kordoc 4.0.8 → 4.2.0
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.
- package/README.md +50 -16
- package/dist/{-LE4RLXVA.js → -SXQIZHYA.js} +24 -10
- package/dist/{parser-TCTCSBYZ.js → chunk-26AJJCFF.js} +1427 -2027
- package/dist/chunk-26AJJCFF.js.map +1 -0
- package/dist/{parser-KMMUS73Q.js → chunk-4JPPLIIV.js} +1425 -2025
- package/dist/chunk-4JPPLIIV.js.map +1 -0
- package/dist/chunk-4VNLZ2GZ.js +13 -0
- package/dist/chunk-4VNLZ2GZ.js.map +1 -0
- package/dist/{chunk-2DZQF6YJ.js → chunk-6SHO3BNV.js} +24 -15
- package/dist/chunk-6SHO3BNV.js.map +1 -0
- package/dist/chunk-7FCNOKOR.js +144 -0
- package/dist/chunk-7FCNOKOR.js.map +1 -0
- package/dist/{chunk-6XAAYAWL.cjs → chunk-CMN3SYZ3.js} +62 -304
- package/dist/chunk-CMN3SYZ3.js.map +1 -0
- package/dist/chunk-DZIXKL7E.js +145 -0
- package/dist/chunk-DZIXKL7E.js.map +1 -0
- package/dist/chunk-ELOV2GIJ.js +82 -0
- package/dist/chunk-ELOV2GIJ.js.map +1 -0
- package/dist/chunk-F2HYH62H.cjs +130 -0
- package/dist/chunk-F2HYH62H.cjs.map +1 -0
- package/dist/{chunk-HMUEU7D7.js → chunk-FSXLSHCW.js} +3 -3
- package/dist/{chunk-HMUEU7D7.js.map → chunk-FSXLSHCW.js.map} +1 -1
- package/dist/chunk-GEPINOL6.js +83 -0
- package/dist/chunk-GEPINOL6.js.map +1 -0
- package/dist/chunk-H7OWEYH2.cjs +243 -0
- package/dist/chunk-H7OWEYH2.cjs.map +1 -0
- package/dist/{chunk-7S3M4N4E.js → chunk-IGGRLWT5.js} +683 -226
- package/dist/chunk-IGGRLWT5.js.map +1 -0
- package/dist/chunk-J2RZ444H.js +243 -0
- package/dist/chunk-J2RZ444H.js.map +1 -0
- package/dist/chunk-JP5YH3Z4.js +130 -0
- package/dist/chunk-JP5YH3Z4.js.map +1 -0
- package/dist/{parser-ITMU2WN5.cjs → chunk-LHROM2BO.cjs} +1409 -2012
- package/dist/chunk-LHROM2BO.cjs.map +1 -0
- package/dist/chunk-NNDVFNMJ.js +246 -0
- package/dist/chunk-NNDVFNMJ.js.map +1 -0
- package/dist/{chunk-RROH5WHO.js → chunk-PLIY23LY.js} +2 -2
- package/dist/{chunk-BNU5QGIZ.js → chunk-V6U6YS5B.js} +12 -4
- package/dist/chunk-V6U6YS5B.js.map +1 -0
- package/dist/{chunk-UEIYETCQ.js → chunk-WF2E5P2G.js} +40 -35
- package/dist/chunk-WF2E5P2G.js.map +1 -0
- package/dist/{chunk-D35VBACN.js → chunk-WQLHPOMJ.js} +7 -4
- package/dist/chunk-WQLHPOMJ.js.map +1 -0
- package/dist/{chunk-OQPILS7B.js → chunk-XTCT6KQ5.cjs} +79 -183
- package/dist/chunk-XTCT6KQ5.cjs.map +1 -0
- package/dist/{chunk-MEPHGCPQ.js → chunk-YOO6ET6M.js} +7 -3
- package/dist/chunk-YOO6ET6M.js.map +1 -0
- package/dist/chunks-ZQEFAZNC.js +10 -0
- package/dist/cli.js +228 -36
- package/dist/cli.js.map +1 -1
- package/dist/{detect-RI2MQ33K.js → detect-YEZX2XCF.js} +2 -2
- package/dist/{formula-RXVSQPXI.js → formula-KSNAEBZ2.js} +23 -232
- package/dist/formula-KSNAEBZ2.js.map +1 -0
- package/dist/{formula-5NKVS2LR.cjs → formula-SB3NHDUM.cjs} +30 -240
- package/dist/formula-SB3NHDUM.cjs.map +1 -0
- package/dist/{formula-JCNF43NE.js → formula-SXC6LRRC.js} +23 -233
- package/dist/formula-SXC6LRRC.js.map +1 -0
- package/dist/index.cjs +1280 -589
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.cts +112 -4
- package/dist/index.d.ts +112 -4
- package/dist/index.js +929 -238
- package/dist/index.js.map +1 -1
- package/dist/mcp.js +278 -68
- package/dist/mcp.js.map +1 -1
- package/dist/models-KA4GLQK5.js +25 -0
- package/dist/parser-ND2YGQZR.js +800 -0
- package/dist/parser-ND2YGQZR.js.map +1 -0
- package/dist/parser-RBZHA6SY.js +804 -0
- package/dist/parser-RBZHA6SY.js.map +1 -0
- package/dist/parser-WDZOMDG4.cjs +803 -0
- package/dist/parser-WDZOMDG4.cjs.map +1 -0
- package/dist/pdf-ocr-2Z7VZIFI.cjs +439 -0
- package/dist/pdf-ocr-2Z7VZIFI.cjs.map +1 -0
- package/dist/pdf-ocr-FJFXFADQ.js +438 -0
- package/dist/pdf-ocr-FJFXFADQ.js.map +1 -0
- package/dist/pdf-ocr-VYE4DHD4.js +386 -0
- package/dist/pdf-ocr-VYE4DHD4.js.map +1 -0
- package/dist/{profile-io-SLN4P76T.js → profile-io-TWHDIVPL.js} +3 -3
- package/dist/rasterize-W2NEZUSW.js +40 -0
- package/dist/rasterize-W2NEZUSW.js.map +1 -0
- package/dist/redact-7ILEAUTS.js +12 -0
- package/dist/redact-7ILEAUTS.js.map +1 -0
- package/dist/render-4MG5QUDG.js +10 -0
- package/dist/render-4MG5QUDG.js.map +1 -0
- package/dist/seal-D2QNXJVO.js +10 -0
- package/dist/seal-D2QNXJVO.js.map +1 -0
- package/dist/{setup-57FB3LSP.js → setup-Q2PRE7UA.js} +5 -3
- package/dist/setup-Q2PRE7UA.js.map +1 -0
- package/dist/{watch-OHQWSVPE.js → watch-VWXJZT2R.js} +78 -23
- package/dist/watch-VWXJZT2R.js.map +1 -0
- package/package.json +2 -1
- package/dist/chunk-2DZQF6YJ.js.map +0 -1
- package/dist/chunk-6XAAYAWL.cjs.map +0 -1
- package/dist/chunk-7S3M4N4E.js.map +0 -1
- package/dist/chunk-BNU5QGIZ.js.map +0 -1
- package/dist/chunk-D35VBACN.js.map +0 -1
- package/dist/chunk-MEPHGCPQ.js.map +0 -1
- package/dist/chunk-OQPILS7B.js.map +0 -1
- package/dist/chunk-UEIYETCQ.js.map +0 -1
- package/dist/formula-5NKVS2LR.cjs.map +0 -1
- package/dist/formula-JCNF43NE.js.map +0 -1
- package/dist/formula-RXVSQPXI.js.map +0 -1
- package/dist/parser-ITMU2WN5.cjs.map +0 -1
- package/dist/parser-KMMUS73Q.js.map +0 -1
- package/dist/parser-TCTCSBYZ.js.map +0 -1
- package/dist/provider-4ZJKV3DC.js +0 -37
- package/dist/provider-4ZJKV3DC.js.map +0 -1
- package/dist/provider-AKROB7WQ.js +0 -39
- package/dist/provider-AKROB7WQ.js.map +0 -1
- package/dist/provider-G4C2V2PD.cjs +0 -39
- package/dist/provider-G4C2V2PD.cjs.map +0 -1
- package/dist/render-25BT623I.js +0 -10
- package/dist/seal-R4TETA4C.js +0 -10
- package/dist/setup-57FB3LSP.js.map +0 -1
- package/dist/watch-OHQWSVPE.js.map +0 -1
- /package/dist/{-LE4RLXVA.js.map → -SXQIZHYA.js.map} +0 -0
- /package/dist/{chunk-RROH5WHO.js.map → chunk-PLIY23LY.js.map} +0 -0
- /package/dist/{detect-RI2MQ33K.js.map → chunks-ZQEFAZNC.js.map} +0 -0
- /package/dist/{profile-io-SLN4P76T.js.map → detect-YEZX2XCF.js.map} +0 -0
- /package/dist/{render-25BT623I.js.map → models-KA4GLQK5.js.map} +0 -0
- /package/dist/{seal-R4TETA4C.js.map → profile-io-TWHDIVPL.js.map} +0 -0
|
@@ -0,0 +1,246 @@
|
|
|
1
|
+
#!/usr/bin/env node
|
|
2
|
+
|
|
3
|
+
// src/pdf/formula/models.ts
|
|
4
|
+
import { createHash } from "crypto";
|
|
5
|
+
import { createReadStream } from "fs";
|
|
6
|
+
import { mkdir, stat, unlink, rename } from "fs/promises";
|
|
7
|
+
import { createWriteStream } from "fs";
|
|
8
|
+
import { homedir } from "os";
|
|
9
|
+
import { join, dirname } from "path";
|
|
10
|
+
import { pipeline } from "stream/promises";
|
|
11
|
+
import { Readable } from "stream";
|
|
12
|
+
var MFD_MODEL = {
|
|
13
|
+
name: "Pix2Text MFD",
|
|
14
|
+
filename: "mfd.onnx",
|
|
15
|
+
url: "https://huggingface.co/breezedeus/pix2text-mfd/resolve/main/mfd-v20240618.onnx",
|
|
16
|
+
sha256: "51a8854743b17ae654729af8db82a630c1ccfa06debf4856c8b28055f87d02c1",
|
|
17
|
+
sizeMb: 42
|
|
18
|
+
};
|
|
19
|
+
var MFR_ENCODER_MODEL = {
|
|
20
|
+
name: "Pix2Text MFR encoder",
|
|
21
|
+
filename: "encoder_model.onnx",
|
|
22
|
+
url: "https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/encoder_model.onnx",
|
|
23
|
+
sha256: "bd8d5c322792e9ec45793af5569e9748f82a3d728a9e00213dbfc56c1486f37d",
|
|
24
|
+
sizeMb: 87
|
|
25
|
+
};
|
|
26
|
+
var MFR_DECODER_MODEL = {
|
|
27
|
+
name: "Pix2Text MFR decoder",
|
|
28
|
+
filename: "decoder_model.onnx",
|
|
29
|
+
url: "https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/decoder_model.onnx",
