kreuzberg 4.0.0.rc2 → 4.0.0

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.
Files changed (446) hide show
  1. checksums.yaml +4 -4
  2. data/.gitignore +14 -14
  3. data/.rspec +3 -3
  4. data/.rubocop.yaml +1 -1
  5. data/.rubocop.yml +543 -538
  6. data/Gemfile +8 -8
  7. data/Gemfile.lock +194 -6
  8. data/README.md +396 -426
  9. data/Rakefile +34 -25
  10. data/Steepfile +51 -47
  11. data/examples/async_patterns.rb +283 -341
  12. data/ext/kreuzberg_rb/extconf.rb +65 -45
  13. data/ext/kreuzberg_rb/native/.cargo/config.toml +23 -0
  14. data/ext/kreuzberg_rb/native/Cargo.lock +7619 -6535
  15. data/ext/kreuzberg_rb/native/Cargo.toml +75 -44
  16. data/ext/kreuzberg_rb/native/README.md +425 -425
  17. data/ext/kreuzberg_rb/native/build.rs +15 -15
  18. data/ext/kreuzberg_rb/native/include/ieeefp.h +11 -11
  19. data/ext/kreuzberg_rb/native/include/msvc_compat/strings.h +14 -14
  20. data/ext/kreuzberg_rb/native/include/strings.h +20 -20
  21. data/ext/kreuzberg_rb/native/include/unistd.h +47 -47
  22. data/ext/kreuzberg_rb/native/src/lib.rs +3802 -2998
  23. data/extconf.rb +60 -28
  24. data/kreuzberg.gemspec +199 -148
  25. data/lib/kreuzberg/api_proxy.rb +126 -142
  26. data/lib/kreuzberg/cache_api.rb +67 -46
  27. data/lib/kreuzberg/cli.rb +47 -55
  28. data/lib/kreuzberg/cli_proxy.rb +117 -127
  29. data/lib/kreuzberg/config.rb +936 -691
  30. data/lib/kreuzberg/error_context.rb +136 -32
  31. data/lib/kreuzberg/errors.rb +116 -118
  32. data/lib/kreuzberg/extraction_api.rb +313 -85
  33. data/lib/kreuzberg/mcp_proxy.rb +177 -186
  34. data/lib/kreuzberg/ocr_backend_protocol.rb +40 -113
  35. data/lib/kreuzberg/post_processor_protocol.rb +15 -86
  36. data/lib/kreuzberg/result.rb +334 -216
  37. data/lib/kreuzberg/setup_lib_path.rb +99 -80
  38. data/lib/kreuzberg/types.rb +170 -0
  39. data/lib/kreuzberg/validator_protocol.rb +16 -89
  40. data/lib/kreuzberg/version.rb +5 -5
  41. data/lib/kreuzberg.rb +96 -103
  42. data/lib/libpdfium.so +0 -0
  43. data/sig/kreuzberg/internal.rbs +184 -184
  44. data/sig/kreuzberg.rbs +561 -520
  45. data/spec/binding/async_operations_spec.rb +473 -0
  46. data/spec/binding/batch_operations_spec.rb +595 -0
  47. data/spec/binding/batch_spec.rb +359 -0
  48. data/spec/binding/cache_spec.rb +227 -227
  49. data/spec/binding/cli_proxy_spec.rb +85 -85
  50. data/spec/binding/cli_spec.rb +55 -55
  51. data/spec/binding/config_result_spec.rb +377 -0
  52. data/spec/binding/config_spec.rb +419 -345
  53. data/spec/binding/config_validation_spec.rb +377 -283
  54. data/spec/binding/embeddings_spec.rb +816 -0
  55. data/spec/binding/error_handling_spec.rb +399 -213
  56. data/spec/binding/error_recovery_spec.rb +488 -0
  57. data/spec/binding/errors_spec.rb +66 -66
  58. data/spec/binding/font_config_spec.rb +220 -0
  59. data/spec/binding/images_spec.rb +738 -0
  60. data/spec/binding/keywords_extraction_spec.rb +600 -0
  61. data/spec/binding/metadata_types_spec.rb +1228 -0
  62. data/spec/binding/pages_extraction_spec.rb +471 -0
  63. data/spec/binding/plugins/ocr_backend_spec.rb +307 -307
  64. data/spec/binding/plugins/postprocessor_spec.rb +269 -269
  65. data/spec/binding/plugins/validator_spec.rb +273 -274
  66. data/spec/binding/tables_spec.rb +641 -0
  67. data/spec/fixtures/config.toml +38 -39
  68. data/spec/fixtures/config.yaml +41 -41
  69. data/spec/fixtures/invalid_config.toml +3 -4
  70. data/spec/smoke/package_spec.rb +177 -178
  71. data/spec/spec_helper.rb +40 -42
  72. data/spec/unit/config/chunking_config_spec.rb +213 -0
  73. data/spec/unit/config/embedding_config_spec.rb +343 -0
  74. data/spec/unit/config/extraction_config_spec.rb +438 -0
  75. data/spec/unit/config/font_config_spec.rb +285 -0
  76. data/spec/unit/config/hierarchy_config_spec.rb +314 -0
  77. data/spec/unit/config/image_extraction_config_spec.rb +209 -0
  78. data/spec/unit/config/image_preprocessing_config_spec.rb +249 -0
  79. data/spec/unit/config/keyword_config_spec.rb +229 -0
  80. data/spec/unit/config/language_detection_config_spec.rb +258 -0
  81. data/spec/unit/config/ocr_config_spec.rb +171 -0
  82. data/spec/unit/config/page_config_spec.rb +221 -0
  83. data/spec/unit/config/pdf_config_spec.rb +267 -0
  84. data/spec/unit/config/postprocessor_config_spec.rb +290 -0
  85. data/spec/unit/config/tesseract_config_spec.rb +181 -0
  86. data/spec/unit/config/token_reduction_config_spec.rb +251 -0
  87. data/test/metadata_types_test.rb +959 -0
  88. data/vendor/Cargo.toml +61 -0
  89. data/vendor/kreuzberg/Cargo.toml +259 -204
  90. data/vendor/kreuzberg/README.md +263 -175
  91. data/vendor/kreuzberg/build.rs +782 -474
  92. data/vendor/kreuzberg/examples/bench_fixes.rs +71 -0
  93. data/vendor/kreuzberg/examples/test_pdfium_fork.rs +62 -0
  94. data/vendor/kreuzberg/src/api/error.rs +81 -81
  95. data/vendor/kreuzberg/src/api/handlers.rs +320 -199
  96. data/vendor/kreuzberg/src/api/mod.rs +94 -79
  97. data/vendor/kreuzberg/src/api/server.rs +518 -353
  98. data/vendor/kreuzberg/src/api/types.rs +206 -170
  99. data/vendor/kreuzberg/src/cache/mod.rs +1167 -1167
  100. data/vendor/kreuzberg/src/chunking/mod.rs +2303 -677
  101. data/vendor/kreuzberg/src/chunking/processor.rs +219 -0
  102. data/vendor/kreuzberg/src/core/batch_mode.rs +95 -95
  103. data/vendor/kreuzberg/src/core/batch_optimizations.rs +385 -0
  104. data/vendor/kreuzberg/src/core/config.rs +1914 -1032
  105. data/vendor/kreuzberg/src/core/config_validation.rs +949 -0
  106. data/vendor/kreuzberg/src/core/extractor.rs +1200 -1024
  107. data/vendor/kreuzberg/src/core/formats.rs +235 -0
  108. data/vendor/kreuzberg/src/core/io.rs +329 -329
  109. data/vendor/kreuzberg/src/core/mime.rs +605 -605
  110. data/vendor/kreuzberg/src/core/mod.rs +61 -45
  111. data/vendor/kreuzberg/src/core/pipeline.rs +1223 -984
  112. data/vendor/kreuzberg/src/core/server_config.rs +1220 -0
  113. data/vendor/kreuzberg/src/embeddings.rs +471 -432
  114. data/vendor/kreuzberg/src/error.rs +431 -431
  115. data/vendor/kreuzberg/src/extraction/archive.rs +959 -954
  116. data/vendor/kreuzberg/src/extraction/capacity.rs +263 -0
  117. data/vendor/kreuzberg/src/extraction/docx.rs +404 -40
  118. data/vendor/kreuzberg/src/extraction/email.rs +855 -854
  119. data/vendor/kreuzberg/src/extraction/excel.rs +697 -688
  120. data/vendor/kreuzberg/src/extraction/html.rs +1830 -553
  121. data/vendor/kreuzberg/src/extraction/image.rs +492 -368
  122. data/vendor/kreuzberg/src/extraction/libreoffice.rs +574 -563
  123. data/vendor/kreuzberg/src/extraction/markdown.rs +216 -213
  124. data/vendor/kreuzberg/src/extraction/mod.rs +93 -81
  125. data/vendor/kreuzberg/src/extraction/office_metadata/app_properties.rs +398 -398
  126. data/vendor/kreuzberg/src/extraction/office_metadata/core_properties.rs +247 -247
  127. data/vendor/kreuzberg/src/extraction/office_metadata/custom_properties.rs +240 -240
  128. data/vendor/kreuzberg/src/extraction/office_metadata/mod.rs +130 -130
  129. data/vendor/kreuzberg/src/extraction/office_metadata/odt_properties.rs +284 -287
  130. data/vendor/kreuzberg/src/extraction/pptx.rs +3102 -3000
  131. data/vendor/kreuzberg/src/extraction/structured.rs +491 -490
  132. data/vendor/kreuzberg/src/extraction/table.rs +329 -328
  133. data/vendor/kreuzberg/src/extraction/text.rs +277 -269
  134. data/vendor/kreuzberg/src/extraction/xml.rs +333 -333
  135. data/vendor/kreuzberg/src/extractors/archive.rs +447 -446
  136. data/vendor/kreuzberg/src/extractors/bibtex.rs +470 -469
  137. data/vendor/kreuzberg/src/extractors/docbook.rs +504 -502
  138. data/vendor/kreuzberg/src/extractors/docx.rs +400 -367
  139. data/vendor/kreuzberg/src/extractors/email.rs +157 -143
  140. data/vendor/kreuzberg/src/extractors/epub.rs +696 -707
  141. data/vendor/kreuzberg/src/extractors/excel.rs +385 -343
  142. data/vendor/kreuzberg/src/extractors/fictionbook.rs +492 -491
  143. data/vendor/kreuzberg/src/extractors/html.rs +419 -393
  144. data/vendor/kreuzberg/src/extractors/image.rs +219 -198
  145. data/vendor/kreuzberg/src/extractors/jats.rs +1054 -1051
  146. data/vendor/kreuzberg/src/extractors/jupyter.rs +368 -367
  147. data/vendor/kreuzberg/src/extractors/latex.rs +653 -652
  148. data/vendor/kreuzberg/src/extractors/markdown.rs +701 -700
  149. data/vendor/kreuzberg/src/extractors/mod.rs +429 -365
  150. data/vendor/kreuzberg/src/extractors/odt.rs +628 -628
  151. data/vendor/kreuzberg/src/extractors/opml.rs +635 -634
