body-reader.js 23 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687
  1. exports.createBodyReader = createBodyReader;
  2. exports._readNumberingProperties = readNumberingProperties;
  3. var dingbatToUnicode = require("dingbat-to-unicode");
  4. var _ = require("underscore");
  5. var documents = require("../documents");
  6. var Result = require("../results").Result;
  7. var warning = require("../results").warning;
  8. var uris = require("./uris");
  9. function createBodyReader(options) {
  10. return {
  11. readXmlElement: function(element) {
  12. return new BodyReader(options).readXmlElement(element);
  13. },
  14. readXmlElements: function(elements) {
  15. return new BodyReader(options).readXmlElements(elements);
  16. }
  17. };
  18. }
  19. function BodyReader(options) {
  20. var complexFieldStack = [];
  21. var currentInstrText = [];
  22. // When a paragraph is marked as deleted, its contents should be combined
  23. // with the following paragraph. See 17.13.5.15 del (Deleted Paragraph) of
  24. // ECMA-376 4th edition Part 1.
  25. var deletedParagraphContents = [];
  26. var relationships = options.relationships;
  27. var contentTypes = options.contentTypes;
  28. var docxFile = options.docxFile;
  29. var files = options.files;
  30. var numbering = options.numbering;
  31. var styles = options.styles;
  32. function readXmlElements(elements) {
  33. var results = elements.map(readXmlElement);
  34. return combineResults(results);
  35. }
  36. function readXmlElement(element) {
  37. if (element.type === "element") {
  38. var handler = xmlElementReaders[element.name];
  39. if (handler) {
  40. return handler(element);
  41. } else if (!Object.prototype.hasOwnProperty.call(ignoreElements, element.name)) {
  42. var message = warning("An unrecognised element was ignored: " + element.name);
  43. return emptyResultWithMessages([message]);
  44. }
  45. }
  46. return emptyResult();
  47. }
  48. function readParagraphProperties(element) {
  49. return readParagraphStyle(element).map(function(style) {
  50. return {
  51. type: "paragraphProperties",
  52. styleId: style.styleId,
  53. styleName: style.name,
  54. alignment: element.firstOrEmpty("w:jc").attributes["w:val"],
  55. numbering: readNumberingProperties(style.styleId, element.firstOrEmpty("w:numPr"), numbering),
  56. indent: readParagraphIndent(element.firstOrEmpty("w:ind"))
  57. };
  58. });
  59. }
  60. function readParagraphIndent(element) {
  61. return {
  62. start: element.attributes["w:start"] || element.attributes["w:left"],
  63. end: element.attributes["w:end"] || element.attributes["w:right"],
  64. firstLine: element.attributes["w:firstLine"],
  65. hanging: element.attributes["w:hanging"]
  66. };
  67. }
  68. function readRunProperties(element) {
  69. return readRunStyle(element).map(function(style) {
  70. var fontSizeString = element.firstOrEmpty("w:sz").attributes["w:val"];
  71. // w:sz gives the font size in half points, so halve the value to get the size in points
  72. var fontSize = /^[0-9]+$/.test(fontSizeString) ? parseInt(fontSizeString, 10) / 2 : null;
  73. return {
  74. type: "runProperties",
  75. styleId: style.styleId,
  76. styleName: style.name,
  77. verticalAlignment: element.firstOrEmpty("w:vertAlign").attributes["w:val"],
  78. font: element.firstOrEmpty("w:rFonts").attributes["w:ascii"],
  79. fontSize: fontSize,
  80. isBold: readBooleanElement(element.first("w:b")),
  81. isUnderline: readUnderline(element.first("w:u")),
  82. isItalic: readBooleanElement(element.first("w:i")),
  83. isStrikethrough: readBooleanElement(element.first("w:strike")),