|
|
30
|
+
sha256: "fd0f92d7a012f3dae41e1ac79421aea0ea888b5a66cb3f9a004e424f82f3daed",
|
|
31
|
+
sizeMb: 30
|
|
32
|
+
};
|
|
33
|
+
var MFR_TOKENIZER = {
|
|
34
|
+
name: "Pix2Text MFR tokenizer",
|
|
35
|
+
filename: "tokenizer.json",
|
|
36
|
+
url: "https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/tokenizer.json",
|
|
37
|
+
sha256: "3e2ab757277d22639bec28c9d7972e352d3d1dba223051fa674002dc5ab64df3",
|
|
38
|
+
sizeMb: 1
|
|
39
|
+
};
|
|
40
|
+
var ALL_FORMULA_MODELS = [
|
|
41
|
+
MFD_MODEL,
|
|
42
|
+
MFR_ENCODER_MODEL,
|
|
43
|
+
MFR_DECODER_MODEL,
|
|
44
|
+
MFR_TOKENIZER
|
|
45
|
+
];
|
|
46
|
+
function getModelsDir(subdir) {
|
|
47
|
+
const override = process.env.KORDOC_MODEL_CACHE;
|
|
48
|
+
if (override && override.trim()) {
|
|
49
|
+
return join(override, subdir);
|
|
50
|
+
}
|
|
51
|
+
return join(homedir(), ".cache", "kordoc", "models", subdir);
|
|
52
|
+
}
|
|
53
|
+
function getFormulaModelsDir() {
|
|
54
|
+
return getModelsDir("pix2text");
|
|
55
|
+
}
|
|
56
|
+
async function getModelStatusIn(dir, specs) {
|
|
57
|
+
const result = [];
|
|
58
|
+
for (const spec of specs) {
|
|
59
|
+
const localPath = join(dir, spec.filename);
|
|
60
|
+
let exists = false;
|
|
61
|
+
try {
|
|
62
|
+
const s = await stat(localPath);
|
|
63
|
+
exists = s.isFile() && s.size > 0;
|
|
64
|
+
} catch {
|
|
65
|
+
exists = false;
|
|
66
|
+
}
|
|
67
|
+
if (!exists) {
|
|
68
|
+
result.push({ spec, localPath, exists: false, verified: false });
|
|
69
|
+
continue;
|
|
70
|
+
}
|
|
71
|
+
try {
|
|
72
|
+
const actual = await sha256OfFile(localPath);
|
|
73
|
+
if (actual === spec.sha256) {
|
|
74
|
+
result.push({ spec, localPath, exists: true, verified: true });
|
|
75
|
+
} else {
|
|
76
|
+
result.push({
|
|
77
|
+
spec,
|
|
78
|
+
localPath,
|
|
79
|
+
exists: true,
|
|
80
|
+
verified: false,
|
|
81
|
+
invalidReason: `SHA256 mismatch: expected ${spec.sha256}, got ${actual}`
|
|
82
|
+
});
|
|
83
|
+
}
|
|
84
|
+
} catch (e) {
|
|
85
|
+
result.push({
|
|
86
|
+
spec,
|
|
87
|
+
localPath,
|
|
88
|
+
exists: true,
|
|
89
|
+
verified: false,
|
|
90
|
+
invalidReason: `SHA compute failed: ${e.message}`
|
|
91
|
+
});
|
|
92
|
+
}
|
|
93
|
+
}
|
|
94
|
+
return result;
|
|
95
|
+
}
|
|
96
|
+
async function getFormulaModelStatus() {
|
|
97
|
+
return getModelStatusIn(getFormulaModelsDir(), ALL_FORMULA_MODELS);
|
|
98
|
+
}
|
|
99
|
+
async function ensureModelsIn(dir, specs, onProgress) {
|
|
100
|
+
await mkdir(dir, { recursive: true });
|
|
101
|
+
for (const spec of specs) {
|
|
102
|
+
const localPath = join(dir, spec.filename);
|
|
103
|
+
if (await isExistingValid(localPath, spec.sha256)) {
|
|
104
|
+
onProgress?.({
|
|
105
|
+
spec,
|
|
106
|
+
downloaded: 0,
|
|
107
|
+
total: null,
|
|
108
|
+
phase: "skip",
|
|
109
|
+
message: "\uC774\uBBF8 \uC874\uC7AC + SHA \uC77C\uCE58"
|
|
110
|
+
});
|
|
111
|
+
continue;
|
|
112
|
+
}
|
|
113
|
+
try {
|
|
114
|
+
await unlink(localPath);
|
|
115
|
+
} catch {
|
|
116
|
+
}
|
|
117
|
+
await downloadToFile(spec, localPath, onProgress);
|
|
118
|
+
}
|
|
119
|
+
}
|
|
120
|
+
async function ensureFormulaModels(onProgress) {
|
|
121
|
+
return ensureModelsIn(getFormulaModelsDir(), ALL_FORMULA_MODELS, onProgress);
|
|
122
|
+
}
|
|
123
|
+
async function ensureSingleModel(spec, onProgress) {
|
|
124
|
+
const dir = getFormulaModelsDir();
|
|
125
|
+
await mkdir(dir, { recursive: true });
|
|
126
|
+
const localPath = join(dir, spec.filename);
|
|
127
|
+
if (await isExistingValid(localPath, spec.sha256)) {
|
|
128
|
+
onProgress?.({ spec, downloaded: 0, total: null, phase: "skip" });
|
|
129
|
+
return;
|
|
130
|
+
}
|
|
131
|
+
try {
|
|
132
|
+
await unlink(localPath);
|
|
133
|
+
} catch {
|
|
134
|
+
}
|
|
135
|
+
await downloadToFile(spec, localPath, onProgress);
|
|
136
|
+
}
|
|
137
|
+
async function isExistingValid(localPath, sha256Expected) {
|
|
138
|
+
try {
|
|
139
|
+
const s = await stat(localPath);
|
|
140
|
+
if (!s.isFile() || s.size === 0) return false;
|
|
141
|
+
} catch {
|
|
142
|
+
return false;
|
|
143
|
+
}
|
|
144
|
+
try {
|
|
145
|
+
const actual = await sha256OfFile(localPath);
|
|
146
|
+
return actual === sha256Expected;
|
|
147
|
+
} catch {
|
|
148
|
+
return false;
|
|
149
|
+
}
|
|
150
|
+
}
|
|
151
|
+
async function downloadToFile(spec, localPath, onProgress) {
|
|
152
|
+
const partPath = `${localPath}.part`;
|
|
153
|
+
await mkdir(dirname(localPath), { recursive: true });
|
|
154
|
+
const resp = await fetch(spec.url, {
|
|
155
|
+
headers: {
|
|
156
|
+
// HF CDN 은 UA 없으면 가끔 403 을 뱉는다
|
|
157
|
+
"User-Agent": "kordoc-formula-ocr/1.0 (+https://github.com/chrisryugj/kordoc)"
|
|
158
|
+
}
|
|
159
|
+
});
|
|
160
|
+
if (!resp.ok || !resp.body) {
|
|
161
|
+
throw new Error(
|
|
162
|
+
`${spec.name} \uB2E4\uC6B4\uB85C\uB4DC \uC2E4\uD328: HTTP ${resp.status} ${resp.statusText} (${spec.url})`
|
|
163
|
+
);
|
|
164
|
+
}
|
|
165
|
+
const lenHeader = resp.headers.get("content-length");
|
|
166
|
+
const total = lenHeader ? Number.parseInt(lenHeader, 10) : null;
|
|
167
|
+
let downloaded = 0;
|
|
168
|
+
const ws = createWriteStream(partPath);
|
|
169
|
+
try {
|
|
170
|
+
const reader = Readable.fromWeb(resp.body);
|
|
171
|
+
reader.on("data", (chunk) => {
|
|
172
|
+
downloaded += chunk.length;
|
|
173
|
+
onProgress?.({
|
|
174
|
+
spec,
|
|
175
|
+
downloaded,
|
|
176
|
+
total,
|
|
177
|
+
phase: "download"
|
|
178
|
+
});
|
|
179
|
+
});
|
|
180
|
+
await pipeline(reader, ws);
|
|
181
|
+
} catch (e) {
|
|
182
|
+
try {
|
|
183
|
+
await unlink(partPath);
|
|
184
|
+
} catch {
|
|
185
|
+
}
|
|
186
|
+
throw new Error(`${spec.name} \uC2A4\uD2B8\uB9AC\uBC0D \uC2E4\uD328: ${e.message}`);
|
|
187
|
+
}
|
|
188
|
+
onProgress?.({
|
|
189
|
+
spec,
|
|
190
|
+
downloaded,
|
|
191
|
+
total,
|
|
192
|
+
phase: "verify"
|
|
193
|
+
});
|
|
194
|
+
let actual;
|
|
195
|
+
try {
|
|
196
|
+
actual = await sha256OfFile(partPath);
|
|
197
|
+
} catch (e) {
|
|
198
|
+
try {
|
|
199
|
+
await unlink(partPath);
|
|
200
|
+
} catch {
|
|
201
|
+
}
|
|
202
|
+
throw new Error(`${spec.name} SHA \uACC4\uC0B0 \uC2E4\uD328: ${e.message}`);
|
|
203
|
+
}
|
|
204
|
+
if (actual !== spec.sha256) {
|
|
205
|
+
try {
|
|
206
|
+
await unlink(partPath);
|
|
207
|
+
} catch {
|
|
208
|
+
}
|
|
209
|
+
throw new Error(
|
|
210
|
+
`${spec.name} SHA256 mismatch: expected ${spec.sha256}, got ${actual} \u2014 \uBAA8\uB378 URL \uC774 \uC624\uC5FC\uB418\uC5C8\uAC70\uB098 \uC804\uC1A1 \uC911 \uC190\uC0C1\uB418\uC5C8\uC2B5\uB2C8\uB2E4.`
|
|
211
|
+
);
|
|
212
|
+
}
|
|
213
|
+
await rename(partPath, localPath);
|
|
214
|
+
onProgress?.({
|
|
215
|
+
spec,
|
|
216
|
+
downloaded,
|
|
217
|
+
total,
|
|
218
|
+
phase: "done"
|
|
219
|
+
});
|
|
220
|
+
}
|
|
221
|
+
async function sha256OfFile(p) {
|
|
222
|
+
const h = createHash("sha256");
|
|
223
|
+
const stream = createReadStream(p);
|
|
224
|
+
await pipeline(stream, async function* (src) {
|
|
225
|
+
for await (const chunk of src) {
|
|
226
|
+
h.update(chunk);
|
|
227
|
+
}
|
|
228
|
+
});
|
|
229
|
+
return h.digest("hex");
|
|
230
|
+
}
|
|
231
|
+
|
|
232
|
+
export {
|
|