  152. data/vendor/kreuzberg/src/extractors/orgmode.rs +529 -528
  153. data/vendor/kreuzberg/src/extractors/pdf.rs +761 -493
  154. data/vendor/kreuzberg/src/extractors/pptx.rs +279 -248
  155. data/vendor/kreuzberg/src/extractors/rst.rs +577 -576
  156. data/vendor/kreuzberg/src/extractors/rtf.rs +809 -810
  157. data/vendor/kreuzberg/src/extractors/security.rs +484 -484
  158. data/vendor/kreuzberg/src/extractors/security_tests.rs +367 -367
  159. data/vendor/kreuzberg/src/extractors/structured.rs +142 -140
  160. data/vendor/kreuzberg/src/extractors/text.rs +265 -260
  161. data/vendor/kreuzberg/src/extractors/typst.rs +651 -650
  162. data/vendor/kreuzberg/src/extractors/xml.rs +147 -135
  163. data/vendor/kreuzberg/src/image/dpi.rs +164 -164
  164. data/vendor/kreuzberg/src/image/mod.rs +6 -6
  165. data/vendor/kreuzberg/src/image/preprocessing.rs +417 -417
  166. data/vendor/kreuzberg/src/image/resize.rs +89 -89
  167. data/vendor/kreuzberg/src/keywords/config.rs +154 -154
  168. data/vendor/kreuzberg/src/keywords/mod.rs +237 -237
  169. data/vendor/kreuzberg/src/keywords/processor.rs +275 -267
  170. data/vendor/kreuzberg/src/keywords/rake.rs +293 -293
  171. data/vendor/kreuzberg/src/keywords/types.rs +68 -68
  172. data/vendor/kreuzberg/src/keywords/yake.rs +163 -163
  173. data/vendor/kreuzberg/src/language_detection/mod.rs +985 -942
  174. data/vendor/kreuzberg/src/language_detection/processor.rs +218 -0
  175. data/vendor/kreuzberg/src/lib.rs +114 -105
  176. data/vendor/kreuzberg/src/mcp/mod.rs +35 -32
  177. data/vendor/kreuzberg/src/mcp/server.rs +2090 -1968
  178. data/vendor/kreuzberg/src/ocr/cache.rs +469 -469
  179. data/vendor/kreuzberg/src/ocr/error.rs +37 -37
  180. data/vendor/kreuzberg/src/ocr/hocr.rs +216 -216
  181. data/vendor/kreuzberg/src/ocr/language_registry.rs +520 -0
  182. data/vendor/kreuzberg/src/ocr/mod.rs +60 -58
  183. data/vendor/kreuzberg/src/ocr/processor.rs +858 -863
  184. data/vendor/kreuzberg/src/ocr/table/mod.rs +4 -4
  185. data/vendor/kreuzberg/src/ocr/table/tsv_parser.rs +144 -144
  186. data/vendor/kreuzberg/src/ocr/tesseract_backend.rs +456 -450
  187. data/vendor/kreuzberg/src/ocr/types.rs +393 -393
  188. data/vendor/kreuzberg/src/ocr/utils.rs +47 -47
  189. data/vendor/kreuzberg/src/ocr/validation.rs +206 -206
  190. data/vendor/kreuzberg/src/panic_context.rs +154 -154
  191. data/vendor/kreuzberg/src/pdf/bindings.rs +306 -0
  192. data/vendor/kreuzberg/src/pdf/bundled.rs +408 -0
  193. data/vendor/kreuzberg/src/pdf/error.rs +214 -122
  194. data/vendor/kreuzberg/src/pdf/fonts.rs +358 -0
  195. data/vendor/kreuzberg/src/pdf/hierarchy.rs +903 -0
  196. data/vendor/kreuzberg/src/pdf/images.rs +139 -139
  197. data/vendor/kreuzberg/src/pdf/metadata.rs +509 -346
  198. data/vendor/kreuzberg/src/pdf/mod.rs +81 -50
  199. data/vendor/kreuzberg/src/pdf/rendering.rs +369 -369
  200. data/vendor/kreuzberg/src/pdf/table.rs +417 -393
  201. data/vendor/kreuzberg/src/pdf/text.rs +553 -158
  202. data/vendor/kreuzberg/src/plugins/extractor.rs +1042 -1013
  203. data/vendor/kreuzberg/src/plugins/mod.rs +212 -209
  204. data/vendor/kreuzberg/src/plugins/ocr.rs +637 -620
  205. data/vendor/kreuzberg/src/plugins/processor.rs +650 -642
  206. data/vendor/kreuzberg/src/plugins/registry.rs +1339 -1337
  207. data/vendor/kreuzberg/src/plugins/traits.rs +258 -258
  208. data/vendor/kreuzberg/src/plugins/validator.rs +967 -956
  209. data/vendor/kreuzberg/src/stopwords/mod.rs +1470 -1470
  210. data/vendor/kreuzberg/src/text/mod.rs +27 -19
  211. data/vendor/kreuzberg/src/text/quality.rs +710 -697
  212. data/vendor/kreuzberg/src/text/quality_processor.rs +231 -0
  213. data/vendor/kreuzberg/src/text/string_utils.rs +229 -217
  214. data/vendor/kreuzberg/src/text/token_reduction/cjk_utils.rs +164 -164
  215. data/vendor/kreuzberg/src/text/token_reduction/config.rs +100 -100
  216. data/vendor/kreuzberg/src/text/token_reduction/core.rs +832 -796
  217. data/vendor/kreuzberg/src/text/token_reduction/filters.rs +923 -902
  218. data/vendor/kreuzberg/src/text/token_reduction/mod.rs +160 -160
  219. data/vendor/kreuzberg/src/text/token_reduction/semantic.rs +619 -619
  220. data/vendor/kreuzberg/src/text/token_reduction/simd_text.rs +148 -147
  221. data/vendor/kreuzberg/src/text/utf8_validation.rs +193 -0
  222. data/vendor/kreuzberg/src/types.rs +1713 -903
  223. data/vendor/kreuzberg/src/utils/mod.rs +31 -17
  224. data/vendor/kreuzberg/src/utils/pool.rs +503 -0
  225. data/vendor/kreuzberg/src/utils/pool_sizing.rs +364 -0
  226. data/vendor/kreuzberg/src/utils/quality.rs +968 -959
  227. data/vendor/kreuzberg/src/utils/string_pool.rs +761 -0
  228. data/vendor/kreuzberg/src/utils/string_utils.rs +381 -381
  229. data/vendor/kreuzberg/stopwords/af_stopwords.json +53 -53
  230. data/vendor/kreuzberg/stopwords/ar_stopwords.json +482 -482
  231. data/vendor/kreuzberg/stopwords/bg_stopwords.json +261 -261
  232. data/vendor/kreuzberg/stopwords/bn_stopwords.json +400 -400
  233. data/vendor/kreuzberg/stopwords/br_stopwords.json +1205 -1205
  234. data/vendor/kreuzberg/stopwords/ca_stopwords.json +280 -280
  235. data/vendor/kreuzberg/stopwords/cs_stopwords.json +425 -425
  236. data/vendor/kreuzberg/stopwords/da_stopwords.json +172 -172
  237. data/vendor/kreuzberg/stopwords/de_stopwords.json +622 -622
  238. data/vendor/kreuzberg/stopwords/el_stopwords.json +849 -849
  239. data/vendor/kreuzberg/stopwords/en_stopwords.json +1300 -1300
  240. data/vendor/kreuzberg/stopwords/eo_stopwords.json +175 -175
  241. data/vendor/kreuzberg/stopwords/es_stopwords.json +734 -734
  242. data/vendor/kreuzberg/stopwords/et_stopwords.json +37 -37
  243. data/vendor/kreuzberg/stopwords/eu_stopwords.json +100 -100
  244. data/vendor/kreuzberg/stopwords/fa_stopwords.json +801 -801
  245. data/vendor/kreuzberg/stopwords/fi_stopwords.json +849 -849
  246. data/vendor/kreuzberg/stopwords/fr_stopwords.json +693 -693
  247. data/vendor/kreuzberg/stopwords/ga_stopwords.json +111 -111
  248. data/vendor/kreuzberg/stopwords/gl_stopwords.json +162 -162
  249. data/vendor/kreuzberg/stopwords/gu_stopwords.json +226 -226
  250. data/vendor/kreuzberg/stopwords/ha_stopwords.json +41 -41
  251. data/vendor/kreuzberg/stopwords/he_stopwords.json +196 -196
  252. data/vendor/kreuzberg/stopwords/hi_stopwords.json +227 -227
  253. data/vendor/kreuzberg/stopwords/hr_stopwords.json +181 -181
  254. data/vendor/kreuzberg/stopwords/hu_stopwords.json +791 -791
  255. data/vendor/kreuzberg/stopwords/hy_stopwords.json +47 -47
  256. data/vendor/kreuzberg/stopwords/id_stopwords.json +760 -760
  257. data/vendor/kreuzberg/stopwords/it_stopwords.json +634 -634
  258. data/vendor/kreuzberg/stopwords/ja_stopwords.json +136 -136
  259. data/vendor/kreuzberg/stopwords/kn_stopwords.json +84 -84
  260. data/vendor/kreuzberg/stopwords/ko_stopwords.json +681 -681
  261. data/vendor/kreuzberg/stopwords/ku_stopwords.json +64 -64
  262. data/vendor/kreuzberg/stopwords/la_stopwords.json +51 -51
  263. data/vendor/kreuzberg/stopwords/lt_stopwords.json +476 -476
  264. data/vendor/kreuzberg/stopwords/lv_stopwords.json +163 -163
  265. data/vendor/kreuzberg/stopwords/ml_stopwords.json +1 -1
  266. data/vendor/kreuzberg/stopwords/mr_stopwords.json +101 -101
  267. data/vendor/kreuzberg/stopwords/ms_stopwords.json +477 -477
  268. data/vendor/kreuzberg/stopwords/ne_stopwords.json +490 -490
  269. data/vendor/kreuzberg/stopwords/nl_stopwords.json +415 -415
  270. data/vendor/kreuzberg/stopwords/no_stopwords.json +223 -223
  271. data/vendor/kreuzberg/stopwords/pl_stopwords.json +331 -331
  272. data/vendor/kreuzberg/stopwords/pt_stopwords.json +562 -562
  273. data/vendor/kreuzberg/stopwords/ro_stopwords.json +436 -436
  274. data/vendor/kreuzberg/stopwords/ru_stopwords.json +561 -561
  275. data/vendor/kreuzberg/stopwords/si_stopwords.json +193 -193
  276. data/vendor/kreuzberg/stopwords/sk_stopwords.json +420 -420
  277. data/vendor/kreuzberg/stopwords/sl_stopwords.json +448 -448
  278. data/vendor/kreuzberg/stopwords/so_stopwords.json +32 -32
  279. data/vendor/kreuzberg/stopwords/st_stopwords.json +33 -33
  280. data/vendor/kreuzberg/stopwords/sv_stopwords.json +420 -420
  281. data/vendor/kreuzberg/stopwords/sw_stopwords.json +76 -76
  282. data/vendor/kreuzberg/stopwords/ta_stopwords.json +129 -129
  283. data/vendor/kreuzberg/stopwords/te_stopwords.json +54 -54
  284. data/vendor/kreuzberg/stopwords/th_stopwords.json +118 -118