  84. isAllCaps: readBooleanElement(element.first("w:caps")),
  85. isSmallCaps: readBooleanElement(element.first("w:smallCaps")),
  86. highlight: readHighlightValue(element.firstOrEmpty("w:highlight").attributes["w:val"])
  87. };
  88. });
  89. }
  90. function readUnderline(element) {
  91. if (element) {
  92. var value = element.attributes["w:val"];
  93. return value !== undefined && value !== "false" && value !== "0" && value !== "none";
  94. } else {
  95. return false;
  96. }
  97. }
  98. function readBooleanElement(element) {
  99. if (element) {
  100. var value = element.attributes["w:val"];
  101. return value !== "false" && value !== "0";
  102. } else {
  103. return false;
  104. }
  105. }
  106. function readHighlightValue(value) {
  107. if (!value || value === "none") {
  108. return null;
  109. } else {
  110. return value;
  111. }
  112. }
  113. function readParagraphStyle(element) {
  114. return readStyle(element, "w:pStyle", "Paragraph", styles.findParagraphStyleById);
  115. }
  116. function readRunStyle(element) {
  117. return readStyle(element, "w:rStyle", "Run", styles.findCharacterStyleById);
  118. }
  119. function readTableStyle(element) {
  120. return readStyle(element, "w:tblStyle", "Table", styles.findTableStyleById);
  121. }
  122. function readStyle(element, styleTagName, styleType, findStyleById) {
  123. var messages = [];
  124. var styleElement = element.first(styleTagName);
  125. var styleId = null;
  126. var name = null;
  127. if (styleElement) {
  128. styleId = styleElement.attributes["w:val"];
  129. if (styleId) {
  130. var style = findStyleById(styleId);
  131. if (style) {
  132. name = style.name;
  133. } else {
  134. messages.push(undefinedStyleWarning(styleType, styleId));
  135. }
  136. }
  137. }
  138. return elementResultWithMessages({styleId: styleId, name: name}, messages);
  139. }
  140. var unknownComplexField = {type: "unknown"};
  141. function readFldChar(element) {
  142. var type = element.attributes["w:fldCharType"];
  143. if (type === "begin") {
  144. complexFieldStack.push(unknownComplexField);
  145. currentInstrText = [];
  146. } else if (type === "end") {
  147. complexFieldStack.pop();
  148. } else if (type === "separate") {
  149. var hyperlinkOptions = parseHyperlinkFieldCode(currentInstrText.join(''));
  150. var complexField = hyperlinkOptions === null ? unknownComplexField : {type: "hyperlink", options: hyperlinkOptions};
  151. complexFieldStack.pop();
  152. complexFieldStack.push(complexField);
  153. }
  154. return emptyResult();
  155. }
  156. function currentHyperlinkOptions() {
  157. var topHyperlink = _.last(complexFieldStack.filter(function(complexField) {
  158. return complexField.type === "hyperlink";
  159. }));
  160. return topHyperlink ? topHyperlink.options : null;
  161. }
  162. function parseHyperlinkFieldCode(code) {
  163. var externalLinkResult = /\s*HYPERLINK "(.*)"/.exec(code);
  164. if (externalLinkResult) {
  165. return {href: externalLinkResult[1]};
  166. }
  167. var internalLinkResult = /\s*HYPERLINK\s+\\l\s+"(.*)"/.exec(code);
  168. if (internalLinkResult) {
  169. return {anchor: internalLinkResult[1]};
  170. }
  171. return null;
  172. }
  173. function readInstrText(element) {
  174. currentInstrText.push(element.text());
  175. return emptyResult();
  176. }
  177. function readSymbol(element) {
  178. // See 17.3.3.30 sym (Symbol Character) of ECMA-376 4th edition Part 1
  179. var font = element.attributes["w:font"];
  180. var char = element.attributes["w:char"];
  181. var unicodeCharacter = dingbatToUnicode.hex(font, char);
  182. if (unicodeCharacter == null && /^F0..$/.test(char)) {