233
|
+
MFD_MODEL,
|
|
234
|
+
MFR_ENCODER_MODEL,
|
|
235
|
+
MFR_DECODER_MODEL,
|
|
236
|
+
MFR_TOKENIZER,
|
|
237
|
+
ALL_FORMULA_MODELS,
|
|
238
|
+
getModelsDir,
|
|
239
|
+
getFormulaModelsDir,
|
|
240
|
+
getModelStatusIn,
|
|
241
|
+
getFormulaModelStatus,
|
|
242
|
+
ensureModelsIn,
|
|
243
|
+
ensureFormulaModels,
|
|
244
|
+
ensureSingleModel
|
|
245
|
+
};
|
|
246
|
+
//# sourceMappingURL=chunk-NNDVFNMJ.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"sources":["../src/pdf/formula/models.ts"],"sourcesContent":["/**\n * 수식 OCR 모델 자동 다운로드 + SHA-256 검증\n *\n * 캐시 위치: `~/.cache/kordoc/models/pix2text/`\n * 총 ~155MB (MFD 44MB + MFR encoder 87MB + MFR decoder 30MB + tokenizer 40KB)\n *\n * HF URL + SHA 는 Docufinder archive/rust-formula-ocr-wip 에서 검증 완료됨.\n * 모델 버전 갱신 시 URL + SHA 를 함께 갱신해야 한다.\n */\n\nimport { createHash } from \"crypto\"\nimport { createReadStream } from \"fs\"\nimport { mkdir, stat, unlink, rename } from \"fs/promises\"\nimport { createWriteStream } from \"fs\"\nimport { homedir } from \"os\"\nimport { join, dirname } from \"path\"\nimport { pipeline } from \"stream/promises\"\nimport { Readable } from \"stream\"\n\nexport interface ModelSpec {\n name: string\n filename: string\n url: string\n sha256: string\n sizeMb: number\n}\n\n/** 모델 스펙 (SHA-256 은 실제 다운로드로 검증됨 — 변경 금지) */\nexport const MFD_MODEL: ModelSpec = {\n name: \"Pix2Text MFD\",\n filename: \"mfd.onnx\",\n url: \"https://huggingface.co/breezedeus/pix2text-mfd/resolve/main/mfd-v20240618.onnx\",\n sha256: \"51a8854743b17ae654729af8db82a630c1ccfa06debf4856c8b28055f87d02c1\",\n sizeMb: 42,\n}\n\nexport const MFR_ENCODER_MODEL: ModelSpec = {\n name: \"Pix2Text MFR encoder\",\n filename: \"encoder_model.onnx\",\n url: \"https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/encoder_model.onnx\",\n sha256: \"bd8d5c322792e9ec45793af5569e9748f82a3d728a9e00213dbfc56c1486f37d\",\n sizeMb: 87,\n}\n\nexport const MFR_DECODER_MODEL: ModelSpec = {\n name: \"Pix2Text MFR decoder\",\n filename: \"decoder_model.onnx\",\n url: \"https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/decoder_model.onnx\",\n sha256: \"fd0f92d7a012f3dae41e1ac79421aea0ea888b5a66cb3f9a004e424f82f3daed\",\n sizeMb: 30,\n}\n\nexport const MFR_TOKENIZER: ModelSpec = {\n name: \"Pix2Text MFR tokenizer\",\n filename: \"tokenizer.json\",\n url: \"https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/tokenizer.json\",\n sha256: \"3e2ab757277d22639bec28c9d7972e352d3d1dba223051fa674002dc5ab64df3\",\n sizeMb: 1,\n}\n\nexport const ALL_FORMULA_MODELS: ReadonlyArray<ModelSpec> = [\n MFD_MODEL,\n MFR_ENCODER_MODEL,\n MFR_DECODER_MODEL,\n MFR_TOKENIZER,\n]\n\n/**\n * 모델 캐시 하위 디렉토리 경로. 환경 변수 `KORDOC_MODEL_CACHE` 가 설정되면\n * 해당 경로를 우선 사용, 없으면 `~/.cache/kordoc/models/<subdir>/`.\n * (수식 OCR 외 텍스트 OCR 등 다른 모델군도 공용 — subdir 로 격리)\n */\nexport function getModelsDir(subdir: string): string {\n const override = process.env.KORDOC_MODEL_CACHE\n if (override && override.trim()) {\n return join(override, subdir)\n }\n return join(homedir(), \".cache\", \"kordoc\", \"models\", subdir)\n}\n\nexport function getFormulaModelsDir(): string {\n return getModelsDir(\"pix2text\")\n}\n\nexport interface ModelStatus {\n spec: ModelSpec\n localPath: string\n exists: boolean\n /** SHA 검증까지 끝났는가 (exists && valid) */\n verified: boolean\n /** 검증 실패 시 이유 */\n invalidReason?: string\n}\n\n/** 지정 모델군의 존재/유효성 상태 반환 (다운로드 없이 확인만) */\nexport async function getModelStatusIn(dir: string, specs: ReadonlyArray<ModelSpec>): Promise<ModelStatus[]> {\n const result: ModelStatus[] = []\n for (const spec of specs) {\n const localPath = join(dir, spec.filename)\n let exists = false\n try {\n const s = await stat(localPath)\n exists = s.isFile() && s.size > 0\n } catch {\n exists = false\n }\n if (!exists) {\n result.push({ spec, localPath, exists: false, verified: false })\n continue\n }\n try {\n const actual = await sha256OfFile(localPath)\n if (actual === spec.sha256) {\n result.push({ spec, localPath, exists: true, verified: true })\n } else {\n result.push({\n spec,\n localPath,\n exists: true,\n verified: false,\n invalidReason: `SHA256 mismatch: expected ${spec.sha256}, got ${actual}`,\n })\n }\n } catch (e) {\n result.push({\n spec,\n localPath,\n exists: true,\n verified: false,\n invalidReason: `SHA compute failed: ${(e as Error).message}`,\n })\n }\n }\n return result\n}\n\n/** 모든 수식 모델의 존재/유효성 상태 반환 (다운로드 없이 확인만) */\nexport async function getFormulaModelStatus(): Promise<ModelStatus[]> {\n return getModelStatusIn(getFormulaModelsDir(), ALL_FORMULA_MODELS)\n}\n\nexport interface DownloadProgress {\n spec: ModelSpec\n /** 현재까지 다운로드한 바이트 */\n downloaded: number\n /** 전체 바이트 (알 수 있으면) */\n total: number | null\n /** \"download\" | \"verify\" | \"done\" | \"skip\" */\n phase: \"download\" | \"verify\" | \"done\" | \"skip\" | \"error\"\n message?: string\n}\n\nexport type ProgressHandler = (p: DownloadProgress) => void\n\n/**\n * 지정 모델군을 다운로드/검증한다.\n * - 이미 있고 SHA 일치 → skip\n * - 없음 → 다운로드 후 SHA 검증. 실패 시 파일 삭제.\n */\nexport async function ensureModelsIn(dir: string, specs: ReadonlyArray<ModelSpec>, onProgress?: ProgressHandler): Promise<void> {\n await mkdir(dir, { recursive: true })\n\n for (const spec of specs) {\n const localPath = join(dir, spec.filename)\n\n if (await isExistingValid(localPath, spec.sha256)) {\n onProgress?.({\n spec,\n downloaded: 0,\n total: null,\n phase: \"skip\",\n message: \"이미 존재 + SHA 일치\",\n })\n continue\n }\n\n // 기존 파일 있지만 SHA 불일치 → 삭제\n try {\n await unlink(localPath)\n } catch {\n // 없을 수 있음\n }\n\n await downloadToFile(spec, localPath, onProgress)\n }\n}\n\n/** 필요한 모든 수식 모델을 다운로드/검증한다. */\nexport async function ensureFormulaModels(onProgress?: ProgressHandler): Promise<void> {\n return ensureModelsIn(getFormulaModelsDir(), ALL_FORMULA_MODELS, onProgress)\n}\n\n/** 단일 모델만 확인/다운로드 (진단 UI 에서 개별 상태 새로고침용) */\nexport async function ensureSingleModel(spec: ModelSpec, onProgress?: ProgressHandler): Promise<void> {\n const dir = getFormulaModelsDir()\n await mkdir(dir, { recursive: true })\n const localPath = join(dir, spec.filename)\n if (await isExistingValid(localPath, spec.sha256)) {\n onProgress?.({ spec, downloaded: 0, total: null, phase: \"skip\" })\n return\n }\n try {\n await unlink(localPath)\n } catch {}\n await downloadToFile(spec, localPath, onProgress)\n}\n\nasync function isExistingValid(localPath: string, sha256Expected: string): Promise<boolean> {\n try {\n const s = await stat(localPath)\n if (!s.isFile() || s.size === 0) return false\n } catch {\n return false\n }\n try {\n const actual = await sha256OfFile(localPath)\n return actual === sha256Expected\n } catch {\n return false\n }\n}\n\nasync function downloadToFile(\n spec: ModelSpec,\n localPath: string,\n onProgress?: ProgressHandler,\n): Promise<void> {\n // 먼저 .part 로 받고 검증 후 rename — 중단된 다운로드가 \"정상 파일\"로 오인되는 걸 방지\n const partPath = `${localPath}.part`\n await mkdir(dirname(localPath), { recursive: true })\n\n const resp = await fetch(spec.url, {\n headers: {\n // HF CDN 은 UA 없으면 가끔 403 을 뱉는다\n \"User-Agent\": \"kordoc-formula-ocr/1.0 (+https://github.com/chrisryugj/kordoc)\",\n },\n })\n if (!resp.ok || !resp.body) {\n throw new Error(\n `${spec.name} 다운로드 실패: HTTP ${resp.status} ${resp.statusText} (${spec.url})`,\n )\n }\n\n const lenHeader = resp.headers.get(\"content-length\")\n const total = lenHeader ? Number.parseInt(lenHeader, 10) : null\n let downloaded = 0\n\n const ws = createWriteStream(partPath)\n try {\n const reader = Readable.fromWeb(resp.body as unknown as import(\"stream/web\").ReadableStream)\n reader.on(\"data\", (chunk: Buffer | Uint8Array) => {\n downloaded += chunk.length\n onProgress?.