  285. data/vendor/kreuzberg/stopwords/tl_stopwords.json +149 -149
  286. data/vendor/kreuzberg/stopwords/tr_stopwords.json +506 -506
  287. data/vendor/kreuzberg/stopwords/uk_stopwords.json +75 -75
  288. data/vendor/kreuzberg/stopwords/ur_stopwords.json +519 -519
  289. data/vendor/kreuzberg/stopwords/vi_stopwords.json +647 -647
  290. data/vendor/kreuzberg/stopwords/yo_stopwords.json +62 -62
  291. data/vendor/kreuzberg/stopwords/zh_stopwords.json +796 -796
  292. data/vendor/kreuzberg/stopwords/zu_stopwords.json +31 -31
  293. data/vendor/kreuzberg/tests/api_embed.rs +360 -0
  294. data/vendor/kreuzberg/tests/api_extract_multipart.rs +52 -52
  295. data/vendor/kreuzberg/tests/api_large_pdf_extraction.rs +471 -0
  296. data/vendor/kreuzberg/tests/api_large_pdf_extraction_diagnostics.rs +289 -0
  297. data/vendor/kreuzberg/tests/api_tests.rs +1472 -966
  298. data/vendor/kreuzberg/tests/archive_integration.rs +545 -543
  299. data/vendor/kreuzberg/tests/batch_orchestration.rs +587 -556
  300. data/vendor/kreuzberg/tests/batch_pooling_benchmark.rs +154 -0
  301. data/vendor/kreuzberg/tests/batch_processing.rs +328 -316
  302. data/vendor/kreuzberg/tests/bibtex_parity_test.rs +421 -421
  303. data/vendor/kreuzberg/tests/concurrency_stress.rs +541 -525
  304. data/vendor/kreuzberg/tests/config_features.rs +612 -598
  305. data/vendor/kreuzberg/tests/config_integration_test.rs +753 -0
  306. data/vendor/kreuzberg/tests/config_loading_tests.rs +416 -415
  307. data/vendor/kreuzberg/tests/core_integration.rs +519 -510
  308. data/vendor/kreuzberg/tests/csv_integration.rs +414 -414
  309. data/vendor/kreuzberg/tests/data/hierarchy_ground_truth.json +294 -0
  310. data/vendor/kreuzberg/tests/docbook_extractor_tests.rs +500 -498
  311. data/vendor/kreuzberg/tests/docx_metadata_extraction_test.rs +122 -122
  312. data/vendor/kreuzberg/tests/docx_vs_pandoc_comparison.rs +370 -370
  313. data/vendor/kreuzberg/tests/email_integration.rs +327 -325
  314. data/vendor/kreuzberg/tests/epub_native_extractor_tests.rs +275 -275
  315. data/vendor/kreuzberg/tests/error_handling.rs +402 -393
  316. data/vendor/kreuzberg/tests/fictionbook_extractor_tests.rs +228 -228
  317. data/vendor/kreuzberg/tests/format_integration.rs +165 -159
  318. data/vendor/kreuzberg/tests/helpers/mod.rs +202 -142
  319. data/vendor/kreuzberg/tests/html_table_test.rs +551 -551
  320. data/vendor/kreuzberg/tests/image_integration.rs +255 -253
  321. data/vendor/kreuzberg/tests/instrumentation_test.rs +139 -139
  322. data/vendor/kreuzberg/tests/jats_extractor_tests.rs +639 -639
  323. data/vendor/kreuzberg/tests/jupyter_extractor_tests.rs +704 -704
  324. data/vendor/kreuzberg/tests/keywords_integration.rs +479 -479
  325. data/vendor/kreuzberg/tests/keywords_quality.rs +509 -509
  326. data/vendor/kreuzberg/tests/latex_extractor_tests.rs +496 -496
  327. data/vendor/kreuzberg/tests/markdown_extractor_tests.rs +490 -490
  328. data/vendor/kreuzberg/tests/mime_detection.rs +429 -428
  329. data/vendor/kreuzberg/tests/ocr_configuration.rs +514 -510
  330. data/vendor/kreuzberg/tests/ocr_errors.rs +698 -676
  331. data/vendor/kreuzberg/tests/ocr_language_registry.rs +191 -0
  332. data/vendor/kreuzberg/tests/ocr_quality.rs +629 -627
  333. data/vendor/kreuzberg/tests/ocr_stress.rs +469 -469
  334. data/vendor/kreuzberg/tests/odt_extractor_tests.rs +674 -695
  335. data/vendor/kreuzberg/tests/opml_extractor_tests.rs +616 -616
  336. data/vendor/kreuzberg/tests/orgmode_extractor_tests.rs +822 -822
  337. data/vendor/kreuzberg/tests/page_markers.rs +297 -0
  338. data/vendor/kreuzberg/tests/pdf_hierarchy_detection.rs +301 -0
  339. data/vendor/kreuzberg/tests/pdf_hierarchy_quality.rs +589 -0
  340. data/vendor/kreuzberg/tests/pdf_integration.rs +45 -43
  341. data/vendor/kreuzberg/tests/pdf_ocr_triggering.rs +301 -0
  342. data/vendor/kreuzberg/tests/pdf_text_merging.rs +475 -0
  343. data/vendor/kreuzberg/tests/pdfium_linking.rs +340 -0
  344. data/vendor/kreuzberg/tests/pipeline_integration.rs +1446 -1411
  345. data/vendor/kreuzberg/tests/plugin_ocr_backend_test.rs +776 -771
  346. data/vendor/kreuzberg/tests/plugin_postprocessor_test.rs +577 -560
  347. data/vendor/kreuzberg/tests/plugin_system.rs +927 -921
  348. data/vendor/kreuzberg/tests/plugin_validator_test.rs +783 -783
  349. data/vendor/kreuzberg/tests/registry_integration_tests.rs +587 -586
  350. data/vendor/kreuzberg/tests/rst_extractor_tests.rs +694 -692
  351. data/vendor/kreuzberg/tests/rtf_extractor_tests.rs +775 -776
  352. data/vendor/kreuzberg/tests/security_validation.rs +416 -415
  353. data/vendor/kreuzberg/tests/stopwords_integration_test.rs +888 -888
  354. data/vendor/kreuzberg/tests/test_fastembed.rs +631 -609
  355. data/vendor/kreuzberg/tests/typst_behavioral_tests.rs +1260 -1259
  356. data/vendor/kreuzberg/tests/typst_extractor_tests.rs +648 -647
  357. data/vendor/kreuzberg/tests/xlsx_metadata_extraction_test.rs +87 -87
  358. data/vendor/kreuzberg-ffi/Cargo.toml +67 -0
  359. data/vendor/kreuzberg-ffi/README.md +851 -0
  360. data/vendor/kreuzberg-ffi/benches/result_view_benchmark.rs +227 -0
  361. data/vendor/kreuzberg-ffi/build.rs +168 -0
  362. data/vendor/kreuzberg-ffi/cbindgen.toml +37 -0
  363. data/vendor/kreuzberg-ffi/kreuzberg-ffi.pc.in +12 -0
  364. data/vendor/kreuzberg-ffi/kreuzberg.h +3012 -0
  365. data/vendor/kreuzberg-ffi/src/batch_streaming.rs +588 -0
  366. data/vendor/kreuzberg-ffi/src/config.rs +1341 -0
  367. data/vendor/kreuzberg-ffi/src/error.rs +901 -0
  368. data/vendor/kreuzberg-ffi/src/extraction.rs +555 -0
  369. data/vendor/kreuzberg-ffi/src/helpers.rs +879 -0
  370. data/vendor/kreuzberg-ffi/src/lib.rs +977 -0
  371. data/vendor/kreuzberg-ffi/src/memory.rs +493 -0
  372. data/vendor/kreuzberg-ffi/src/mime.rs +329 -0
  373. data/vendor/kreuzberg-ffi/src/panic_shield.rs +265 -0
  374. data/vendor/kreuzberg-ffi/src/plugins/document_extractor.rs +442 -0
  375. data/vendor/kreuzberg-ffi/src/plugins/mod.rs +14 -0
  376. data/vendor/kreuzberg-ffi/src/plugins/ocr_backend.rs +628 -0
  377. data/vendor/kreuzberg-ffi/src/plugins/post_processor.rs +438 -0
  378. data/vendor/kreuzberg-ffi/src/plugins/validator.rs +329 -0
  379. data/vendor/kreuzberg-ffi/src/result.rs +510 -0
  380. data/vendor/kreuzberg-ffi/src/result_pool.rs +639 -0
  381. data/vendor/kreuzberg-ffi/src/result_view.rs +773 -0
  382. data/vendor/kreuzberg-ffi/src/string_intern.rs +568 -0
  383. data/vendor/kreuzberg-ffi/src/types.rs +363 -0
  384. data/vendor/kreuzberg-ffi/src/util.rs +210 -0
  385. data/vendor/kreuzberg-ffi/src/validation.rs +848 -0
  386. data/vendor/kreuzberg-ffi/tests.disabled/README.md +48 -0
  387. data/vendor/kreuzberg-ffi/tests.disabled/config_loading_tests.rs +299 -0
  388. data/vendor/kreuzberg-ffi/tests.disabled/config_tests.rs +346 -0
  389. data/vendor/kreuzberg-ffi/tests.disabled/extractor_tests.rs +232 -0
  390. data/vendor/kreuzberg-ffi/tests.disabled/plugin_registration_tests.rs +470 -0
  391. data/vendor/kreuzberg-tesseract/.commitlintrc.json +13 -0
  392. data/vendor/kreuzberg-tesseract/.crate-ignore +2 -0
  393. data/vendor/kreuzberg-tesseract/Cargo.lock +2933 -0
  394. data/vendor/kreuzberg-tesseract/Cargo.toml +57 -0
  395. data/vendor/{rb-sys/LICENSE-MIT → kreuzberg-tesseract/LICENSE} +22 -21
  396. data/vendor/kreuzberg-tesseract/README.md +399 -0
  397. data/vendor/kreuzberg-tesseract/build.rs +1127 -0
  398. data/vendor/kreuzberg-tesseract/patches/README.md +71 -0
  399. data/vendor/kreuzberg-tesseract/patches/tesseract.diff +199 -0
  400. data/vendor/kreuzberg-tesseract/src/api.rs +1371 -0
  401. data/vendor/kreuzberg-tesseract/src/choice_iterator.rs +77 -0
  402. data/vendor/kreuzberg-tesseract/src/enums.rs +297 -0
  403. data/vendor/kreuzberg-tesseract/src/error.rs +81 -0
  404. data/vendor/kreuzberg-tesseract/src/lib.rs +145 -0
  405. data/vendor/kreuzberg-tesseract/src/monitor.rs +57 -0
  406. data/vendor/kreuzberg-tesseract/src/mutable_iterator.rs +197 -0
  407. data/vendor/kreuzberg-tesseract/src/page_iterator.rs +253 -0
  408. data/vendor/kreuzberg-tesseract/src/result_iterator.rs +286 -0
  409. data/vendor/kreuzberg-tesseract/src/result_renderer.rs +183 -0
  410. data/vendor/kreuzberg-tesseract/tests/integration_test.rs +211 -0
  411. metadata +196 -45
  412. data/vendor/kreuzberg/benches/otel_overhead.rs +0 -48
  413. data/vendor/kreuzberg/src/extractors/fictionbook.rs.backup2 +0 -738
  414. data/vendor/rb-sys/.cargo-ok +0 -1
