChunks
4
+ } from "./chunk-ZYYTU7NY.js";
5
+ import "./chunk-BPJSTJBN.js";
6
+ import "./chunk-ZJ75YDW7.js";
7
+ export {
8
+ blocksToChunks
9
+ };
10
+ //# sourceMappingURL=chunks-JEYQWXIZ.js.map