({\n spec,\n downloaded,\n total,\n phase: \"download\",\n })\n })\n await pipeline(reader, ws)\n } catch (e) {\n try {\n await unlink(partPath)\n } catch {}\n throw new Error(`${spec.name} 스트리밍 실패: ${(e as Error).message}`)\n }\n\n onProgress?.({\n spec,\n downloaded,\n total,\n phase: \"verify\",\n })\n\n // SHA 검증\n let actual: string\n try {\n actual = await sha256OfFile(partPath)\n } catch (e) {\n try {\n await unlink(partPath)\n } catch {}\n throw new Error(`${spec.name} SHA 계산 실패: ${(e as Error).message}`)\n }\n\n if (actual !== spec.sha256) {\n try {\n await unlink(partPath)\n } catch {}\n throw new Error(\n `${spec.name} SHA256 mismatch: expected ${spec.sha256}, got ${actual} — 모델 URL 이 오염되었거나 전송 중 손상되었습니다.`,\n )\n }\n\n await rename(partPath, localPath)\n onProgress?.({\n spec,\n downloaded,\n total,\n phase: \"done\",\n })\n}\n\nasync function sha256OfFile(p: string): Promise<string> {\n const h = createHash(\"sha256\")\n const stream = createReadStream(p)\n await pipeline(stream, async function* (src) {\n for await (const chunk of src) {\n h.update(chunk)\n // yield 안 함 (소비만)\n }\n })\n return h.digest(\"hex\")\n}\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,EAAC;AACT,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,IAAC;AACT,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,IAAC;AACT,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,IAAC;AACT,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
|
import {
|
|
3
3
|
KordocError
|
|
4
|
-
} from "./chunk-
|
|
4
|
+
} from "./chunk-FSXLSHCW.js";
|
|
5
5
|
|
|
6
6
|
// src/hwpx/profile-io.ts
|
|
7
7
|
import { z } from "zod";
|
|
@@ -89,4 +89,4 @@ export {
|
|
|
89
89
|
formatProfileSchema,
|
|
90
90
|
parseFormatProfileJson
|
|
91
91
|
};
|
|
92
|
-
//# sourceMappingURL=chunk-
|
|
92
|
+
//# sourceMappingURL=chunk-PLIY23LY.js.map
|
|
@@ -1,12 +1,18 @@
|
|
|
1
1
|
#!/usr/bin/env node
|
|
2
2
|
import {
|
|
3
3
|
KordocError
|
|
4
|
-
} from "./chunk-
|
|
4
|
+
} from "./chunk-FSXLSHCW.js";
|
|
5
5
|
|
|
6
6
|
// src/hwpx/parser-shared.ts
|
|
7
7
|
import { DOMParser } from "@xmldom/xmldom";
|
|
8
8
|
var MAX_DECOMPRESS_SIZE = 100 * 1024 * 1024;
|
|
9
9
|
var MAX_ZIP_ENTRIES = 500;
|
|
10
|
+
var ZipBombError = class extends KordocError {
|
|
11
|
+
constructor(message) {
|
|
12
|
+
super(message);
|
|
13
|
+
this.name = "ZipBombError";
|
|
14
|
+
}
|
|
15
|
+
};
|
|
10
16
|
function clampSpan(val, max) {
|
|
11
17
|
return Math.max(1, Math.min(val, max));
|
|
12
18
|
}
|
|
@@ -42,14 +48,15 @@ function findChildByLocalName(parent, name) {
|
|
|
42
48
|
}
|
|
43
49
|
return null;
|
|
44
50
|
}
|
|
45
|
-
function extractTextFromNode(node) {
|
|
51
|
+
function extractTextFromNode(node, depth = 0) {
|
|
46
52
|
let result = "";
|
|
53
|
+
if (depth > MAX_XML_DEPTH) return result;
|
|
47
54
|
const children = node.childNodes;
|
|
48
55
|
if (!children) return result;
|
|
49
56
|
for (let i = 0; i < children.length; i++) {
|
|
50
57
|
const child = children[i];
|
|
51
58
|
if (child.nodeType === 3) result += child.textContent || "";
|
|
52
|
-
else if (child.nodeType === 1) result += extractTextFromNode(child);
|
|
59
|
+
else if (child.nodeType === 1) result += extractTextFromNode(child, depth + 1);
|
|
53
60
|
}
|
|
54
61
|
return result.trim();
|
|
55
62
|
}
|
|
@@ -254,6 +261,7 @@ function parseRenderStyles(headXml) {
|
|
|
254
261
|
export {
|
|
255
262
|
MAX_DECOMPRESS_SIZE,
|
|
256
263
|
MAX_ZIP_ENTRIES,
|
|
264
|
+
ZipBombError,
|
|
257
265
|
clampSpan,
|
|
258
266
|
MAX_XML_DEPTH,
|
|
259
267
|
createSectionShared,
|
|
@@ -265,4 +273,4 @@ export {
|
|
|
265
273
|
DEFAULT_CHAR,
|
|
266
274
|
parseRenderStyles
|
|
267
275
|
};
|
|
268
|
-
//# sourceMappingURL=chunk-
|
|
276
|
+
//# sourceMappingURL=chunk-V6U6YS5B.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"sources":["../src/hwpx/parser-shared.ts","../src/render/head-styles.ts"],"sourcesContent":["/**\n * HWPX 파서 공유 상수/타입/유틸 (parser.ts에서 분리).\n * ZIP 한도, 섹션 공유 상태(SectionShared), walk 컨텍스트(WalkCtx), XML 유틸.\n */\n\nimport { DOMParser } from \"@xmldom/xmldom\"\nimport { KordocError } from \"../utils.js\"\nimport type { CellContext, IRBlock, ParseWarning } from \"../types.js\"\n// WalkCtx.styleMap 타입 참조 — 타입 전용이라 styles.ts와의 순환은 컴파일 시 소거됨\nimport type { HwpxStyleMap } from \"./styles.js\"\n\nexport const MAX_DECOMPRESS_SIZE = 100 * 1024 * 1024\n/** 손상 ZIP 복구 시 최대 엔트리 수 */\nexport const MAX_ZIP_ENTRIES = 500\n\n/** ZIP bomb 가드 전용 에러 — per-section catch가 XML fatalError(PARTIAL_PARSE로 강등)와\n * 구분해 이것만 재던진다. KordocError 서브클래스라 sanitizeError allowlist는 그대로 통과 */\nexport class ZipBombError extends KordocError {\n constructor(message: string) {\n super(message)\n this.name = \"ZipBombError\"\n }\n}\n\n/** colSpan/rowSpan을 안전한 범위로 클램핑 */\nexport function clampSpan(val: number, max: number): number {\n return Math.max(1, Math.min(val, max))\n}\n\n/** XML DOM 재귀 최대 깊이 — 악성 파일의 스택 오버플로 방지.\n * 좌표계가 다른 hwp5 MAX_NEST_DEPTH(8, 표 중첩 단계)·filler/소스맵 16\n * (표 중첩 단계)과 달리 이건 \"XML 요소\" 깊이라 표 1단이 여러 depth를\n * 소모한다 — 상수 통일 금지 (의미가 다름) */\nexport const MAX_XML_DEPTH = 200\n\n/** 셀 컨텍스트 확장 — 중첩표/이미지/다중문단 블록과 제목셀 여부를 IRCell로 전달 (v3.0) */\nexport interface CellCtxEx extends CellContext {\n blocks?: IRBlock[]\n /** 중첩표/이미지 등 구조 콘텐츠 존재 — true일 때만 IRCell.blocks로 attach */\n hasStructure?: boolean\n isHeader?: boolean\n}\n\nexport interface TableState {\n rows: CellContext[][]\n currentRow: CellContext[]\n cell: CellCtxEx | null\n /** hp:caption 텍스트 — IRTable.caption으로 전달 (v3.0) */\n caption?: string\n}\n\n/** 섹션 간 공유 상태 — 자동번호 카운터, 머리말/꼬리말, 변경추적 */\nexport interface SectionShared {\n /** numbering id → 레벨별(1..10) 카운터. -1 = 미사용(start값으로 초기화 — 0은 start=\"0\"의 유효값) */\n numState: Map<string, number[]>\n pageText: { headers: string[]; footers: string[] }\n track: { deleteDepth: number; warned: boolean }\n /** content.hpf kordoc-layout 메타 (\"default\"|\"gongmun\") — 자사 생성 파일 왕복 채널\n * 게이트. null/미설정 = 외래 파일 (id 기반 인라인 강조·인용 복원 꺼짐) */\n kordocLayout?: string | null\n /** 표 후행 빈 열(앵커 있는 입력란) 보존 — ParseOptions.keepTrailingEmptyCols (#47) */\n keepTrailingEmptyCols?: boolean\n}\n\nexport function createSectionShared(): SectionShared {\n return { numState: new Map(), pageText: { headers: [], footers: [] }, track: { deleteDepth: 0, warned: false } }\n}\n\n/** walk 함수들이 공유하는 파싱 컨텍스트 — 개별 optional 파라미터를 하나로 묶어 시그니처 안정화 */\nexport interface WalkCtx {\n styleMap?: HwpxStyleMap\n warnings?: ParseWarning[]\n sectionNum?: number\n shared: SectionShared\n /** secPr outlineShapeIDRef — 개요(OUTLINE) 문단이 사용하는 numbering id */\n outlineNumId?: string\n}\n\n/** xmldom DOMParser 생성 — onError 콜백으로 malformed XML 경고 수집 */\nexport function createXmlParser(warnings?: ParseWarning[]): DOMParser {\n return new DOMParser({\n onError(level: \"warning\" | \"error\" | \"fatalError\", msg: string) {\n if (level === \"fatalError\") throw new KordocError(`XML 파싱 실패: ${msg}`)\n warnings?.push({ code: \"MALFORMED_XML\", message: `XML ${level === \"warning\" ? \"경고\" : \"오류\"}: ${msg}` })\n },\n })\n}\n\n/** 수집된 머리말/꼬리말을 본문 앞/뒤 문단으로 배치 */\nexport function applyPageText(blocks: IRBlock[], shared: SectionShared): void {\n const { headers, footers } = shared.pageText\n if (headers.length > 0) {\n blocks.unshift(...headers.map(t => ({ type: \"paragraph\" as const, text: t, pageNumber: 1 })))\n }\n if (footers.length > 0) {\n blocks.push(...footers.map(t => ({ type: \"paragraph\" as const, text: t })))\n }\n}\n\n/** 자식 중 지정된 localName(접두사 제거)을 가진 첫 번째 Element 반환 */\nexport function findChildByLocalName(parent: Element, name: string): Element | null {\n const children = parent.childNodes\n if (!children) return null\n for (let i = 0; i < children.length; i++) {\n const ch = children[i] as Element\n if (ch.nodeType !