  415. data/vendor/rb-sys/.cargo_vcs_info.json +0 -6
  416. data/vendor/rb-sys/Cargo.lock +0 -393
  417. data/vendor/rb-sys/Cargo.toml +0 -70
  418. data/vendor/rb-sys/Cargo.toml.orig +0 -57
  419. data/vendor/rb-sys/LICENSE-APACHE +0 -190
  420. data/vendor/rb-sys/bin/release.sh +0 -21
  421. data/vendor/rb-sys/build/features.rs +0 -108
  422. data/vendor/rb-sys/build/main.rs +0 -246
  423. data/vendor/rb-sys/build/stable_api_config.rs +0 -153
  424. data/vendor/rb-sys/build/version.rs +0 -48
  425. data/vendor/rb-sys/readme.md +0 -36
  426. data/vendor/rb-sys/src/bindings.rs +0 -21
  427. data/vendor/rb-sys/src/hidden.rs +0 -11
  428. data/vendor/rb-sys/src/lib.rs +0 -34
  429. data/vendor/rb-sys/src/macros.rs +0 -371
  430. data/vendor/rb-sys/src/memory.rs +0 -53
  431. data/vendor/rb-sys/src/ruby_abi_version.rs +0 -38
  432. data/vendor/rb-sys/src/special_consts.rs +0 -31
  433. data/vendor/rb-sys/src/stable_api/compiled.c +0 -179
  434. data/vendor/rb-sys/src/stable_api/compiled.rs +0 -257
  435. data/vendor/rb-sys/src/stable_api/ruby_2_6.rs +0 -316
  436. data/vendor/rb-sys/src/stable_api/ruby_2_7.rs +0 -316
  437. data/vendor/rb-sys/src/stable_api/ruby_3_0.rs +0 -324
  438. data/vendor/rb-sys/src/stable_api/ruby_3_1.rs +0 -317
  439. data/vendor/rb-sys/src/stable_api/ruby_3_2.rs +0 -315
  440. data/vendor/rb-sys/src/stable_api/ruby_3_3.rs +0 -326
  441. data/vendor/rb-sys/src/stable_api/ruby_3_4.rs +0 -327
  442. data/vendor/rb-sys/src/stable_api.rs +0 -261
  443. data/vendor/rb-sys/src/symbol.rs +0 -31
  444. data/vendor/rb-sys/src/tracking_allocator.rs +0 -332
  445. data/vendor/rb-sys/src/utils.rs +0 -89
  446. data/vendor/rb-sys/src/value_type.rs +0 -7
@@ -0,0 +1,1127 @@
1
+ #![allow(clippy::uninlined_format_args)]
2
+
3
+ #[cfg(feature = "build-tesseract")]
4
+ mod build_tesseract {
5
+ use cmake::Config;
6
+ use std::env;
7
+ use std::fs;
8
+ use std::path::{Path, PathBuf};
9
+
10
+ const LEPTONICA_VERSION: &str = "1.86.0";
11
+ const TESSERACT_VERSION: &str = "5.5.1";
12
+ #[allow(dead_code)]
13
+ const EMSDK_COMMIT: &str = "974d5c096bd56e42045d052a47b28198ce75e2a8";
14
+ #[allow(dead_code)]
15
+ const EMSDK_VERSION: &str = "3.1.31";
16
+ #[allow(dead_code)]
17
+ const EMSDK_REPOSITORY: &str = "https://github.com/emscripten-core/emsdk.git";
18
+
19
+ fn leptonica_url() -> String {
20
+ format!(
21
+ "https://codeload.github.com/DanBloomberg/leptonica/zip/refs/tags/{}",
22
+ LEPTONICA_VERSION
23
+ )
24
+ }
25
+
26
+ fn tesseract_url() -> String {
27
+ format!(
28
+ "https://codeload.github.com/tesseract-ocr/tesseract/zip/refs/tags/{}",
29
+ TESSERACT_VERSION
30
+ )
31
+ }
32
+
33
+ fn workspace_cache_dir_from_out_dir() -> Option<PathBuf> {
34
+ let out_dir = env::var_os("OUT_DIR")?;
35
+ let mut path = PathBuf::from(out_dir);
36
+ for _ in 0..4 {
37
+ if !path.pop() {
38
+ return None;
39
+ }
40
+ }
41
+ Some(path.join("tesseract-rs-cache"))
42
+ }
43
+
44
+ fn get_preferred_out_dir() -> PathBuf {
45
+ if let Ok(custom) = env::var("TESSERACT_RS_CACHE_DIR") {
46
+ return PathBuf::from(custom);
47
+ }
48
+
49
+ if cfg!(target_os = "windows") {
50
+ return PathBuf::from("C:\\tess");
51
+ }
52
+
53
+ if let Some(workspace_cache) = workspace_cache_dir_from_out_dir() {
54
+ return workspace_cache;
55
+ }
56
+
57
+ if cfg!(target_os = "macos") {
58
+ let home_dir = env::var("HOME").unwrap_or_else(|_| {
59
+ env::var("USER")
60
+ .map(|user| format!("/Users/{}", user))
61
+ .expect("Neither HOME nor USER environment variable set")
62
+ });
63
+ PathBuf::from(home_dir)
64
+ .join("Library")
65
+ .join("Application Support")
66
+ .join("tesseract-rs")
67
+ } else if cfg!(target_os = "linux") {
68
+ let home_dir = env::var("HOME").unwrap_or_else(|_| {
69
+ env::var("USER")
70
+ .map(|user| format!("/home/{}", user))
71
+ .expect("Neither HOME nor USER environment variable set")
72
+ });
73
+ PathBuf::from(home_dir).join(".tesseract-rs")
74
+ } else {
75
+ panic!("Unsupported operating system");
76
+ }
77
+ }
78
+
79
+ fn target_triple() -> String {
80
+ env::var("TARGET").unwrap_or_else(|_| env::var("HOST").unwrap_or_default())
81
+ }
82
+
83
+ fn target_matches(target: &str, needle: &str) -> bool {
84
+ target.contains(needle)
85
+ }
86
+
87
+ fn is_windows_target(target: &str) -> bool {
88
+ target_matches(target, "windows")
89
+ }
90
+
91
+ fn is_macos_target(target: &str) -> bool {
92
+ target_matches(target, "apple-darwin")
93
+ }
94
+
95
+ fn is_linux_target(target: &str) -> bool {
96
+ target_matches(target, "linux")
97
+ }
98
+
99
+ fn is_msvc_target(target: &str) -> bool {
100
+ is_windows_target(target) && target_matches(target, "msvc")
101
+ }
102
+
103
+ fn is_mingw_target(target: &str) -> bool {
104
+ is_windows_target(target) && target_matches(target, "gnu")
105
+ }
106
+
107
+ fn is_wasm_target(target: &str) -> bool {
108
+ target_matches(target, "wasm32") || target_matches(target, "wasm64")
109
+ }
110
+
111
+ fn prepare_out_dir() -> PathBuf {
112
+ let preferred = get_preferred_out_dir();
113
+ match fs::create_dir_all(&preferred) {
114
+ Ok(_) => preferred,
115
+ Err(err) => {
116
+ println!(
117
+ "cargo:warning=Failed to create cache dir {:?}: {}. Falling back to temp dir.",
118
+ preferred, err
119
+ );
120
+ let fallback = env::temp_dir().join("tesseract-rs-cache");
121
+ fs::create_dir_all(&fallback).expect("Failed to create fallback cache directory in temp dir");
122
+ fallback
123
+ }
124
+ }
125
+ }
126
+
127
+ #[allow(dead_code)]
128
+ fn find_emsdk() -> Result<PathBuf, String> {
129
+ if let Ok(emsdk_path) = env::var("EMSDK") {
130
+ let path = PathBuf::from(emsdk_path);
131
+ if path.exists() {
132
+ return Ok(path);
133
+ }
134
+ }
135
+
136
+ let common_paths = vec![
137
+ PathBuf::from("/opt/homebrew/opt/emscripten"),
138
+ PathBuf::from("/usr/local/opt/emscripten"),
139
+ PathBuf::from(env::var("HOME").unwrap_or_default()).join(".emsdk"),
140
+ PathBuf::from("/usr/lib/emscripten"),
141
+ PathBuf::from("/opt/emsdk"),
142
+ ];
143
+
144
+ for path in common_paths {
145
+ if path.exists() {
146
+ return Ok(path);
147
+ }
148
+ }
149
+
150
+ Err("EMSDK not found. Please install via: brew install emscripten (macOS) or apt-get install emscripten (Linux)".to_string())
151
+ }
152
+
153
+ pub fn build() {
154
+ let target = target_triple();
155
+
156
+ if is_wasm_target(&target) {
157
+ println!(
158
+ "cargo:warning=Detected WASM target: {}, routing to build_wasm()",
159
+ target
160
+ );
161
+ return build_wasm();
162
+ }
163
+
164
+ let custom_out_dir = prepare_out_dir();
165
+ let windows_target = is_windows_target(&target);
166
+ let msvc_target = is_msvc_target(&target);
167
+ let mingw_target = is_mingw_target(&target);
168
+
169
+ println!("cargo:warning=custom_out_dir: {:?}", custom_out_dir);
170
+
171
+ let cache_dir = custom_out_dir.join("cache");
172
+
173
+ if env::var("CARGO_CLEAN").is_ok() {
174
+ clean_cache(&cache_dir);
175
+ }
176
+
177
+ std::fs::create_dir_all(&cache_dir).expect("Failed to create cache directory");
178
+
179
+ let out_dir = custom_out_dir.clone();
180
+ let project_dir = custom_out_dir.clone();
181
+ let third_party_dir = project_dir.join("third_party");
182
+
183
+ let leptonica_dir = if third_party_dir.join("leptonica").exists() {
184
+ println!("cargo:warning=Using existing leptonica source");
185
+ third_party_dir.join("leptonica")
186
+ } else {
187
+ fs::create_dir_all(&third_party_dir).expect("Failed to create third_party directory");
188
+ download_and_extract(&third_party_dir, &leptonica_url(), "leptonica")
189
+ };
190
+
191
+ let tesseract_dir = if third_party_dir.join("tesseract").exists() {
192
+ println!("cargo:warning=Using existing tesseract source");
193
+ third_party_dir.join("tesseract")
194
+ } else {
195
+ fs::create_dir_all(&third_party_dir).expect("Failed to create third_party directory");
196
+ download_and_extract(&third_party_dir, &tesseract_url(), "tesseract")
197
+ };
198
+
199
+ let (cmake_cxx_flags, additional_defines) = get_os_specific_config();
200
+
201
+ let leptonica_install_dir = out_dir.join("leptonica");
202
+ let leptonica_cache_dir = cache_dir.join("leptonica");
203
+
204
+ let leptonica_link_name = build_or_use_cached(
205
+ "leptonica",
206
+ &leptonica_cache_dir,
207
+ &leptonica_install_dir,
208
+ || {
209
+ let mut leptonica_config = Config::new(&leptonica_dir);
210
+
211
+ let leptonica_src_dir = leptonica_dir.join("src");
212
+ let environ_h_path = leptonica_src_dir.join("environ.h");