  183. unicodeCharacter = dingbatToUnicode.hex(font, char.substring(2));
  184. }
  185. if (unicodeCharacter == null) {
  186. return emptyResultWithMessages([warning(
  187. "A w:sym element with an unsupported character was ignored: char " + char + " in font " + font
  188. )]);
  189. } else {
  190. return elementResult(new documents.Text(unicodeCharacter.string));
  191. }
  192. }
  193. function noteReferenceReader(noteType) {
  194. return function(element) {
  195. var noteId = element.attributes["w:id"];
  196. return elementResult(new documents.NoteReference({
  197. noteType: noteType,
  198. noteId: noteId
  199. }));
  200. };
  201. }
  202. function readCommentReference(element) {
  203. return elementResult(documents.commentReference({
  204. commentId: element.attributes["w:id"]
  205. }));
  206. }
  207. function readChildElements(element) {
  208. return readXmlElements(element.children);
  209. }
  210. var xmlElementReaders = {
  211. "w:p": function(element) {
  212. var paragraphPropertiesElement = element.firstOrEmpty("w:pPr");
  213. var isDeleted = !!paragraphPropertiesElement
  214. .firstOrEmpty("w:rPr")
  215. .first("w:del");
  216. if (isDeleted) {
  217. element.children.forEach(function(child) {
  218. deletedParagraphContents.push(child);
  219. });
  220. return emptyResult();
  221. } else {
  222. var childrenXml = element.children;
  223. if (deletedParagraphContents.length > 0) {
  224. childrenXml = deletedParagraphContents.concat(childrenXml);
  225. deletedParagraphContents = [];
  226. }
  227. return ReadResult.map(
  228. readParagraphProperties(paragraphPropertiesElement),
  229. readXmlElements(childrenXml),
  230. function(properties, children) {
  231. return new documents.Paragraph(children, properties);
  232. }
  233. ).insertExtra();
  234. }
  235. },
  236. "w:r": function(element) {
  237. return ReadResult.map(
  238. readRunProperties(element.firstOrEmpty("w:rPr")),
  239. readXmlElements(element.children),
  240. function(properties, children) {
  241. var hyperlinkOptions = currentHyperlinkOptions();
  242. if (hyperlinkOptions !== null) {
  243. children = [new documents.Hyperlink(children, hyperlinkOptions)];
  244. }
  245. return new documents.Run(children, properties);
  246. }
  247. );
  248. },
  249. "w:fldChar": readFldChar,
  250. "w:instrText": readInstrText,
  251. "w:t": function(element) {
  252. return elementResult(new documents.Text(element.text()));
  253. },
  254. "w:tab": function(element) {
  255. return elementResult(new documents.Tab());
  256. },
  257. "w:noBreakHyphen": function() {
  258. return elementResult(new documents.Text("\u2011"));
  259. },
  260. "w:softHyphen": function(element) {
  261. return elementResult(new documents.Text("\u00AD"));
  262. },
  263. "w:sym": readSymbol,
  264. "w:hyperlink": function(element) {
  265. var relationshipId = element.attributes["r:id"];
  266. var anchor = element.attributes["w:anchor"];
  267. return readXmlElements(element.children).map(function(children) {
  268. function create(options) {
  269. var targetFrame = element.attributes["w:tgtFrame"] || null;
  270. return new documents.Hyperlink(
  271. children,
  272. _.extend({targetFrame: targetFrame}, options)
  273. );
  274. }
  275. if (relationshipId) {
  276. var href = relationships.findTargetByRelationshipId(relationshipId);
  277. if (anchor) {
  278. href = uris.replaceFragment(href, anchor);
  279. }
  280. return create({href: href});
  281. } else if (anchor) {
  282. return create({anchor: anchor});
  283. } else {
  284. return children;
  285. }
  286. });
  287. },
  288. "w:tbl": readTable,
  289. "w:tr": readTableRow,