== 1) continue\n const tag = (ch.tagName || ch.localName || \"\").replace(/^[^:]+:/, \"\")\n if (tag === name) return ch\n }\n return null\n}\n\n/** 노드 내 모든 텍스트를 재귀적으로 추출 (MAX_XML_DEPTH 가드 — 악성 심층 XML 스택 오버플로 방지) */\nexport function extractTextFromNode(node: Node, depth: number = 0): string {\n let result = \"\"\n if (depth > MAX_XML_DEPTH) return result\n const children = node.childNodes\n if (!children) return result\n for (let i = 0; i < children.length; i++) {\n const child = children[i]\n if (child.nodeType === 3) result += child.textContent || \"\"\n else if (child.nodeType === 1) result += extractTextFromNode(child, depth + 1)\n }\n return result.trim()\n}\n","/**\n * 레이아웃 보존 렌더 — header.xml 스타일 테이블 (charPr/paraPr/borderFill).\n * 파서 IR과 무관하게 렌더에 필요한 속성만 추출한다.\n */\n\nimport { createXmlParser, findChildByLocalName } from \"../hwpx/parser-shared.js\"\n\nexport interface RenderCharStyle {\n /** 글자 크기 (1/100pt — 1000 = 10pt) */\n height: number\n bold: boolean\n italic: boolean\n underline: boolean\n /** #RRGGBB — 검정이면 undefined */\n color?: string\n /** 장평 % */\n ratio: number\n /** 자간 % */\n spacing: number\n /** CSS font-family 스택 — charPr fontRef(hangul)에서 해석. 없으면 root 기본 폴백 */\n fontFamily?: string\n /** HWP 원본 글꼴명 (fontfaces face) — reflow 폭 테이블 선택용 */\n face?: string\n}\n\nexport type ParaAlign = \"JUSTIFY\" | \"LEFT\" | \"RIGHT\" | \"CENTER\" | \"DISTRIBUTE\" | \"DISTRIBUTE_SPACE\"\n\n/** reflow(Tier-2)용 문단 기하 — 줄간격·여백. paraPr `<hh:margin>`·`<hh:lineSpacing>`은\n * `<hp:switch>/<hp:case>` 안(손자)이라 재귀 탐색으로 뽑는다. 단위 HWPUNIT. */\nexport interface RenderParaGeom {\n /** PERCENT(기본 160=160%) / FIXED / BETWEEN_LINES / AT_LEAST */\n lineSpacingType: string\n lineSpacingValue: number\n /** 왼쪽 들여쓰기(HWPUNIT) */\n marginLeft: number\n marginRight: number\n /** 첫 줄 들여쓰기/내어쓰기 — 음수=둘째 줄부터 더 들어감(hanging) */\n marginIntent: number\n /** 문단 위 간격(prev) */\n spaceBefore: number\n /** 문단 아래 간격(next) */\n spaceAfter: number\n /** breakSetting의 한글 줄나눔 — 'keep'(어절)/'charAll'(글자). 없으면 호출자 모드.\n * 주의: 저장 속성 의미는 이름과 반대 — breakNonLatinWord \"BREAK_WORD\"=어절,\n * \"KEEP_WORD\"=글자 (한글 COM 실렌더 실측) */\n wrapMode?: \"keep\" | \"charAll\"\n}\n\nexport const DEFAULT_PARA_GEOM: RenderParaGeom = {\n lineSpacingType: \"PERCENT\", lineSpacingValue: 160,\n marginLeft: 0, marginRight: 0, marginIntent: 0, spaceBefore: 0, spaceAfter: 0,\n}\n\nexport interface RenderBorderEdge {\n /** SOLID/DASH/DOT… — NONE은 edge 자체를 생략 */\n type: string\n /** pt 단위 굵기 */\n widthPt: number\n color: string\n}\n\nexport interface RenderBorderFill {\n left?: RenderBorderEdge\n right?: RenderBorderEdge\n top?: RenderBorderEdge\n bottom?: RenderBorderEdge\n /** 배경색 #RRGGBB (없으면 undefined) */\n fill?: string\n}\n\nexport interface RenderStyles {\n charPr: Map<string, RenderCharStyle>\n paraAlign: Map<string, ParaAlign>\n /** reflow용 문단 기하 (줄간격·여백) — id별 */\n paraGeom: Map<string, RenderParaGeom>\n borderFill: Map<string, RenderBorderFill>\n}\n\nexport const DEFAULT_CHAR: RenderCharStyle = { height: 1000, bold: false, italic: false, underline: false, ratio: 100, spacing: 0 }\n\n// ─── 글꼴 매핑 (HWP fontfaces face → CSS font-family 스택) ──────────────\n//\n// HWPX charPr 는 <hh:fontRef hangul=\"N\">으로 fontfaces 테이블의 실제 글꼴명을 가리킨다.\n// 렌더는 이 글꼴명을 SVG <text font-family>로 내보내 뷰어(WebView2/브라우저)가 원본과\n// 같은 글꼴로 그리게 한다. 미설치 글꼴은 계열(명조=serif / 고딕=sans) 폴백으로 수렴시켜\n// 최소한 획 계열(바탕↔돋움)은 원본과 일치시킨다 — 공문서 제목 고딕이 바탕체로 나오던 회귀 해소.\n\n/** 자주 쓰는 글꼴의 별칭 스택 (한글명↔영문 시스템명 병기 — 뷰어 OS별 등록명 차이 흡수) */\nconst FONT_ALIASES: Record<string, string> = {\n \"함초롬바탕\": \"'HCR Batang','함초롬바탕','한컴바탕'\",\n \"한컴바탕\": \"'HCR Batang','함초롬바탕','한컴바탕'\",\n \"함초롬돋움\": \"'HCR Dotum','함초롬돋움','한컴돋움'\",\n \"한컴돋움\": \"'HCR Dotum','함초롬돋움','한컴돋움'\",\n \"맑은 고딕\": \"'Malgun Gothic','맑은 고딕'\",\n \"맑은고딕\": \"'Malgun Gothic','맑은 고딕'\",\n \"굴림\": \"'Gulim','굴림'\",\n \"굴림체\": \"'GulimChe','굴림체','Gulim'\",\n \"돋움\": \"'Dotum','돋움'\",\n \"돋움체\": \"'DotumChe','돋움체','Dotum'\",\n \"바탕\": \"'Batang','바탕'\",\n \"바탕체\": \"'BatangChe','바탕체','Batang'\",\n \"궁서\": \"'Gungsuh','궁서'\",\n \"궁서체\": \"'GungsuhChe','궁서체','Gungsuh'\",\n \"나눔고딕\": \"'NanumGothic','나눔고딕'\",\n \"나눔명조\": \"'NanumMyeongjo','나눔명조'\",\n \"맑은 고딕 Semilight\": \"'Malgun Gothic Semilight','맑은 고딕'\",\n}\n\n/** 명조/바탕 계열(serif) 여부 — 아니면 고딕/돋움(sans)으로 본다 */\nfunction isSerifFace(face: string): boolean {\n return /바탕|명조|Batang|Myeong|Mincho|궁서|Gungsuh|Serif|신명|순명|Song|송/i.test(face)\n}\n\n/** CSS font-family 토큰 인용 — 영숫자·하이픈만이면 무인용, 그 외(공백·한글)는 작은따옴표 */\nfunction cssQuote(name: string): string {\n return /^[A-Za-z0-9_-]+$/.test(name) ? name : `'${name.replace(/['\"\\\\]/g, \"\")}'`\n}\n\n/**\n * HWP 글꼴명 → CSS font-family 스택. 정확한 글꼴 먼저, 이어서 계열 폴백(명조=serif /\n * 고딕=sans, 각 한국어 렌더 가능 글꼴 포함)을 붙여 미설치 시에도 획 계열을 보존한다.\n */\nexport function hwpFaceToCssStack(face: string | null | undefined): string {\n const trimmed = (face ?? \"\").trim()\n if (!trimmed) return \"\"\n const generic = isSerifFace(trimmed)\n ? \"'HCR Batang','Batang','Noto Serif KR',serif\"\n : \"'Malgun Gothic','HCR Dotum','Noto Sans KR',sans-serif\"\n const head = FONT_ALIASES[trimmed] ?? cssQuote(trimmed)\n return `${head},${generic}`\n}\n\n/** fontfaces 테이블에서 HANGUL 그룹의 font id → 글꼴명 맵 (없으면 첫 그룹) */\nfunction collectHangulFonts(root: Element): Map<string, string> {\n const map = new Map<string, string>()\n const findFaces = (el: Element, depth: number): Element | null => {\n if (depth > 24) return null\n for (const ch of Array.from(el.childNodes)) {\n if (ch.nodeType !== 1) continue\n const e = ch as Element\n if ((e.tagName || \"\").replace(/^[^:]+:/, \"\") === \"fontfaces\") return e\n const f = findFaces(e, depth + 1)\n if (f) return f\n }\n return null\n }\n const faces = findFaces(root, 0)\n if (!faces) return map\n let group: Element | null = null\n let firstGroup: Element | null = null\n for (const ch of Array.from(faces.childNodes)) {\n if (ch.nodeType !== 1) continue\n const e = ch as Element\n if ((e.tagName || \"\").replace(/^[^:]+:/, \"\") !== \"fontface\") continue\n if (!firstGroup) firstGroup = e\n if ((e.getAttribute(\"lang\") ?? \"\").toUpperCase() === \"HANGUL\") { group = e; break }\n }\n group = group ?? firstGroup\n if (!group) return map\n for (const ch of Array.from(group.childNodes)) {\n if (ch.nodeType !