213
+
214
+ if environ_h_path.exists() {
215
+ let environ_h = std::fs::read_to_string(&environ_h_path)
216
+ .expect("Failed to read environ.h")
217
+ .replace("#define HAVE_LIBZ 1", "#define HAVE_LIBZ 0")
218
+ .replace("#ifdef NO_CONSOLE_IO", "#define NO_CONSOLE_IO\n#ifdef NO_CONSOLE_IO");
219
+ std::fs::write(environ_h_path, environ_h).expect("Failed to write environ.h");
220
+ }
221
+
222
+ let makefile_static_path = leptonica_dir.join("prog").join("makefile.static");
223
+
224
+ let leptonica_src_cmakelists = leptonica_dir.join("src").join("CMakeLists.txt");
225
+
226
+ if leptonica_src_cmakelists.exists() {
227
+ let cmakelists = std::fs::read_to_string(&leptonica_src_cmakelists)
228
+ .expect("Failed to read leptonica src CMakeLists.txt");
229
+ let patched = cmakelists.replace(
230
+ "if(MINGW)\n set_target_properties(\n leptonica PROPERTIES SUFFIX\n \"-${PROJECT_VERSION}${CMAKE_SHARED_LIBRARY_SUFFIX}\")\nendif(MINGW)\n",
231
+ "if(MINGW AND BUILD_SHARED_LIBS)\n set_target_properties(\n leptonica PROPERTIES SUFFIX\n \"-${PROJECT_VERSION}${CMAKE_SHARED_LIBRARY_SUFFIX}\")\nendif()\n",
232
+ );
233
+ if patched != cmakelists {
234
+ std::fs::write(&leptonica_src_cmakelists, patched)
235
+ .expect("Failed to patch leptonica src CMakeLists.txt");
236
+ }
237
+ }
238
+
239
+ if makefile_static_path.exists() {
240
+ let makefile_static = std::fs::read_to_string(&makefile_static_path)
241
+ .expect("Failed to read makefile.static")
242
+ .replace(
243
+ "ALL_LIBS = $(LEPTLIB) -ltiff -ljpeg -lpng -lz -lm",
244
+ "ALL_LIBS = $(LEPTLIB) -lm",
245
+ );
246
+ std::fs::write(makefile_static_path, makefile_static).expect("Failed to write makefile.static");
247
+ }
248
+
249
+ if windows_target {
250
+ if mingw_target {
251
+ leptonica_config.generator("Unix Makefiles");
252
+ leptonica_config.define("CMAKE_MAKE_PROGRAM", "mingw32-make");
253
+ leptonica_config.define("MSYS2_ARG_CONV_EXCL", "/MD;/MDd;/D;-D;-I;-L");
254
+ } else if msvc_target && env::var("VSINSTALLDIR").is_ok() {
255
+ leptonica_config.generator("NMake Makefiles");
256
+ }
257
+ leptonica_config.define("CMAKE_CL_SHOWINCLUDES_PREFIX", "");
258
+ }
259
+
260
+ if env::var("CI").is_err() && env::var("RUSTC_WRAPPER").unwrap_or_default() == "sccache" {
261
+ leptonica_config.env("CC", "sccache cc").env("CXX", "sccache c++");
262
+ }
263
+
264
+ let leptonica_install_dir_cmake = normalize_cmake_path(&leptonica_install_dir);
265
+
266
+ leptonica_config
267
+ .define("CMAKE_POLICY_VERSION_MINIMUM", "3.5")
268
+ .define("CMAKE_BUILD_TYPE", "Release")
269
+ .define("BUILD_PROG", "OFF")
270
+ .define("BUILD_SHARED_LIBS", "OFF")
271
+ .define("ENABLE_ZLIB", "OFF")
272
+ .define("ENABLE_PNG", "OFF")
273
+ .define("ENABLE_JPEG", "OFF")
274
+ .define("ENABLE_TIFF", "OFF")
275
+ .define("ENABLE_WEBP", "OFF")
276
+ .define("ENABLE_OPENJPEG", "OFF")
277
+ .define("ENABLE_GIF", "OFF")
278
+ .define("NO_CONSOLE_IO", "ON")
279
+ .define("CMAKE_CXX_FLAGS", &cmake_cxx_flags)
280
+ .define("MINIMUM_SEVERITY", "L_SEVERITY_NONE")
281
+ .define("SW_BUILD", "OFF")
282
+ .define("HAVE_LIBZ", "0")
283
+ .define("ENABLE_LTO", "OFF")
284
+ .define("CMAKE_INSTALL_PREFIX", &leptonica_install_dir_cmake);
285
+
286
+ if windows_target {
287
+ if msvc_target {
288
+ leptonica_config
289
+ .define("CMAKE_C_FLAGS_RELEASE", "/MD /O2")
290
+ .define("CMAKE_C_FLAGS_DEBUG", "/MDd /Od");
291
+ } else if mingw_target {
292
+ leptonica_config
293
+ .define("CMAKE_C_FLAGS_RELEASE", "-O2 -DNDEBUG")
294
+ .define("CMAKE_C_FLAGS_DEBUG", "-O0 -g");
295
+ } else {
296
+ leptonica_config
297
+ .define("CMAKE_C_FLAGS_RELEASE", "-O2")
298
+ .define("CMAKE_C_FLAGS_DEBUG", "-O0 -g");
299
+ }
300
+ }
301
+
302
+ for (key, value) in &additional_defines {
303
+ leptonica_config.define(key, value);
304
+ }
305
+
306
+ leptonica_config.build();
307
+ },
308
+ );
309
+
310
+ let leptonica_include_dir = leptonica_install_dir.join("include");
311
+ let leptonica_lib_dir = leptonica_install_dir.join("lib");
312
+ let tesseract_install_dir = out_dir.join("tesseract");
313
+ let tesseract_cache_dir = cache_dir.join("tesseract");
314
+ let tessdata_prefix = project_dir.clone();
315
+
316
+ let leptonica_install_dir_cmake = normalize_cmake_path(&leptonica_install_dir);
317
+ let leptonica_include_dir_cmake = normalize_cmake_path(&leptonica_include_dir);
318
+ let leptonica_lib_dir_cmake = normalize_cmake_path(&leptonica_lib_dir);
319
+ let tesseract_install_dir_cmake = normalize_cmake_path(&tesseract_install_dir);
320
+ let tessdata_prefix_cmake = normalize_cmake_path(&tessdata_prefix);
321
+
322
+ let tesseract_link_name =
323
+ build_or_use_cached("tesseract", &tesseract_cache_dir, &tesseract_install_dir, || {
324
+ let cmakelists_path = tesseract_dir.join("CMakeLists.txt");
325
+ let cmakelists = std::fs::read_to_string(&cmakelists_path)
326
+ .expect("Failed to read CMakeLists.txt")
327
+ .replace("set(HAVE_TIFFIO_H ON)", "");
328
+ std::fs::write(&cmakelists_path, cmakelists).expect("Failed to write CMakeLists.txt");
329
+
330
+ let mut tesseract_config = Config::new(&tesseract_dir);
331
+ if windows_target {
332
+ if mingw_target {
333
+ tesseract_config.generator("Unix Makefiles");
334
+ tesseract_config.define("CMAKE_MAKE_PROGRAM", "mingw32-make");
335
+ tesseract_config.define("MSYS2_ARG_CONV_EXCL", "/MD;/MDd;/D;-D;-I;-L");
336
+ } else if msvc_target && env::var("VSINSTALLDIR").is_ok() {
337
+ tesseract_config.generator("NMake Makefiles");
338
+ }
339
+ tesseract_config.define("CMAKE_CL_SHOWINCLUDES_PREFIX", "");
340
+ }
341
+
342
+ if env::var("CI").is_err() && env::var("RUSTC_WRAPPER").unwrap_or_default() == "sccache" {
343
+ tesseract_config.env("CC", "sccache cc").env("CXX", "sccache c++");
344
+ }
345
+ tesseract_config
346
+ .define("CMAKE_POLICY_VERSION_MINIMUM", "3.5")
347
+ .define("CMAKE_BUILD_TYPE", "Release")
348
+ .define("BUILD_TRAINING_TOOLS", "OFF")
349
+ .define("BUILD_SHARED_LIBS", "OFF")
350
+ .define("DISABLE_ARCHIVE", "ON")
351
+ .define("DISABLE_CURL", "ON")
352
+ .define("DISABLE_OPENCL", "ON")
353
+ .define("Leptonica_DIR", &leptonica_install_dir_cmake)
354
+ .define("LEPTONICA_INCLUDE_DIR", &leptonica_include_dir_cmake)
355
+ .define("LEPTONICA_LIBRARY", &leptonica_lib_dir_cmake)
356
+ .define("CMAKE_PREFIX_PATH", &leptonica_install_dir_cmake)
357
+ .define("CMAKE_INSTALL_PREFIX", &tesseract_install_dir_cmake)
358
+ .define("TESSDATA_PREFIX", &tessdata_prefix_cmake)
359
+ .define("DISABLE_TIFF", "ON")
360
+ .define("DISABLE_PNG", "ON")
361
+ .define("DISABLE_JPEG", "ON")
362
+ .define("DISABLE_WEBP", "ON")
363
+ .define("DISABLE_OPENJPEG", "ON")
364
+ .define("DISABLE_ZLIB", "ON")
365
+ .define("DISABLE_LIBXML2", "ON")
366
+ .define("DISABLE_LIBICU", "ON")
367
+ .define("DISABLE_LZMA", "ON")
368
+ .define("DISABLE_GIF", "ON")
369
+ .define("DISABLE_DEBUG_MESSAGES", "ON")
370
+ .define("debug_file", "/dev/null")
371
+ .define("HAVE_LIBARCHIVE", "OFF")
372
+ .define("HAVE_LIBCURL", "OFF")
373
+ .define("HAVE_TIFFIO_H", "OFF")
374
+ .define("GRAPHICS_DISABLED", "ON")
375
+ .define("DISABLED_LEGACY_ENGINE", "OFF")
376
+ .define("USE_OPENCL", "OFF")
377
+ .define("OPENMP_BUILD", "OFF")
378
+ .define("BUILD_TESTS", "OFF")
379
+ .define("ENABLE_LTO", "OFF")
380
+ .define("BUILD_PROG", "OFF")
381
+ .define("SW_BUILD", "OFF")
382
+ .define("LEPT_TIFF_RESULT", "FALSE")
383
+ .define("INSTALL_CONFIGS", "ON")
384
+ .define("USE_SYSTEM_ICU", "ON")
385
+ .define("CMAKE_CXX_FLAGS", &cmake_cxx_flags);
386
+
387
+ for (key, value) in &additional_defines {
388
+ tesseract_config.define(key, value);
389
+ }
390
+
391
+ tesseract_config.build();
392
+ });
393
+
394
+ println!("cargo:rerun-if-changed=build.rs");
395
+ println!("cargo:rerun-if-changed={}", third_party_dir.display());
396
+ println!("cargo:rerun-if-changed={}", leptonica_dir.display());
397
+ println!("cargo:rerun-if-changed={}", tesseract_dir.display());
398
+
399
+ println!("cargo:rustc-link-search=native={}", leptonica_lib_dir.display());
400
+ println!(
401
+ "cargo:rustc-link-search=native={}",
402
+ tesseract_install_dir.join("lib").display()
403
+ );
404
+
405
+ // Link libraries in the correct order for static linking:
406
+ // 1. tesseract first (depends on leptonica and C++ stdlib)
407
+ // 2. leptonica (depends on C++ stdlib)
408
+ // 3. C++ standard library and system libraries (via set_os_specific_link_flags)
409
+ //
410
+ // IMPORTANT: For static linking, the linker resolves symbols in order.