  290. "w:tc": readTableCell,
  291. "w:footnoteReference": noteReferenceReader("footnote"),
  292. "w:endnoteReference": noteReferenceReader("endnote"),
  293. "w:commentReference": readCommentReference,
  294. "w:br": function(element) {
  295. var breakType = element.attributes["w:type"];
  296. if (breakType == null || breakType === "textWrapping") {
  297. return elementResult(documents.lineBreak);
  298. } else if (breakType === "page") {
  299. return elementResult(documents.pageBreak);
  300. } else if (breakType === "column") {
  301. return elementResult(documents.columnBreak);
  302. } else {
  303. return emptyResultWithMessages([warning("Unsupported break type: " + breakType)]);
  304. }
  305. },
  306. "w:bookmarkStart": function(element){
  307. var name = element.attributes["w:name"];
  308. if (name === "_GoBack") {
  309. return emptyResult();
  310. } else {
  311. return elementResult(new documents.BookmarkStart({name: name}));
  312. }
  313. },
  314. "mc:AlternateContent": function(element) {
  315. return readChildElements(element.first("mc:Fallback"));
  316. },
  317. "w:sdt": function(element) {
  318. return readXmlElements(element.firstOrEmpty("w:sdtContent").children);
  319. },
  320. "w:ins": readChildElements,
  321. "w:object": readChildElements,
  322. "w:smartTag": readChildElements,
  323. "w:drawing": readChildElements,
  324. "w:pict": function(element) {
  325. return readChildElements(element).toExtra();
  326. },
  327. "v:roundrect": readChildElements,
  328. "v:shape": readChildElements,
  329. "v:textbox": readChildElements,
  330. "w:txbxContent": readChildElements,
  331. "wp:inline": readDrawingElement,
  332. "wp:anchor": readDrawingElement,
  333. "v:imagedata": readImageData,
  334. "v:group": readChildElements,
  335. "v:rect": readChildElements
  336. };
  337. return {
  338. readXmlElement: readXmlElement,
  339. readXmlElements: readXmlElements
  340. };
  341. function readTable(element) {
  342. var propertiesResult = readTableProperties(element.firstOrEmpty("w:tblPr"));
  343. return readXmlElements(element.children)
  344. .flatMap(calculateRowSpans)
  345. .flatMap(function(children) {
  346. return propertiesResult.map(function(properties) {
  347. return documents.Table(children, properties);
  348. });
  349. });
  350. }
  351. function readTableProperties(element) {
  352. return readTableStyle(element).map(function(style) {
  353. return {
  354. styleId: style.styleId,
  355. styleName: style.name
  356. };
  357. });
  358. }
  359. function readTableRow(element) {
  360. var properties = element.firstOrEmpty("w:trPr");
  361. var isHeader = !!properties.first("w:tblHeader");
  362. return readXmlElements(element.children).map(function(children) {
  363. return documents.TableRow(children, {isHeader: isHeader});
  364. });
  365. }
  366. function readTableCell(element) {
  367. return readXmlElements(element.children).map(function(children) {
  368. var properties = element.firstOrEmpty("w:tcPr");
  369. var gridSpan = properties.firstOrEmpty("w:gridSpan").attributes["w:val"];
  370. var colSpan = gridSpan ? parseInt(gridSpan, 10) : 1;
  371. var cell = documents.TableCell(children, {colSpan: colSpan});
  372. cell._vMerge = readVMerge(properties);
  373. return cell;
  374. });
  375. }
  376. function readVMerge(properties) {
  377. var element = properties.first("w:vMerge");
  378. if (element) {
  379. var val = element.attributes["w:val"];
  380. return val === "continue" || !val;
  381. } else {
  382. return null;
  383. }
  384. }
  385. function calculateRowSpans(rows) {
  386. var unexpectedNonRows = _.any(rows, function(row) {