== 1) continue\n const e = ch as Element\n if ((e.tagName || \"\").replace(/^[^:]+:/, \"\") !== \"font\") continue\n const id = e.getAttribute(\"id\")\n const face = e.getAttribute(\"face\")\n if (id != null && face) map.set(id, face)\n }\n return map\n}\n\n/** \"0.12 mm\" / \"0.5 mm\" → pt */\nfunction borderWidthPt(v: string | null | undefined): number {\n const n = parseFloat(v ?? \"\")\n if (!Number.isFinite(n)) return 0.34\n return n * 2.834645 // mm → pt\n}\n\nfunction parseEdge(el: Element | null): RenderBorderEdge | undefined {\n if (!el) return undefined\n const type = el.getAttribute(\"type\") ?? \"NONE\"\n if (type === \"NONE\") return undefined\n return { type, widthPt: borderWidthPt(el.getAttribute(\"width\")), color: el.getAttribute(\"color\") ?? \"#000000\" }\n}\n\n/** 서브트리에서 localName이 일치하는 첫 요소를 재귀 탐색 (switch/case 래핑 대응) */\nfunction findDeep(el: Element, name: string, depth = 0): Element | null {\n if (depth > 32) return null\n const children = el.childNodes\n if (!children) return null\n for (let i = 0; i < children.length; i++) {\n const ch = children[i]\n if (ch.nodeType !== 1) continue\n const e = ch as Element\n if ((e.tagName || \"\").replace(/^[^:]+:/, \"\") === name) return e\n const found = findDeep(e, name, depth + 1)\n if (found) return found\n }\n return null\n}\n\n/** paraPr 요소 → 문단 기하 (margin·lineSpacing은 hp:switch/case 안이라 재귀 탐색) */\nfunction parseParaGeom(el: Element): RenderParaGeom {\n const g: RenderParaGeom = { ...DEFAULT_PARA_GEOM }\n const ls = findDeep(el, \"lineSpacing\")\n if (ls) {\n g.lineSpacingType = ls.getAttribute(\"type\") ?? \"PERCENT\"\n g.lineSpacingValue = Number(ls.getAttribute(\"value\")) || 160\n }\n const margin = findDeep(el, \"margin\")\n if (margin) {\n const v = (n: string): number => {\n const c = findDeep(margin, n)\n return c ? Number(c.getAttribute(\"value\")) || 0 : 0\n }\n g.marginLeft = v(\"left\")\n g.marginRight = v(\"right\")\n g.marginIntent = v(\"intent\")\n g.spaceBefore = v(\"prev\")\n g.spaceAfter = v(\"next\")\n }\n const bs = findDeep(el, \"breakSetting\")\n if (bs) {\n // 속성 의미 역전 주의(위 RenderParaGeom.wrapMode 주석): BREAK_WORD=어절, KEEP_WORD=글자\n const nl = bs.getAttribute(\"breakNonLatinWord\")\n if (nl === \"BREAK_WORD\") g.wrapMode = \"keep\"\n else if (nl === \"KEEP_WORD\") g.wrapMode = \"charAll\"\n }\n return g\n}\n\n/** header.xml(구버전 head.xml) → 렌더용 스타일 맵 */\nexport function parseRenderStyles(headXml: string): RenderStyles {\n const styles: RenderStyles = { charPr: new Map(), paraAlign: new Map(), paraGeom: new Map(), borderFill: new Map() }\n const doc = createXmlParser().parseFromString(headXml, \"text/xml\")\n const root = doc.documentElement as unknown as Element | null\n if (!root) return styles\n\n // fontfaces(HANGUL) 선파싱 — charPr fontRef 해석에 쓴다\n const hangulFonts = collectHangulFonts(root)\n\n const walk = (el: Element): void => {\n const tag = (el.tagName || \"\").replace(/^[^:]+:/, \"\")\n if (tag === \"charPr\") {\n const id = el.getAttribute(\"id\")\n if (id != null) {\n const ratioEl = findChildByLocalName(el, \"ratio\")\n const spacingEl = findChildByLocalName(el, \"spacing\")\n const underlineEl = findChildByLocalName(el, \"underline\")\n const textColor = el.getAttribute(\"textColor\")\n const fontRef = findChildByLocalName(el, \"fontRef\")\n const fontId = fontRef?.getAttribute(\"hangul\") ?? fontRef?.getAttribute(\"latin\")\n const face = fontId != null ? hangulFonts.get(fontId) : undefined\n styles.charPr.set(id, {\n height: Number(el.getAttribute(\"height\")) || 1000,\n bold: findChildByLocalName(el, \"bold\") != null,\n italic: findChildByLocalName(el, \"italic\") != null,\n underline: underlineEl != null && (underlineEl.getAttribute(\"type\") ?? \"NONE\") !== \"NONE\",\n color: textColor && textColor !== \"#000000\" && textColor.toLowerCase() !== \"none\" ? textColor : undefined,\n ratio: Number(ratioEl?.getAttribute(\"hangul\")) || 100,\n spacing: Number(spacingEl?.getAttribute(\"hangul\")) || 0,\n fontFamily: face ? hwpFaceToCssStack(face) : undefined,\n face,\n })\n }\n } else if (tag === \"paraPr\") {\n const id = el.getAttribute(\"id\")\n if (id != null) {\n const align = findChildByLocalName(el, \"align\")\n styles.paraAlign.set(id, (align?.getAttribute(\"horizontal\") as ParaAlign) || \"JUSTIFY\")\n styles.paraGeom.set(id, parseParaGeom(el))\n }\n } else if (tag === \"borderFill\") {\n const id = el.getAttribute(\"id\")\n if (id != null) {\n const bf: RenderBorderFill = {\n left: parseEdge(findChildByLocalName(el, \"leftBorder\")),\n right: parseEdge(findChildByLocalName(el, \"rightBorder\")),\n top: parseEdge(findChildByLocalName(el, \"topBorder\")),\n bottom: parseEdge(findChildByLocalName(el, \"bottomBorder\")),\n }\n const fillBrush = findChildByLocalName(el, \"fillBrush\")\n const winBrush = fillBrush ? findChildByLocalName(fillBrush, \"winBrush\") : null\n const face = winBrush?.getAttribute(\"faceColor\")\n if (face && face.toLowerCase() !== \"none\") bf.fill = face\n styles.borderFill.set(id, bf)\n }\n }\n const children = el.childNodes\n for (let i = 0; i < children.length; i++) {\n const ch = children[i]\n if (ch.nodeType === 1) walk(ch as Element)\n }\n }\n walk(root)\n return styles\n}\n"],"mappings":";;;;;;AAKA,SAAS,iBAAiB;AAMnB,IAAM,sBAAsB,MAAM,OAAO;AAEzC,IAAM,kBAAkB;AAIxB,IAAM,eAAN,cAA2B,YAAY;AAAA,EAC5C,YAAY,SAAiB;AAC3B,UAAM,OAAO;AACb,SAAK,OAAO;AAAA,EACd;AACF;AAGO,SAAS,UAAU,KAAa,KAAqB;AAC1D,SAAO,KAAK,IAAI,GAAG,KAAK,IAAI,KAAK,GAAG,CAAC;AACvC;AAMO,IAAM,gBAAgB;AA+BtB,SAAS,sBAAqC;AACnD,SAAO,EAAE,UAAU,oBAAI,IAAI,GAAG,UAAU,EAAE,SAAS,CAAC,GAAG,SAAS,CAAC,EAAE,GAAG,OAAO,EAAE,aAAa,GAAG,QAAQ,MAAM,EAAE;AACjH;AAaO,SAAS,gBAAgB,UAAsC;AACpE,SAAO,IAAI,UAAU;AAAA,IACnB,QAAQ,OAA2C,KAAa;AAC9D,UAAI,UAAU,aAAc,OAAM,IAAI,YAAY,kCAAc,GAAG,EAAE;AACrE,gBAAU,KAAK,EAAE,MAAM,iBAAiB,SAAS,OAAO,UAAU,YAAY,iBAAO,cAAI,KAAK,GAAG,GAAG,CAAC;AAAA,IACvG;AAAA,EACF,CAAC;AACH;AAGO,SAAS,cAAc,QAAmB,QAA6B;AAC5E,QAAM,EAAE,SAAS,QAAQ,IAAI,OAAO;AACpC,MAAI,QAAQ,SAAS,GAAG;AACtB,WAAO,QAAQ,GAAG,QAAQ,IAAI,QAAM,EAAE,MAAM,aAAsB,MAAM,GAAG,YAAY,EAAE,EAAE,CAAC;AAAA,EAC9F;AACA,MAAI,QAAQ,SAAS,GAAG;AACtB,WAAO,KAAK,GAAG,QAAQ,IAAI,QAAM,EAAE,MAAM,aAAsB,MAAM,EAAE,EAAE,CAAC;AAAA,EAC5E;AACF;AAGO,SAAS,qBAAqB,QAAiB,MAA8B;AAClF,QAAM,WAAW,OAAO;AACxB,MAAI,CAAC,SAAU,QAAO;AACtB,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,UAAM,KAAK,SAAS,CAAC;AACrB,QAAI,GAAG,aAAa,EAAG;AACvB,UAAM,OAAO,GAAG,WAAW,GAAG,aAAa,IAAI,QAAQ,WAAW,EAAE;AACpE,QAAI,QAAQ,KAAM,QAAO;AAAA,EAC3B;AACA,SAAO;AACT;AAGO,SAAS,oBAAoB,MAAY,QAAgB,GAAW;AACzE,MAAI,SAAS;AACb,MAAI,QAAQ,cAAe,QAAO;AAClC,QAAM,WAAW,KAAK;AACtB,MAAI,CAAC,SAAU,QAAO;AACtB,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,UAAM,QAAQ,SAAS,CAAC;AACxB,QAAI,MAAM,aAAa,EAAG,WAAU,MAAM,eAAe;AAAA,aAChD,MAAM,aAAa,EAAG,WAAU,oBAAoB,OAAO,QAAQ,CAAC;AAAA,EAC/E;AACA,SAAO,OAAO,KAAK;AACrB;;;AC5EO,IAAM,oBAAoC;AAAA,EAC/C,iBAAiB;AAAA,EAAW,kBAAkB;AAAA,EAC9C,YAAY;AAAA,EAAG,aAAa;AAAA,EAAG,cAAc;AAAA,EAAG,aAAa;AAAA,EAAG,YAAY;AAC9E;AA2BO,IAAM,eAAgC,EAAE,QAAQ,KAAM,MAAM,OAAO,QAAQ,OAAO,WAAW,OAAO,OAAO,KAAK,SAAS,EAAE;AAUlI,IAAM,eAAuC;AAAA,EAC3C,kCAAS;AAAA,EACT,4BAAQ;AAAA,EACR,kCAAS;AAAA,EACT,4BAAQ;AAAA,EACR,6BAAS;AAAA,EACT,4BAAQ;AAAA,EACR,gBAAM;AAAA,EACN,sBAAO;AAAA,EACP,gBAAM;AAAA,EACN,sBAAO;AAAA,EACP,gBAAM;AAAA,EACN,sBAAO;AAAA,EACP,gBAAM;AAAA,EACN,sBAAO;AAAA,EACP,4BAAQ;AAAA,EACR,4BAAQ;AAAA,EACR,uCAAmB;AACrB;AAGA,SAAS,YAAY,MAAuB;AAC1C,SAAO,4DAA4D,KAAK,IAAI;AAC9E;AAGA,SAAS,SAAS,MAAsB;AACtC,SAAO,mBAAmB,KAAK,IAAI,IAAI,OAAO,IAAI,KAAK,QAAQ,WAAW,EAAE,CAAC;AAC/E;AAMO,SAAS,kBAAkB,MAAyC;AACzE,QAAM,WAAW,QAAQ,IAAI,KAAK;AAClC,MAAI,CAAC,QAAS,QAAO;AACrB,QAAM,UAAU,YAAY,OAAO,IAC/B,gDACA;AACJ,QAAM,OAAO,aAAa,OAAO,KAAK,SAAS,OAAO;AACtD,SAAO,GAAG,IAAI,IAAI,OAAO;AAC3B;AAGA,SAAS,mBAAmB,MAAoC;AAC9D,QAAM,MAAM,oBAAI,IAAoB;AACpC,QAAM,YAAY,CAAC,IAAa,UAAkC;AAChE,QAAI,QAAQ,GAAI,QAAO;AACvB,eAAW,MAAM,MAAM,KAAK,GAAG,UAAU,GAAG;AAC1C,UAAI,GAAG,aAAa,EAAG;AACvB,YAAM,IAAI;AACV,WAAK,EAAE,WAAW,IAAI,QAAQ,WAAW,EAAE,MAAM,YAAa,QAAO;AACrE,YAAM,IAAI,UAAU,GAAG,QAAQ,CAAC;AAChC,UAAI,EAAG,QAAO;AAAA,IAChB;AACA,WAAO;AAAA,EACT;AACA,QAAM,QAAQ,UAAU,MAAM,CAAC;AAC/B,MAAI,CAAC,MAAO,QAAO;AACnB,MAAI,QAAwB;AAC5B,MAAI,aAA6B;AACjC,aAAW,MAAM,MAAM,KAAK,MAAM,UAAU,GAAG;AAC7C,QAAI,GAAG,aAAa,EAAG;AACvB,UAAM,IAAI;AACV,SAAK,EAAE,WAAW,IAAI,QAAQ,WAAW,EAAE,MAAM,WAAY;AAC7D,QAAI,CAAC,WAAY,cAAa;AAC9B,SAAK,EAAE,aAAa,MAAM,KAAK,IAAI,YAAY,MAAM,UAAU;AAAE,cAAQ;AAAG;AAAA,IAAM;AAAA,EACpF;AACA,UAAQ,SAAS;AACjB,MAAI,CAAC,MAAO,QAAO;AACnB,aAAW,MAAM,MAAM,KAAK,MAAM,UAAU,GAAG;AAC7C,QAAI,GAAG,aAAa,EAAG;AACvB,UAAM,IAAI;AACV,SAAK,EAAE,WAAW,IAAI,QAAQ,WAAW,EAAE,MAAM,OAAQ;AACzD,UAAM,KAAK,EAAE,aAAa,IAAI;AAC9B,UAAM,OAAO,EAAE,aAAa,MAAM;AAClC,QAAI,MAAM,QAAQ,KAAM,KAAI,IAAI,IAAI,IAAI;AAAA,EAC1C;AACA,SAAO;AACT;AAGA,SAAS,cAAc,GAAsC;AAC3D,QAAM,IAAI,WAAW,KAAK,EAAE;AAC5B,MAAI,CAAC,OAAO,SAAS,CAAC,EAAG,QAAO;AAChC,SAAO,IAAI;AACb;AAEA,SAAS,UAAU,IAAkD;AACnE,MAAI,CAAC,GAAI,QAAO;AAChB,QAAM,OAAO,GAAG,aAAa,MAAM,KAAK;AACxC,MAAI,SAAS,OAAQ,QAAO;AAC5B,SAAO,EAAE,MAAM,SAAS,cAAc,GAAG,aAAa,OAAO,CAAC,GAAG,OAAO,GAAG,aAAa,OAAO,KAAK,UAAU;AAChH;AAGA,SAAS,SAAS,IAAa,MAAc,QAAQ,GAAmB;AACtE,MAAI,QAAQ,GAAI,QAAO;AACvB,QAAM,WAAW,GAAG;AACpB,MAAI,CAAC,SAAU,QAAO;AACtB,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,UAAM,KAAK,SAAS,CAAC;AACrB,QAAI,GAAG,aAAa,EAAG;AACvB,UAAM,IAAI;AACV,SAAK,EAAE,WAAW,IAAI,QAAQ,WAAW,EAAE,MAAM,KAAM,QAAO;AAC9D,UAAM,QAAQ,SAAS,GAAG,MAAM,QAAQ,CAAC;AACzC,QAAI,MAAO,QAAO;AAAA,EACpB;AACA,SAAO;AACT;AAGA,SAAS,cAAc,IAA6B;AAClD,QAAM,IAAoB,EAAE,GAAG,kBAAkB;AACjD,QAAM,KAAK,SAAS,IAAI,aAAa;AACrC,MAAI,IAAI;AACN,MAAE,kBAAkB,GAAG,aAAa,MAAM,KAAK;AAC/C,MAAE,mBAAmB,OAAO,GAAG,aAAa,OAAO,CAAC,KAAK;AAAA,EAC3D;AACA,QAAM,SAAS,SAAS,IAAI,QAAQ;AACpC,MAAI,QAAQ;AACV,UAAM,IAAI,CAAC,MAAsB;AAC/B,YAAM,IAAI,SAAS,QAAQ,CAAC;AAC5B,aAAO,IAAI,OAAO,EAAE,aAAa,OAAO,CAAC,KAAK,IAAI;AAAA,IACpD;AACA,MAAE,aAAa,EAAE,MAAM;AACvB,MAAE,cAAc,EAAE,OAAO;AACzB,MAAE,eAAe,EAAE,QAAQ;AAC3B,MAAE,cAAc,EAAE,MAAM;AACxB,MAAE,aAAa,EAAE,MAAM;AAAA,EACzB;AACA,QAAM,KAAK,SAAS,IAAI,cAAc;AACtC,MAAI,IAAI;AAEN,UAAM,KAAK,GAAG,aAAa,mBAAmB;AAC9C,QAAI,OAAO,aAAc,GAAE,WAAW;AAAA,aAC7B,OAAO,YAAa,GAAE,WAAW;AAAA,EAC5C;AACA,SAAO;AACT;AAGO,SAAS,kBAAkB,SAA+B;AAC/D,QAAM,SAAuB,EAAE,QAAQ,oBAAI,IAAI,GAAG,WAAW,oBAAI,IAAI,GAAG,UAAU,oBAAI,IAAI,GAAG,YAAY,oBAAI,IAAI,EAAE;AACnH,QAAM,MAAM,gBAAgB,EAAE,gBAAgB,SAAS,UAAU;AACjE,QAAM,OAAO,IAAI;AACjB,MAAI,CAAC,KAAM,QAAO;AAGlB,QAAM,cAAc,mBAAmB,IAAI;AAE3C,QAAM,OAAO,CAAC,OAAsB;AAClC,UAAM,OAAO,GAAG,WAAW,IAAI,QAAQ,WAAW,EAAE;AACpD,QAAI,QAAQ,UAAU;AACpB,YAAM,KAAK,GAAG,aAAa,IAAI;AAC/B,UAAI,MAAM,MAAM;AACd,cAAM,UAAU,qBAAqB,IAAI,OAAO;AAChD,cAAM,YAAY,qBAAqB,IAAI,SAAS;AACpD,cAAM,cAAc,qBAAqB,IAAI,WAAW;AACxD,cAAM,YAAY,GAAG,aAAa,WAAW;AAC7C,cAAM,UAAU,qBAAqB,IAAI,SAAS;AAClD,cAAM,SAAS,SAAS,aAAa,QAAQ,KAAK,SAAS,aAAa,OAAO;AAC/E,cAAM,OAAO,UAAU,OAAO,YAAY,IAAI,MAAM,IAAI;AACxD,eAAO,OAAO,IAAI,IAAI;AAAA,UACpB,QAAQ,OAAO,GAAG,aAAa,QAAQ,CAAC,KAAK;AAAA,UAC7C,MAAM,qBAAqB,IAAI,MAAM,KAAK;AAAA,UAC1C,QAAQ,qBAAqB,IAAI,QAAQ,KAAK;AAAA,UAC9C,WAAW,eAAe,SAAS,YAAY,aAAa,MAAM,KAAK,YAAY;AAAA,UACnF,OAAO,aAAa,cAAc,aAAa,UAAU,YAAY,MAAM,SAAS,YAAY;AAAA,UAChG,OAAO,OAAO,SAAS,aAAa,QAAQ,CAAC,KAAK;AAAA,UAClD,SAAS,OAAO,WAAW,aAAa,QAAQ,CAAC,KAAK;AAAA,UACtD,YAAY,OAAO,kBAAkB,IAAI,IAAI;AAAA,UAC7C;AAAA,QACF,CAAC;AAAA,MACH;AAAA,IACF,WAAW,QAAQ,UAAU;AAC3B,YAAM,KAAK,GAAG,aAAa,IAAI;AAC/B,UAAI,MAAM,MAAM;AACd,cAAM,QAAQ,qBAAqB,IAAI,OAAO;AAC9C,eAAO,UAAU,IAAI,IAAK,OAAO,aAAa,YAAY,KAAmB,SAAS;AACtF,eAAO,SAAS,IAAI,IAAI,cAAc,EAAE,CAAC;AAAA,MAC3C;AAAA,IACF,WAAW,QAAQ,cAAc;AAC/B,YAAM,KAAK,GAAG,aAAa,IAAI;AAC/B,UAAI,MAAM,MAAM;AACd,cAAM,KAAuB;AAAA,UAC3B,MAAM,UAAU,qBAAqB,IAAI,YAAY,CAAC;AAAA,UACtD,OAAO,UAAU,qBAAqB,IAAI,aAAa,CAAC;AAAA,UACxD,KAAK,UAAU,qBAAqB,IAAI,WAAW,CAAC;AAAA,UACpD,QAAQ,UAAU,qBAAqB,IAAI,cAAc,CAAC;AAAA,QAC5D;AACA,cAAM,YAAY,qBAAqB,IAAI,WAAW;AACtD,cAAM,WAAW,YAAY,qBAAqB,WAAW,UAAU,IAAI;AAC3E,cAAM,OAAO,UAAU,aAAa,WAAW;AAC/C,YAAI,QAAQ,KAAK,YAAY,MAAM,OAAQ,IAAG,OAAO;AACrD,eAAO,WAAW,IAAI,IAAI,EAAE;AAAA,MAC9B;AAAA,IACF;AACA,UAAM,WAAW,GAAG;AACpB,aAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,YAAM,KAAK,SAAS,CAAC;AACrB,UAAI,GAAG,aAAa,EAAG,MAAK,EAAa;AAAA,IAC3C;AAAA,EACF;AACA,OAAK,IAAI;AACT,SAAO;AACT;","names":[]}
|
|
@@ -1,7 +1,6 @@
|
|
|
1
|
-
#!/usr/bin/env node
|
|
2
1
|
import {
|
|
3
2
|
sanitizeHref
|
|
4
|
-
} from "./chunk-
|
|
3
|
+
} from "./chunk-JP5YH3Z4.js";
|
|
5
4
|
|
|
6
5
|
// src/shared/pua.ts
|
|
7
6
|
var BMP_SYMBOL_MAP = {
|
|
@@ -121,8 +120,10 @@ var SUPPLEMENTARY_MAP = {
|
|
|
121
120
|
// ▸
|
|
122
121
|
985103: "\u2501",
|
|
123
122
|
// ━
|
|
124
|
-
985127: "\u25A0"
|
|
123
|
+
985127: "\u25A0",
|
|
125
124
|
// ■
|
|
125
|
+
984005: "\u25A1"
|
|
126
|
+
// □ 글머리 — HWP3→HWP5 한컴 변환본 보존 코드, 한컴오피스 표시값 (rhwp #1105)
|
|
126
127
|
};
|
|
127
128
|
function mapPuaChar(code) {
|
|
128
129
|
if (code >= 61472 && code <= 61695) {
|
|
@@ -145,11 +146,11 @@ function mapPuaText(text) {
|
|
|
145
146
|
// src/table/builder.ts
|
|
146
147
|
var MAX_COLS = 200;
|
|
147
148
|
var MAX_ROWS = 1e4;
|
|
148
|
-
function buildTable(rows) {
|
|
149
|
+
function buildTable(rows, options) {
|
|
149
150
|
if (rows.length > MAX_ROWS) rows = rows.slice(0, MAX_ROWS);
|
|
150
151
|
const numRows = rows.length;
|
|
151
152
|
const hasAddr = rows.some((row) => row.some((c) => c.colAddr !== void 0 && c.rowAddr !== void 0));
|
|
152
|
-
if (hasAddr) return buildTableDirect(rows, numRows);
|
|
153
|
+
if (hasAddr) return buildTableDirect(rows, numRows, options);
|
|
153
154
|
let maxCols = 0;
|
|
154
155
|
const tempOccupied = Array.from({ length: numRows }, () => []);
|
|
155
156
|
for (let rowIdx = 0; rowIdx < numRows; rowIdx++) {
|
|
@@ -172,6 +173,7 @@ function buildTable(rows) {
|
|
|
172
173
|
() => Array.from({ length: maxCols }, () => ({ text: "", colSpan: 1, rowSpan: 1 }))
|
|
173
174
|
);
|
|
174
175
|
const occupied = Array.from({ length: numRows }, () => Array(maxCols).fill(false));
|
|
176
|
+
const anchorCols = /* @__PURE__ */ new Set();
|
|
175
177
|
for (let rowIdx = 0; rowIdx < numRows; rowIdx++) {
|
|
176
178
|
let colIdx = 0;
|
|
177
179
|
let cellIdx = 0;
|
|
@@ -179,6 +181,7 @@ function buildTable(rows) {
|
|
|
179
181
|
while (colIdx < maxCols && occupied[rowIdx][colIdx]) colIdx++;
|
|
180
182
|
if (colIdx >= maxCols) break;
|
|
181
183
|
const cell = rows[rowIdx][cellIdx];
|
|
184
|
+
anchorCols.add(colIdx);
|
|
182
185
|
grid[rowIdx][colIdx] = {
|
|
183
186
|
text: cell.text.trim(),
|
|
184
187
|
colSpan: cell.colSpan,
|
|
@@ -193,9 +196,9 @@ function buildTable(rows) {
|
|
|
193
196
|
cellIdx++;