411
+ // Libraries must be listed BEFORE the libraries they depend on.
412
+ // The C++ stdlib must come LAST because both tesseract and leptonica
413
+ // depend on it for symbols like operator new, operator delete, etc.
414
+ #[cfg(feature = "dynamic-linking")]
415
+ let link_type = "dylib";
416
+ #[cfg(not(feature = "dynamic-linking"))]
417
+ let link_type = "static";
418
+
419
+ println!("cargo:rustc-link-lib={}={}", link_type, tesseract_link_name);
420
+ println!(
421
+ "cargo:warning=Linking with tesseract ({} linking): {}",
422
+ link_type, tesseract_link_name
423
+ );
424
+ println!("cargo:rustc-link-lib={}={}", link_type, leptonica_link_name);
425
+ println!(
426
+ "cargo:warning=Linking with leptonica ({} linking): {}",
427
+ link_type, leptonica_link_name
428
+ );
429
+
430
+ // Link C++ standard library and system libraries AFTER tesseract and leptonica.
431
+ // This is critical for static linking on Linux (especially aarch64) where
432
+ // tesseract's C++ code needs symbols like operator new/delete from libstdc++.
433
+ set_os_specific_link_flags();
434
+
435
+ println!("cargo:warning=Leptonica include dir: {:?}", leptonica_include_dir);
436
+ println!("cargo:warning=Leptonica lib dir: {:?}", leptonica_lib_dir);
437
+ println!("cargo:warning=Tesseract install dir: {:?}", tesseract_install_dir);
438
+ println!("cargo:warning=Tessdata dir: {:?}", tessdata_prefix);
439
+ }
440
+
441
+ fn get_os_specific_config() -> (String, Vec<(String, String)>) {
442
+ let mut cmake_cxx_flags = String::new();
443
+ let mut additional_defines = Vec::new();
444
+ let target = target_triple();
445
+ let target_macos = is_macos_target(&target);
446
+ let target_linux = is_linux_target(&target);
447
+ let target_windows = is_windows_target(&target);
448
+ let target_msvc = is_msvc_target(&target);
449
+ let target_mingw = is_mingw_target(&target);
450
+ let target_musl = target.contains("musl");
451
+
452
+ if target_macos {
453
+ cmake_cxx_flags.push_str("-stdlib=libc++ ");
454
+ cmake_cxx_flags.push_str("-std=c++17 ");
455
+ } else if target_linux {
456
+ cmake_cxx_flags.push_str("-std=c++17 ");
457
+ if target_musl || env::var("CC").map(|cc| cc.contains("clang")).unwrap_or(false) {
458
+ cmake_cxx_flags.push_str("-stdlib=libc++ ");
459
+ let cxx_compiler = env::var("CXX").unwrap_or_else(|_| {
460
+ if let Ok(target) = env::var("TARGET") {
461
+ if target != env::var("HOST").unwrap_or_default() {
462
+ format!("{}-clang++", target)
463
+ } else {
464
+ "clang++".to_string()
465
+ }
466
+ } else {
467
+ "clang++".to_string()
468
+ }
469
+ });
470
+ additional_defines.push(("CMAKE_CXX_COMPILER".to_string(), cxx_compiler));
471
+ } else {
472
+ let cxx_compiler = env::var("CXX").unwrap_or_else(|_| {
473
+ if let Ok(target) = env::var("TARGET") {
474
+ if target != env::var("HOST").unwrap_or_default() {
475
+ format!("{}-g++", target)
476
+ } else {
477
+ "g++".to_string()
478
+ }
479
+ } else {
480
+ "g++".to_string()
481
+ }
482
+ });
483
+ additional_defines.push(("CMAKE_CXX_COMPILER".to_string(), cxx_compiler));
484
+ }
485
+ } else if target_windows {
486
+ if target_msvc {
487
+ cmake_cxx_flags.push_str("/EHsc /MP /std:c++17 /DTESSERACT_STATIC ");
488
+ additional_defines.push(("CMAKE_C_FLAGS_RELEASE".to_string(), "/MD /O2".to_string()));
489
+ additional_defines.push(("CMAKE_C_FLAGS_DEBUG".to_string(), "/MDd /Od".to_string()));
490
+ additional_defines.push((
491
+ "CMAKE_CXX_FLAGS_RELEASE".to_string(),
492
+ "/MD /O2 /DTESSERACT_STATIC".to_string(),
493
+ ));
494
+ additional_defines.push((
495
+ "CMAKE_CXX_FLAGS_DEBUG".to_string(),
496
+ "/MDd /Od /DTESSERACT_STATIC".to_string(),
497
+ ));
498
+ additional_defines.push(("CMAKE_MSVC_RUNTIME_LIBRARY".to_string(), "MultiThreadedDLL".to_string()));
499
+ } else if target_mingw {
500
+ cmake_cxx_flags.push_str("-std=c++17 -DTESSERACT_STATIC ");
501
+ additional_defines.push(("CMAKE_C_FLAGS_RELEASE".to_string(), "-O2 -DNDEBUG".to_string()));
502
+ additional_defines.push(("CMAKE_C_FLAGS_DEBUG".to_string(), "-O0 -g".to_string()));
503
+ additional_defines.push(("CMAKE_C_COMPILER".to_string(), "gcc".to_string()));
504
+ additional_defines.push(("CMAKE_CXX_COMPILER".to_string(), "g++".to_string()));
505
+ additional_defines.push(("CMAKE_SYSTEM_NAME".to_string(), "Windows".to_string()));
506
+ additional_defines.push((
507
+ "CMAKE_CXX_FLAGS_RELEASE".to_string(),
508
+ "-O2 -DNDEBUG -DTESSERACT_STATIC".to_string(),
509
+ ));
510
+ additional_defines.push((
511
+ "CMAKE_CXX_FLAGS_DEBUG".to_string(),
512
+ "-O0 -g -DTESSERACT_STATIC".to_string(),
513
+ ));
514
+ } else {
515
+ cmake_cxx_flags.push_str("-std=c++17 -DTESSERACT_STATIC ");
516
+ additional_defines.push(("CMAKE_C_FLAGS_RELEASE".to_string(), "-O2 -DNDEBUG".to_string()));
517
+ additional_defines.push(("CMAKE_C_FLAGS_DEBUG".to_string(), "-O0 -g".to_string()));
518
+ additional_defines.push((
519
+ "CMAKE_CXX_FLAGS_RELEASE".to_string(),
520
+ "-O2 -DNDEBUG -DTESSERACT_STATIC".to_string(),
521
+ ));
522
+ additional_defines.push((
523
+ "CMAKE_CXX_FLAGS_DEBUG".to_string(),
524
+ "-O0 -g -DTESSERACT_STATIC".to_string(),
525
+ ));
526
+ }
527
+ }
528
+
529
+ cmake_cxx_flags.push_str("-DUSE_STD_NAMESPACE ");
530
+ additional_defines.push(("CMAKE_POSITION_INDEPENDENT_CODE".to_string(), "ON".to_string()));
531
+
532
+ if target_windows && target_msvc {
533
+ cmake_cxx_flags.push_str("/permissive- ");
534
+ additional_defines.push(("CMAKE_EXE_LINKER_FLAGS".to_string(), "/INCREMENTAL:NO".to_string()));
535
+ additional_defines.push(("CMAKE_SHARED_LINKER_FLAGS".to_string(), "/INCREMENTAL:NO".to_string()));
536
+ additional_defines.push(("CMAKE_MODULE_LINKER_FLAGS".to_string(), "/INCREMENTAL:NO".to_string()));
537
+ }
538
+
539
+ (cmake_cxx_flags, additional_defines)
540
+ }
541
+
542
+ fn set_os_specific_link_flags() {
543
+ let target = target_triple();
544
+ let target_macos = is_macos_target(&target);
545
+ let target_linux = is_linux_target(&target);
546
+ let target_windows = is_windows_target(&target);
547
+ let target_mingw = is_mingw_target(&target);
548
+ let target_musl = target.contains("musl");
549
+
550
+ if target_macos {
551
+ println!("cargo:rustc-link-lib=c++");
552
+ } else if target_linux {
553
+ if target_musl || env::var("CC").map(|cc| cc.contains("clang")).unwrap_or(false) {
554
+ println!("cargo:rustc-link-lib=c++");
555
+ } else {
556
+ println!("cargo:rustc-link-lib=stdc++");
557
+ println!("cargo:rustc-link-lib=stdc++fs");
558
+ }
559
+ println!("cargo:rustc-link-lib=pthread");
560
+ println!("cargo:rustc-link-lib=m");
561
+ println!("cargo:rustc-link-lib=dl");
562
+ } else if target_windows {
563
+ if target_mingw {
564
+ println!("cargo:rustc-link-lib=stdc++");
565
+ }
566
+ println!("cargo:rustc-link-lib=user32");
567
+ println!("cargo:rustc-link-lib=gdi32");
568
+ println!("cargo:rustc-link-lib=ws2_32");
569
+ println!("cargo:rustc-link-lib=advapi32");
570
+ println!("cargo:rustc-link-lib=shell32");
571
+ }
572
+
573
+ println!("cargo:rustc-link-search=native={}", env::var("OUT_DIR").unwrap());
574
+ }
575
+
576
+ fn download_and_extract(target_dir: &Path, url: &str, name: &str) -> PathBuf {
577
+ use zip::ZipArchive;
578
+
579
+ fs::create_dir_all(target_dir).expect("Failed to create target directory");
580
+
581
+ let client = reqwest::blocking::Client::builder()
582
+ .timeout(std::time::Duration::from_secs(300))
583
+ .http1_only()
584
+ .build()
585
+ .expect("Failed to create HTTP client");
586
+
587
+ println!("cargo:warning=Downloading {} from {}", name, url);
588
+ let max_attempts = 5;
589
+ let mut content = None;
590
+
591
+ for attempt in 1..=max_attempts {
592
+ let err_msg = match client.get(url).send() {
593
+ Ok(resp) => {
594
+ if resp.status().is_success() {
595
+ match resp.bytes() {
596
+ Ok(bytes) => {
597
+ content = Some(bytes.to_vec());
598
+ break;
599
+ }
600
+ Err(err) => format!("Failed to read response: {}", err),
601
+ }
602
+ } else {
603
+ format!("HTTP {}", resp.status().as_u16())
604
+ }
605
+ }
606
+ Err(err) => err.to_string(),
607
+ };
608
+
609
+ if attempt == max_attempts {
610
+ panic!(
611
+ "Failed to download {} after {} attempts: {}",
612
+ name, max_attempts, err_msg
613
+ );
614
+ }
615
+
616
+ let backoff = 2u64.pow((attempt - 1).min(4));
617
+ println!(
618
+ "cargo:warning=Download attempt {}/{} for {} failed ({}). Retrying in {}s...",
619
+ attempt, max_attempts, name, err_msg, backoff
620
+ );
621
+ std::thread::sleep(std::time::Duration::from_secs(backoff));
622
+ }
623
+
624
+ let content = content.expect("unreachable: download loop must either succeed or panic");
625
+
626
+ println!("cargo:warning=Downloaded {} bytes for {}", content.len(), name);
627
+
628
+ let temp_file = target_dir.join(format!("{}.zip", name));
629
+ fs::write(&temp_file, content).expect("Failed to write archive to file");
630
+
631
+ let extract_dir = target_dir.join(name);
632
+ if extract_dir.exists() {
633
+ fs::remove_dir_all(&extract_dir).expect("Failed to remove existing directory");
634
+ }
635
+ fs::create_dir_all(&extract_dir).expect("Failed to create extraction directory");
636
+
637
+ let mut archive = ZipArchive::new(fs::File::open(&temp_file).unwrap()).unwrap();
638
+
639
+ for i in 0..archive.len() {
640
+ let mut file = archive.by_index(i).unwrap();
641
+ let file_path = file.mangled_name();
642
+ let file_path = file_path.to_str().unwrap();
643
+
644
+ let path = Path::new(file_path);
645
+ let path = path.strip_prefix(path.components().next().unwrap()).unwrap();
646
+
647
+ if path.as_os_str().is_empty() {
648
+ continue;
649
+ }
650
+
651
+ let target_path = extract_dir.join(path);
652
+
653
+ if file.is_dir() {
654
+ fs::create_dir_all(target_path).unwrap();
655
+ } else {
656
+ if let Some(parent) = target_path.parent() {
657
+ fs::create_dir_all(parent).unwrap();
658
+ }
659
+ let mut outfile = fs::File::create(target_path).unwrap();
660
+ std::io::copy(&mut file, &mut outfile).unwrap();
661
+ }
662
+ }
663
+
664
+ fs::remove_file(temp_file).expect("Failed to remove temporary zip file");
665
+
666
+ extract_dir
667
+ }
668
+
669
+ fn normalize_cmake_path(path: &Path) -> String {
670
+ path.to_string_lossy().replace('\\', "/")
671
+ }
672
+
673
+ fn clean_cache(cache_dir: &Path) {
674
+ println!("Cleaning cache directory: {:?}", cache_dir);
675
+ if cache_dir.exists() {
676
+ fs::remove_dir_all(cache_dir).expect("Failed to remove cache directory");
677
+ }
678
+ }
679
+
680
+ #[allow(dead_code)]
681
+ fn apply_patches(src_dir: &Path, patch_dir: &Path) -> std::io::Result<()> {
682
+ use std::process::Command;
683
+
684
+ let patch_file = patch_dir.join("tesseract.diff");
685
+ if !patch_file.exists() {
686
+ println!("cargo:warning=Patch file not found: {}", patch_file.display());
687
+ return Ok(());
688
+ }
689
+
690
+ println!("cargo:warning=Applying patches from: {}", patch_file.display());
691
+
692
+ let output = Command::new("patch")
693
+ .arg("-p1")
694
+ .current_dir(src_dir)
695
+ .stdin(fs::File::open(&patch_file)?)