  387. return row.type !== documents.types.tableRow;
  388. });
  389. if (unexpectedNonRows) {
  390. return elementResultWithMessages(rows, [warning(
  391. "unexpected non-row element in table, cell merging may be incorrect"
  392. )]);
  393. }
  394. var unexpectedNonCells = _.any(rows, function(row) {
  395. return _.any(row.children, function(cell) {
  396. return cell.type !== documents.types.tableCell;
  397. });
  398. });
  399. if (unexpectedNonCells) {
  400. return elementResultWithMessages(rows, [warning(
  401. "unexpected non-cell element in table row, cell merging may be incorrect"
  402. )]);
  403. }
  404. var columns = {};
  405. rows.forEach(function(row) {
  406. var cellIndex = 0;
  407. row.children.forEach(function(cell) {
  408. if (cell._vMerge && columns[cellIndex]) {
  409. columns[cellIndex].rowSpan++;
  410. } else {
  411. columns[cellIndex] = cell;
  412. cell._vMerge = false;
  413. }
  414. cellIndex += cell.colSpan;
  415. });
  416. });
  417. rows.forEach(function(row) {
  418. row.children = row.children.filter(function(cell) {
  419. return !cell._vMerge;
  420. });
  421. row.children.forEach(function(cell) {
  422. delete cell._vMerge;
  423. });
  424. });
  425. return elementResult(rows);
  426. }
  427. function readDrawingElement(element) {
  428. var blips = element
  429. .getElementsByTagName("a:graphic")
  430. .getElementsByTagName("a:graphicData")
  431. .getElementsByTagName("pic:pic")
  432. .getElementsByTagName("pic:blipFill")
  433. .getElementsByTagName("a:blip");
  434. return combineResults(blips.map(readBlip.bind(null, element)));
  435. }
  436. function readBlip(element, blip) {
  437. var properties = element.first("wp:docPr").attributes;
  438. var altText = isBlank(properties.descr) ? properties.title : properties.descr;
  439. var blipImageFile = findBlipImageFile(blip);
  440. if (blipImageFile === null) {
  441. return emptyResultWithMessages([warning("Could not find image file for a:blip element")]);
  442. } else {
  443. return readImage(blipImageFile, altText);
  444. }
  445. }
  446. function isBlank(value) {
  447. return value == null || /^\s*$/.test(value);
  448. }
  449. function findBlipImageFile(blip) {
  450. var embedRelationshipId = blip.attributes["r:embed"];
  451. var linkRelationshipId = blip.attributes["r:link"];
  452. if (embedRelationshipId) {
  453. return findEmbeddedImageFile(embedRelationshipId);
  454. } else if (linkRelationshipId) {
  455. var imagePath = relationships.findTargetByRelationshipId(linkRelationshipId);
  456. return {
  457. path: imagePath,
  458. read: files.read.bind(files, imagePath)
  459. };
  460. } else {
  461. return null;
  462. }
  463. }
  464. function readImageData(element) {
  465. var relationshipId = element.attributes['r:id'];
  466. if (relationshipId) {
  467. return readImage(
  468. findEmbeddedImageFile(relationshipId),
  469. element.attributes["o:title"]);
  470. } else {
  471. return emptyResultWithMessages([warning("A v:imagedata element without a relationship ID was ignored")]);
  472. }
  473. }
  474. function findEmbeddedImageFile(relationshipId) {
  475. var path = uris.uriToZipEntryName("word", relationships.findTargetByRelationshipId(relationshipId));
  476. return {
  477. path: path,
  478. read: docxFile.read.bind(docxFile, path)
  479. };
  480. }
  481. function readImage(imageFile, altText) {
  482. var contentType = contentTypes.findContentType(imageFile.path);
  483. var image = documents.Image({
  484. readImage: imageFile.read,
  485. altText: altText,
  486. contentType: contentType
  487. });
  488. var warnings = supportedImageTypes[contentType] ?