|
|
194
197
|
}
|
|
195
198
|
}
|
|
196
|
-
return trimAndReturn(grid, numRows, maxCols);
|
|
199
|
+
return trimAndReturn(grid, numRows, maxCols, anchorCols, options);
|
|
197
200
|
}
|
|
198
|
-
function buildTableDirect(rows, numRows) {
|
|
201
|
+
function buildTableDirect(rows, numRows, options) {
|
|
199
202
|
let maxCols = 0;
|
|
200
203
|
for (const row of rows) {
|
|
201
204
|
for (const cell of row) {
|
|
@@ -209,11 +212,13 @@ function buildTableDirect(rows, numRows) {
|
|
|
209
212
|
{ length: numRows },
|
|
210
213
|
() => Array.from({ length: maxCols }, () => ({ text: "", colSpan: 1, rowSpan: 1 }))
|
|
211
214
|
);
|
|
215
|
+
const anchorCols = /* @__PURE__ */ new Set();
|
|
212
216
|
for (const row of rows) {
|
|
213
217
|
for (const cell of row) {
|
|
214
218
|
const r = cell.rowAddr ?? 0;
|
|
215
219
|
const c = cell.colAddr ?? 0;
|
|
216
220
|
if (r >= numRows || c >= maxCols || r < 0 || c < 0) continue;
|
|
221
|
+
anchorCols.add(c);
|
|
217
222
|
grid[r][c] = { text: cell.text.trim(), colSpan: cell.colSpan, rowSpan: cell.rowSpan };
|
|
218
223
|
for (let dr = 0; dr < cell.rowSpan; dr++) {
|
|
219
224
|
for (let dc = 0; dc < cell.colSpan; dc++) {
|
|
@@ -225,13 +230,14 @@ function buildTableDirect(rows, numRows) {
|
|
|
225
230
|
}
|
|
226
231
|
}
|
|
227
232
|
}
|
|
228
|
-
return trimAndReturn(grid, numRows, maxCols);
|
|
233
|
+
return trimAndReturn(grid, numRows, maxCols, anchorCols, options);
|
|
229
234
|
}
|
|
230
|
-
function trimAndReturn(grid, numRows, maxCols) {
|
|
235
|
+
function trimAndReturn(grid, numRows, maxCols, anchorCols, options) {
|
|
231
236
|
let effectiveCols = maxCols;
|
|
232
237
|
while (effectiveCols > 0) {
|
|
233
238
|
const colEmpty = grid.every((row) => !row[effectiveCols - 1]?.text?.trim());
|
|
234
239
|
if (!colEmpty) break;
|
|
240
|
+
if (options?.keepAnchoredEmptyCols && anchorCols.has(effectiveCols - 1)) break;
|
|
235
241
|
effectiveCols--;
|
|
236
242
|
}
|
|
237
243
|
if (effectiveCols < maxCols && effectiveCols > 0) {
|
|
@@ -248,14 +254,14 @@ function convertTableToText(rows) {
|
|
|
248
254
|
function escapeGfm(text) {
|
|
249
255
|
const NUL = String.fromCharCode(0);
|
|
250
256
|
const spans = [];
|
|
251
|
-
const masked = text.replace(
|
|
257
|
+
const masked = text.replace(/!\[[^\]]*\]\([^)\n]*\)|\]\((?:https?:|mailto:|tel:|#)[^)\n]*\)|\$\$[^$]*\$\$|\$[^$\n]*\$/gi, (m) => {
|
|
252
258
|
spans.push(m);
|
|
253
259
|
return NUL + (spans.length - 1) + NUL;
|
|
254
260
|
});
|
|
255
|
-
const escaped = masked.replace(/([~*])/g, "\\$1");
|
|
261
|
+
const escaped = masked.replace(/([~*_`])/g, "\\$1");
|
|
256
262
|
return escaped.replace(new RegExp(NUL + "(\\d+)" + NUL, "g"), (_, n) => spans[Number(n)]);
|
|
257
263
|
}
|
|
258
|
-
var HWP_SHAPE_ALT_TEXT_RE =
|
|
264
|
+
var HWP_SHAPE_ALT_TEXT_RE = /^(?:모서리가 둥근 |둥근 )?(?:사각형|직사각형|정사각형|원|타원|삼각형|이등변 삼각형|직각 삼각형|선|직선|곡선|화살표|굵은 화살표|이중 화살표|오각형|육각형|팔각형|별|[4-8]점별|십자|십자형|구름|구름형|마름모|도넛|평행사변형|사다리꼴|부채꼴|호|반원|물결|번개|하트|빗금|블록 화살표|수식|표|그림|개체|그리기\s?개체|묶음\s?개체|글상자|수식\s?개체|OLE\s?개체)\s?입니다\.?$/gm;
|
|
259
265
|
function sanitizeText(text) {
|
|
260
266
|
let result = mapPuaText(text).replace(/[\u{F0000}-\u{FFFFD}]/gu, "").replace(HWP_SHAPE_ALT_TEXT_RE, "").replace(/ +/g, " ").trim();
|
|
261
267
|
if (result.length <= 30 && result.includes(" ")) {
|
|
@@ -478,7 +484,7 @@ function cellInnerHtml(cell) {
|
|
|
478
484
|
return sanitizeText(cell.text).replace(/\n/g, "<br>");
|
|
479
485
|
}
|
|
480
486
|
function containsInlineMath(text) {
|
|
481
|
-
return /(^|[^\\])\$(?=\S)(
|
|
487
|
+
return /(^|[^\\])\$(?=\S)(?:[^$\n\\]|\\.)+?\S\$/.test(text);
|
|
482
488
|
}
|
|
483
489
|
function tableContainsInlineMath(table) {
|
|
484
490
|
for (const row of table.cells) {
|
|
@@ -560,8 +566,15 @@ function tableToMarkdown(table) {
|
|
|
560
566
|
let pendingLabelRow = null;
|
|
561
567
|
for (let r = 0; r < display.length; r++) {
|
|
562
568
|
const row = display[r];
|
|
563
|
-
|
|
564
|
-
|
|
569
|
+
if (row.every((cell) => cell === "")) {
|
|
570
|
+
if (row.some((_, c) => skip.has(`${r},${c}`))) continue;
|
|
571
|
+
if (pendingLabelRow) {
|
|
572
|
+
uniqueRows.push(pendingLabelRow);
|
|
573
|
+
pendingLabelRow = null;
|
|
574
|
+
}
|
|
575
|
+
uniqueRows.push(row);
|
|
576
|
+
continue;
|
|
577
|
+
}
|
|
565
578
|
const nonEmptyCols = row.filter((cell) => cell !== "");
|
|
566
579
|
const hasSkipInRow = row.some((_, c) => skip.has(`${r},${c}`));
|
|
567
580
|
if (!hasSkipInRow && nonEmptyCols.length === 1 && row[0] !== "" && row.slice(1).every((c) => c === "")) {
|
|
@@ -587,11 +600,6 @@ function tableToMarkdown(table) {
|
|
|
587
600
|
return md.join("\n");
|
|
588
601
|
}
|
|
589
602
|
|
|
590
|
-
// src/types.ts
|
|
591
|
-
var HEADING_RATIO_H1 = 1.5;
|
|
592
|
-
var HEADING_RATIO_H2 = 1.3;
|
|
593
|
-
var HEADING_RATIO_H3 = 1.15;
|
|
594
|
-
|
|
595
603
|
// src/image/transcode.ts
|
|
596
604
|
import { deflateSync } from "zlib";
|
|
597
605
|
var CRC_TABLE = (() => {
|
|
@@ -702,12 +710,9 @@ function encodePng(width, height, rgba) {
|
|
|
702
710
|
out.set(iendChunk, o);
|
|
703
711
|
return out;
|
|
704
712
|
}
|
|
705
|
-
function escapeRegExp(s) {
|
|
706
|
-
return s.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
|
|
707
|
-
}
|
|
708
713
|
function inlineImagesIntoMarkdown(markdown, images, opts) {
|
|
709
714
|
const compress = opts?.compress !== false;
|
|
710
|
-
|
|
715
|
+
const uris = /* @__PURE__ */ new Map();
|
|
711
716
|
for (const img of images) {
|
|
712
717
|
let bytes = img.data;
|
|
713
718
|
let mime = img.mimeType;
|
|
@@ -718,14 +723,17 @@ function inlineImagesIntoMarkdown(markdown, images, opts) {
|
|
|
718
723
|
mime = "image/png";
|
|
719
724
|
}
|
|
720
725
|
}
|
|
721
|
-
|
|
722
|
-
const dataUri = `data:${mime};base64,${base64}`;
|
|
723
|
-
const name = escapeRegExp(img.filename);
|
|
724
|
-
const mdRe = new RegExp(`!\\[image\\]\\((?:images/)?${name}\\)`, "g");
|
|
725
|
-
out = out.replace(mdRe, () => ``);
|
|
726
|
-
const imgTagRe = new RegExp(`(<img\\b[^>]*\\bsrc=")(?:images/)?${name}(")`, "g");
|
|
727
|
-
out = out.replace(imgTagRe, (_m, pre, post) => `${pre}${dataUri}${post}`);
|
|
726
|
+
uris.set(img.filename, `data:${mime};base64,${Buffer.from(bytes).toString("base64")}`);
|
|
728
727
|
}
|
|
728
|
+
if (uris.size === 0) return markdown;
|
|
729
|
+
let out = markdown.replace(/!\[image\]\((?:images\/)?([^()\s]+)\)/g, (m, name) => {
|
|
730
|
+
const uri = uris.get(name);
|
|
731
|
+
return uri !== void 0 ? `` : m;
|
|
732
|
+
});
|
|
733
|
+
out = out.replace(/(<img\b[^>]*\bsrc=")(?:images\/)?([^"]*)(")/g, (m, pre, name, post) => {
|
|
734
|
+
const uri = uris.get(name);
|
|
735
|
+
return uri !== void 0 ? `${pre}${uri}${post}` : m;
|
|
736
|
+
});
|
|
729
737
|
return out;
|
|
730
738
|
}
|
|
731
739
|
|
|
@@ -738,10 +746,7 @@ export {
|
|
|
738
746
|
flattenLayoutTables,
|
|
739
747
|
dedupeRunningHeaders,
|
|
740
748
|
blocksToMarkdown,
|
|
741
|
-
HEADING_RATIO_H1,
|
|
742
|
-
HEADING_RATIO_H2,
|
|
743
|
-
HEADING_RATIO_H3,
|
|
744
749
|
encodePng,
|
|
745
750
|
inlineImagesIntoMarkdown
|
|
746
751
|
};
|
|
747
|
-
//# sourceMappingURL=chunk-
|
|
752
|
+
//# sourceMappingURL=chunk-WF2E5P2G.js.map
|