696
+ .output()?;
697
+
698
+ if !output.status.success() {
699
+ let stderr = String::from_utf8_lossy(&output.stderr);
700
+ println!("cargo:warning=Patch output: {}", stderr);
701
+ }
702
+
703
+ Ok(())
704
+ }
705
+
706
+ #[allow(dead_code)]
707
+ fn setup_emsdk_environment(emsdk_root: &Path) {
708
+ let emsdk_root_str = emsdk_root.to_string_lossy().to_string();
709
+
710
+ unsafe {
711
+ env::set_var("EMSDK", &emsdk_root_str);
712
+ }
713
+ println!("cargo:warning=Set EMSDK={}", emsdk_root_str);
714
+
715
+ let emscripten_path = format!("{}/upstream/emscripten", emsdk_root_str);
716
+ unsafe {
717
+ env::set_var("EMSCRIPTEN", &emscripten_path);
718
+ }
719
+ println!("cargo:warning=Set EMSCRIPTEN={}", emscripten_path);
720
+
721
+ let em_config = format!("{}/.emscripten", emsdk_root_str);
722
+ unsafe {
723
+ env::set_var("EM_CONFIG", &em_config);
724
+ }
725
+ println!("cargo:warning=Set EM_CONFIG={}", em_config);
726
+
727
+ let emsdk_bin_paths = [
728
+ format!("{}/upstream/emscripten", emsdk_root_str),
729
+ format!("{}/upstream/bin", emsdk_root_str),
730
+ ];
731
+
732
+ let current_path = env::var("PATH").unwrap_or_default();
733
+ let new_path = format!(
734
+ "{}{}{}",
735
+ emsdk_bin_paths.join(":"),
736
+ if current_path.is_empty() { "" } else { ":" },
737
+ current_path
738
+ );
739
+ unsafe {
740
+ env::set_var("PATH", &new_path);
741
+ }
742
+ println!("cargo:warning=Updated PATH with EMSDK tools");
743
+ }
744
+
745
+ #[allow(dead_code)]
746
+ fn build_leptonica_wasm(leptonica_src: &Path, leptonica_install: &Path, emsdk_dir: &Path) {
747
+ let toolchain_file = emsdk_dir.join("upstream/emscripten/cmake/Modules/Platform/Emscripten.cmake");
748
+
749
+ let mut config = Config::new(leptonica_src);
750
+
751
+ config.define("CMAKE_TOOLCHAIN_FILE", &toolchain_file);
752
+
753
+ config
754
+ .define("CMAKE_BUILD_TYPE", "Release")
755
+ .define("CMAKE_POLICY_VERSION_MINIMUM", "3.5")
756
+ .define("LIBWEBP_SUPPORT", "OFF")
757
+ .define("OPENJPEG_SUPPORT", "OFF")
758
+ .define("ENABLE_ZLIB", "OFF")
759
+ .define("ENABLE_PNG", "OFF")
760
+ .define("ENABLE_JPEG", "OFF")
761
+ .define("ENABLE_TIFF", "OFF")
762
+ .define("ENABLE_WEBP", "OFF")
763
+ .define("ENABLE_OPENJPEG", "OFF")
764
+ .define("ENABLE_GIF", "OFF")
765
+ .define("BUILD_PROG", "OFF")
766
+ .define("BUILD_SHARED_LIBS", "OFF")
767
+ .define("NO_CONSOLE_IO", "ON")
768
+ .define("HAVE_LIBZ", "0")
769
+ .define("ENABLE_LTO", "OFF")
770
+ .define("CMAKE_INSTALL_PREFIX", leptonica_install);
771
+
772
+ config.build();
773
+ }
774
+
775
+ #[allow(dead_code)]
776
+ fn build_wasm() {
777
+ println!("cargo:warning=Building for WASM target with Emscripten SDK");
778
+
779
+ let custom_out_dir = prepare_out_dir();
780
+ let cache_dir = custom_out_dir.join("cache");
781
+ fs::create_dir_all(&cache_dir).expect("Failed to create cache directory");
782
+
783
+ let project_dir = custom_out_dir.clone();
784
+ let third_party_dir = project_dir.join("third_party");
785
+
786
+ println!("cargo:warning=Looking for pre-installed Emscripten SDK...");
787
+ let emsdk_dir = match find_emsdk() {
788
+ Ok(path) => {
789
+ println!("cargo:warning=Found EMSDK at: {}", path.display());
790
+ path
791
+ }
792
+ Err(err) => {
793
+ panic!("{}
794
+
795
+ Installation instructions:
796
+ macOS: brew install emscripten
797
+ Ubuntu: sudo apt-get install emscripten
798
+ Arch: sudo pacman -S emscripten
799
+ Manual: git clone https://github.com/emscripten-core/emsdk.git ~/.emsdk && cd ~/.emsdk && ./emsdk install latest && ./emsdk activate latest
800
+
801
+ After installation, set EMSDK environment variable or ensure it's in a standard location.", err);
802
+ }
803
+ };
804
+
805
+ setup_emsdk_environment(&emsdk_dir);
806
+
807
+ let leptonica_dir = if third_party_dir.join("leptonica").exists() {
808
+ println!("cargo:warning=Using existing leptonica source");
809
+ third_party_dir.join("leptonica")
810
+ } else {
811
+ fs::create_dir_all(&third_party_dir).expect("Failed to create third_party directory");
812
+ download_and_extract(&third_party_dir, &leptonica_url(), "leptonica")
813
+ };
814
+
815
+ let tesseract_dir = if third_party_dir.join("tesseract").exists() {
816
+ println!("cargo:warning=Using existing tesseract source");
817
+ third_party_dir.join("tesseract")
818
+ } else {
819
+ fs::create_dir_all(&third_party_dir).expect("Failed to create third_party directory");
820
+ download_and_extract(&third_party_dir, &tesseract_url(), "tesseract")
821
+ };
822
+
823
+ let leptonica_install_dir = custom_out_dir.join("leptonica");
824
+ let leptonica_cache_dir = cache_dir.join("leptonica");
825
+
826
+ let _leptonica_link_name =
827
+ build_or_use_cached("leptonica", &leptonica_cache_dir, &leptonica_install_dir, || {
828
+ println!("cargo:warning=Building Leptonica for WASM...");
829
+ build_leptonica_wasm(&leptonica_dir, &leptonica_install_dir, &emsdk_dir);
830
+ });
831
+
832
+ let tesseract_install_dir = custom_out_dir.join("tesseract");
833
+ let tesseract_cache_dir = cache_dir.join("tesseract");
834
+
835
+ let _tesseract_link_name =
836
+ build_or_use_cached("tesseract", &tesseract_cache_dir, &tesseract_install_dir, || {
837
+ println!("cargo:warning=Building Tesseract for WASM (SIMD enabled)...");
838
+ build_tesseract_wasm(
839
+ &tesseract_dir,
840
+ &tesseract_install_dir,
841
+ &leptonica_install_dir,
842
+ &emsdk_dir,
843
+ true,
844
+ );
845
+ });
846
+
847
+ let leptonica_lib_dir = leptonica_install_dir.join("lib");
848
+ let tesseract_lib_dir = tesseract_install_dir.join("lib");
849
+
850
+ println!("cargo:rustc-link-search=native={}", leptonica_lib_dir.display());
851
+ println!("cargo:rustc-link-search=native={}", tesseract_lib_dir.display());
852
+
853
+ println!("cargo:rustc-link-lib=static=tesseract");
854
+ println!("cargo:rustc-link-lib=static=leptonica");
855
+
856
+ println!("cargo:warning=WASM build completed successfully!");
857
+ println!("cargo:warning=Leptonica install dir: {:?}", leptonica_install_dir);
858
+ println!("cargo:warning=Tesseract install dir: {:?}", tesseract_install_dir);
859
+ }
860
+
861
+ #[allow(dead_code)]
862
+ fn build_tesseract_wasm(
863
+ src_dir: &Path,
864
+ tesseract_install: &Path,
865
+ leptonica_install: &Path,
866
+ emsdk_dir: &Path,
867
+ enable_simd: bool,
868
+ ) {
869
+ let toolchain_file = emsdk_dir.join("upstream/emscripten/cmake/Modules/Platform/Emscripten.cmake");
870
+
871
+ let mut config = Config::new(src_dir);
872
+
873
+ config.define("CMAKE_TOOLCHAIN_FILE", &toolchain_file);
874
+
875
+ config.define("Leptonica_DIR", leptonica_install);
876
+ config.define("CMAKE_PREFIX_PATH", leptonica_install);
877
+
878
+ let mut cxx_flags = String::from("-DTESSERACT_IMAGEDATA_AS_PIX ");
879
+ if enable_simd {
880
+ cxx_flags.push_str("-msimd128 ");
881
+ }
882
+ cxx_flags.push_str("-fno-exceptions -fPIC -Os");
883
+
884
+ let c_flags = "-fno-exceptions -fPIC -Os";
885
+
886
+ config
887
+ .define("CMAKE_BUILD_TYPE", "Release")
888
+ .define("CMAKE_POLICY_VERSION_MINIMUM", "3.5")
889
+ .define("BUILD_TESSERACT_BINARY", "OFF")
890
+ .define("BUILD_TRAINING_TOOLS", "OFF")
891
+ .define("INSTALL_CONFIGS", "ON")
892
+ .define("BUILD_TESTS", "OFF")
893
+ .define("BUILD_PROG", "OFF")
894
+ .define("SYNTAX_LOG", "OFF")
895
+ .define("DISABLE_ARCHIVE", "ON")
896
+ .define("DISABLE_CURL", "ON")
897
+ .define("DISABLE_OPENCL", "ON")
898
+ .define("DISABLE_TIFF", "ON")
899
+ .define("DISABLE_PNG", "ON")
900
+ .define("DISABLE_JPEG", "ON")
901
+ .define("DISABLE_WEBP", "ON")
902
+ .define("DISABLE_OPENJPEG", "ON")
903
+ .define("DISABLE_ZLIB", "ON")
904
+ .define("DISABLE_LIBXML2", "ON")