  489. [] : warning("Image of type " + contentType + " is unlikely to display in web browsers");
  490. return elementResultWithMessages(image, warnings);
  491. }
  492. function undefinedStyleWarning(type, styleId) {
  493. return warning(
  494. type + " style with ID " + styleId + " was referenced but not defined in the document");
  495. }
  496. }
  497. function readNumberingProperties(styleId, element, numbering) {
  498. var level = element.firstOrEmpty("w:ilvl").attributes["w:val"];
  499. var numId = element.firstOrEmpty("w:numId").attributes["w:val"];
  500. if (level !== undefined && numId !== undefined) {
  501. return numbering.findLevel(numId, level);
  502. }
  503. if (styleId != null) {
  504. var levelByStyleId = numbering.findLevelByParagraphStyleId(styleId);
  505. if (levelByStyleId != null) {
  506. return levelByStyleId;
  507. }
  508. }
  509. return null;
  510. }
  511. var supportedImageTypes = {
  512. "image/png": true,
  513. "image/gif": true,
  514. "image/jpeg": true,
  515. "image/svg+xml": true,
  516. "image/tiff": true
  517. };
  518. var ignoreElements = {
  519. "office-word:wrap": true,
  520. "v:shadow": true,
  521. "v:shapetype": true,
  522. "w:annotationRef": true,
  523. "w:bookmarkEnd": true,
  524. "w:sectPr": true,
  525. "w:proofErr": true,
  526. "w:lastRenderedPageBreak": true,
  527. "w:commentRangeStart": true,
  528. "w:commentRangeEnd": true,
  529. "w:del": true,
  530. "w:footnoteRef": true,
  531. "w:endnoteRef": true,
  532. "w:pPr": true,
  533. "w:rPr": true,
  534. "w:tblPr": true,
  535. "w:tblGrid": true,
  536. "w:trPr": true,
  537. "w:tcPr": true
  538. };
  539. function emptyResultWithMessages(messages) {
  540. return new ReadResult(null, null, messages);
  541. }
  542. function emptyResult() {
  543. return new ReadResult(null);
  544. }
  545. function elementResult(element) {
  546. return new ReadResult(element);
  547. }
  548. function elementResultWithMessages(element, messages) {
  549. return new ReadResult(element, null, messages);
  550. }
  551. function ReadResult(element, extra, messages) {
  552. this.value = element || [];
  553. this.extra = extra || [];
  554. this._result = new Result({
  555. element: this.value,
  556. extra: extra
  557. }, messages);
  558. this.messages = this._result.messages;
  559. }
  560. ReadResult.prototype.toExtra = function() {
  561. return new ReadResult(null, joinElements(this.extra, this.value), this.messages);
  562. };
  563. ReadResult.prototype.insertExtra = function() {
  564. var extra = this.extra;
  565. if (extra && extra.length) {
  566. return new ReadResult(joinElements(this.value, extra), null, this.messages);
  567. } else {
  568. return this;
  569. }
  570. };
  571. ReadResult.prototype.map = function(func) {
  572. var result = this._result.map(function(value) {
  573. return func(value.element);
  574. });
  575. return new ReadResult(result.value, this.extra, result.messages);
  576. };
  577. ReadResult.prototype.flatMap = function(func) {
  578. var result = this._result.flatMap(function(value) {
  579. return func(value.element)._result;
  580. });
  581. return new ReadResult(result.value.element, joinElements(this.extra, result.value.extra), result.messages);
  582. };
  583. ReadResult.map = function(first, second, func) {
  584. return new ReadResult(
  585. func(first.value, second.value),
  586. joinElements(first.extra, second.extra),
  587. first.messages.concat(second.messages)
  588. );
  589. };
  590. function combineResults(results) {
  591. var result = Result.combine(_.pluck(results, "_result"));
  592. return new ReadResult(
  593. _.flatten(_.pluck(result.value, "element")),
  594. _.filter(_.flatten(_.pluck(result.value, "extra")), identity),
  595. result.messages
  596. );
  597. }
  598. function joinElements(first, second) {
  599. return _.flatten([first, second]);
  600. }
  601. function identity(value) {
  602. return value;
  603. }