905
+ .define("DISABLE_LIBICU", "ON")
906
+ .define("DISABLE_LZMA", "ON")
907
+ .define("DISABLE_GIF", "ON")
908
+ .define("DISABLE_DEBUG_MESSAGES", "ON")
909
+ .define("GRAPHICS_DISABLED", "ON")
910
+ .define("DISABLED_LEGACY_ENGINE", "ON")
911
+ .define("USE_OPENCL", "OFF")
912
+ .define("OPENMP_BUILD", "OFF")
913
+ .define("ENABLE_LTO", "ON")
914
+ .define("HAVE_SSE4_1", if enable_simd { "ON" } else { "OFF" })
915
+ .define("HAVE_AVX", "OFF")
916
+ .define("HAVE_AVX2", "OFF")
917
+ .define("HAVE_AVX512F", "OFF")
918
+ .define("HAVE_FMA", "OFF")
919
+ .define("CMAKE_INSTALL_PREFIX", tesseract_install)
920
+ .define("CMAKE_CXX_FLAGS", &cxx_flags)
921
+ .define("CMAKE_C_FLAGS", c_flags);
922
+
923
+ config.build();
924
+ }
925
+
926
+ fn build_or_use_cached<F>(name: &str, cache_dir: &Path, install_dir: &Path, build_fn: F) -> String
927
+ where
928
+ F: FnOnce(),
929
+ {
930
+ let target_env = env::var("CARGO_CFG_TARGET_ENV").unwrap_or_default();
931
+ let target_triple = env::var("TARGET")
932
+ .unwrap_or_else(|_| env::var("CARGO_CFG_TARGET_ARCH").unwrap_or_else(|_| "unknown".to_string()));
933
+ let is_windows = target_triple.contains("windows");
934
+ let is_windows_gnu = is_windows && target_env == "gnu";
935
+
936
+ let lib_name = if is_windows && !is_windows_gnu {
937
+ format!("{}.lib", name)
938
+ } else {
939
+ format!("lib{}.a", name)
940
+ };
941
+
942
+ let cached_path = cache_dir.join(&lib_name);
943
+ let marker_path = cache_dir.join(format!("{}.target", name));
944
+ let out_path = install_dir.join("lib").join(&lib_name);
945
+
946
+ let possible_lib_names: Vec<String> = if is_windows {
947
+ let mut base = match name {
948
+ "leptonica" => vec![
949
+ "leptonica.lib".to_string(),
950
+ "libleptonica.lib".to_string(),
951
+ "leptonica-static.lib".to_string(),
952
+ "leptonica-1.84.1.lib".to_string(),
953
+ "leptonica-1.86.0.lib".to_string(),
954
+ "leptonicad.lib".to_string(),
955
+ "libleptonica_d.lib".to_string(),
956
+ "leptonica-1.84.1d.lib".to_string(),
957
+ "leptonica-1.86.0d.lib".to_string(),
958
+ ],
959
+ "tesseract" => vec![
960
+ "tesseract.lib".to_string(),
961
+ "libtesseract.lib".to_string(),
962
+ "tesseract-static.lib".to_string(),
963
+ "tesseract53.lib".to_string(),
964
+ "tesseract54.lib".to_string(),
965
+ "tesseract55.lib".to_string(),
966
+ "tesseractd.lib".to_string(),
967
+ "libtesseract_d.lib".to_string(),
968
+ "tesseract53d.lib".to_string(),
969
+ "tesseract54d.lib".to_string(),
970
+ "tesseract55d.lib".to_string(),
971
+ ],
972
+ _ => vec![format!("{}.lib", name)],
973
+ };
974
+
975
+ if is_windows_gnu {
976
+ match name {
977
+ "leptonica" => {
978
+ base.push(format!("libleptonica-{}.a", LEPTONICA_VERSION));
979
+ base.push("libleptonica.a".to_string());
980
+ }
981
+ "tesseract" => {
982
+ base.push(format!("libtesseract{}.a", TESSERACT_VERSION.replace('.', "")));
983
+ base.push("libtesseract.a".to_string());
984
+ base.push("libtesseract55.a".to_string());
985
+ }
986
+ _ => {
987
+ base.push(format!("lib{}.a", name));
988
+ }
989
+ }
990
+ }
991
+
992
+ base
993
+ } else {
994
+ vec![format!("lib{}.a", name)]
995
+ };
996
+
997
+ fs::create_dir_all(cache_dir).expect("Failed to create cache directory");
998
+ fs::create_dir_all(out_path.parent().unwrap()).expect("Failed to create output directory");
999
+
1000
+ let candidate_lib_dirs = [
1001
+ install_dir.join("lib"),
1002
+ install_dir.join("lib64"),
1003
+ install_dir.join("lib").join("tesseract"),
1004
+ ];
1005
+
1006
+ let cache_valid = cached_path.exists()
1007
+ && {
1008
+ match fs::read_to_string(&marker_path) {
1009
+ Ok(cached_target) => {
1010
+ let valid = cached_target.trim() == target_triple;
1011
+ if !valid {
1012
+ println!(
1013
+ "cargo:warning=Cached {} library is for wrong architecture (cached: {}, current: {}), rebuilding",
1014
+ name,
1015
+ cached_target.trim(),
1016
+ target_triple
1017
+ );
1018
+ let _ = fs::remove_file(&cached_path);
1019
+ let _ = fs::remove_file(&marker_path);
1020
+ }
1021
+ valid
1022
+ }
1023
+ Err(_) => {
1024
+ println!(
1025
+ "cargo:warning=Cached {} library missing target marker, rebuilding",
1026
+ name
1027
+ );
1028
+ let _ = fs::remove_file(&cached_path);
1029
+ false
1030
+ }
1031
+ }
1032
+ };
1033
+
1034
+ let link_name_to_use = if cache_valid {
1035
+ println!("cargo:warning=Using cached {} library for {}", name, target_triple);
1036
+ if let Err(e) = fs::copy(&cached_path, &out_path) {
1037
+ println!("cargo:warning=Failed to copy cached library: {}", e);
1038
+ build_fn();
1039
+ }
1040
+ name.to_string()
1041
+ } else {
1042
+ println!("Building {} library", name);
1043
+ build_fn();
1044
+
1045
+ let mut found_lib_name = None;
1046
+ 'search: for lib_name in &possible_lib_names {
1047
+ for dir in &candidate_lib_dirs {
1048
+ let lib_path = dir.join(lib_name);
1049
+ if lib_path.exists() {
1050
+ println!("cargo:warning=Found {} library at: {}", name, lib_path.display());
1051
+ let link_name = if lib_name.ends_with(".lib") {
1052
+ lib_name.strip_suffix(".lib").unwrap_or(lib_name).to_string()
1053
+ } else if lib_name.ends_with(".a") {
1054
+ lib_name
1055
+ .strip_prefix("lib")
1056
+ .and_then(|s| s.strip_suffix(".a"))
1057
+ .unwrap_or(lib_name)
1058
+ .to_string()
1059
+ } else {
1060
+ lib_name.to_string()
1061
+ };
1062
+ found_lib_name = Some((lib_path, link_name));
1063
+ break 'search;
1064
+ }
1065
+ }
1066
+ }
1067
+
1068
+ if let Some((lib_path, link_name)) = found_lib_name {
1069
+ if out_path.exists() {
1070
+ println!(
1071
+ "cargo:warning=Library already available at expected location: {}",
1072
+ out_path.display()
1073
+ );
1074
+ } else if let Err(e) = fs::copy(&lib_path, &out_path) {
1075
+ println!("cargo:warning=Failed to copy library to standard location: {}", e);
1076
+ }
1077
+ if let Err(e) = fs::copy(&lib_path, &cached_path) {
1078
+ println!("cargo:warning=Failed to cache library: {}", e);
1079
+ } else if let Err(e) = fs::write(&marker_path, &target_triple) {
1080
+ println!("cargo:warning=Failed to write cache marker: {}", e);
1081
+ } else {
1082
+ println!("cargo:warning=Cached {} library for {}", name, target_triple);
1083
+ }
1084
+ link_name
1085
+ } else {
1086
+ println!(
1087
+ "cargo:warning=Library {} not found! Searched for: {:?}",
1088
+ name, possible_lib_names
1089
+ );
1090
+ for dir in &candidate_lib_dirs {
1091
+ println!("cargo:warning=Checked directory: {}", dir.display());
1092
+ if let Ok(entries) = fs::read_dir(dir) {
1093
+ println!("cargo:warning=Files in {}:", dir.display());
1094
+ for entry in entries.flatten() {
1095
+ println!("cargo:warning= - {}", entry.file_name().to_string_lossy());
1096
+ }
1097
+ } else {
1098
+ println!("cargo:warning=Directory not accessible: {}", dir.display());
1099
+ }
1100
+ }
1101
+ name.to_string()
1102
+ }
1103
+ };
1104
+
1105
+ for dir in candidate_lib_dirs.iter().filter(|d| d.exists()) {
1106
+ println!("cargo:rustc-link-search=native={}", dir.display());
1107
+ }
1108
+
1109
+ // Return the link name instead of outputting the link directive here
1110
+ // This allows the caller to control the linking order
1111
+ link_name_to_use
1112
+ }
1113
+ }
1114
+
1115
+ fn main() {
1116
+ #[cfg(feature = "build-tesseract")]
1117
+ {
1118
+ build_tesseract::build();
1119
+ }
1120
+
1121
+ #[cfg(all(feature = "dynamic-linking", not(feature = "build-tesseract")))]
1122
+ {
1123
+ println!("cargo:warning=Using dynamic linking with system-installed Tesseract libraries");
1124
+ println!("cargo:rustc-link-lib=dylib=tesseract");
1125
+ println!("cargo:rustc-link-lib=dylib=leptonica");
1126
+ }
1127
+ }