#19 Identify links to tables and figures

src/docx2jats/jats/Figure.php

@@ -1,9 +1,9 @@
 <?php namespace docx2jats\jats;
 
 /**
-* @file src/docx2jats/jats/Table.php
+* @file src/docx2jats/jats/Figure.php
 *
-* Copyright (c) 2018-2019 Vitalii Bezsheiko
+* Copyright (c) 2018-2020 Vitalii Bezsheiko
 * Distributed under the GNU GPL v3.
 *
 * @brief represent JATS XML image
@@ -13,6 +13,8 @@
 use docx2jats\objectModel\body\Image as FigureObject;
 
 class Figure extends Element {
+	const JATS_FIGURE_ID_PREFIX = 'fig';
+
 	/* @var $dataObject FigureObject */
 	var $figureObject;
 
@@ -23,7 +25,11 @@ public function __construct(DataObject $dataObject) {
 	}
 
 	function setContent() {
-		$dataObject = $this->getDataObject(); /* @var $dataObject \docx2jats\objectModel\body\Table */
+		$dataObject = $this->getDataObject(); /* @var $dataObject \docx2jats\objectModel\body\Image */
+
+		if ($dataObject->getId()) {
+			$this->setAttribute('id', self::JATS_FIGURE_ID_PREFIX . $dataObject->getId());
+		}
 
 		if ($dataObject->getLabel()) {
 			$this->appendChild($this->ownerDocument->createElement('label', $dataObject->getLabel()));

src/docx2jats/jats/Par.php

@@ -3,7 +3,7 @@
 /**
  * @file src/docx2jats/jats/Row.php
  *
- * Copyright (c) 2018-2019 Vitalii Bezsheiko
+ * Copyright (c) 2018-2020 Vitalii Bezsheiko
  * Distributed under the GNU GPL v3.
  *
  * @brief represent JATS XML paragraph; can't be nested. To be included into body, sections, lists and table cells.
@@ -25,6 +25,7 @@ public function setContent() {
 
 		foreach ($this->getDataObject()->getContent() as $content) {
 			if (get_class($content) === 'docx2jats\objectModel\body\Field') {
+				// Write links to references
 				if ($content->getType() === Field::DOCX_FIELD_CSL) {
 					$lastKey = array_key_last($content->getRefIds());
 					foreach ($content->getRefIds() as $key => $id) {
@@ -38,6 +39,21 @@ public function setContent() {
 						}
 					}
 				}
+				// Write links to table and figures
+				elseif ($content->getType() === Field::DOCX_FIELD_BOOKMARK_REF) {
+					$refEl = $this->ownerDocument->createElement('xref');
+					$this->appendChild($refEl);
+					foreach ($content->getContent() as $text) { /* @var $text \docx2jats\objectModel\body\Text */
+						JatsText::extractText($text, $refEl);
+					}
+					if ($tableIdRef = $content->tableIdRef) {
+						$refEl->setAttribute('ref-type', 'table');
+						$refEl->setAttribute('rid', Table::JATS_TABLE_ID_PREFIX . $tableIdRef);
+					} elseif ($figureIdRef = $content->figureIdRef) {
+						$refEl->setAttribute('ref-type', 'fig');
+						$refEl->setAttribute('rid', Figure::JATS_FIGURE_ID_PREFIX . $figureIdRef);
+					}
+				}
 			} else {
 				JatsText::extractText($content, $this);
 			}

src/docx2jats/jats/Table.php

@@ -3,7 +3,7 @@
 /**
  * @file src/docx2jats/jats/Table.php
  *
- * Copyright (c) 2018-2019 Vitalii Bezsheiko
+ * Copyright (c) 2018-2020 Vitalii Bezsheiko
  * Distributed under the GNU GPL v3.
  *
  * @brief represent JATS XML table
@@ -14,13 +14,19 @@
 
 class Table extends Element {
 
+	const JATS_TABLE_ID_PREFIX = 'tbl';
+
 	public function __construct(DataObject $dataObject) {
 		parent::__construct($dataObject);
 	}
 
 	public function setContent() {
 		$dataObject = $this->getDataObject(); /* @var $dataObject \docx2jats\objectModel\body\Table */
 
+		if ($dataObject->getId()) {
+			$this->setAttribute('id', self::JATS_TABLE_ID_PREFIX . $dataObject->getId());
+		}
+
 		if ($dataObject->getLabel()) {
 			$this->appendChild($this->ownerDocument->createElement('label', $dataObject->getLabel()));
 		}

src/docx2jats/objectModel/Document.php

@@ -3,12 +3,13 @@
 /**
  * @file src/docx2jats/objectModel/Document.php
  *
- * Copyright (c) 2018-2019 Vitalii Bezsheiko
+ * Copyright (c) 2018-2020 Vitalii Bezsheiko
  * Distributed under the GNU GPL v3.
  *
  * @brief representation of an article; extracts all main elements from DOCX document.xml
  */
 
+use docx2jats\jats\Figure;
 use docx2jats\objectModel\DataObject;
 use docx2jats\objectModel\body\Par;
 use docx2jats\objectModel\body\Table;
@@ -43,6 +44,19 @@ class Document {
 	private $references = array();
 	private $refCount = 0;
 
+	// Set unique IDs for tables and figure in order of appearance
+	private $currentFigureId = 1;
+	private $currentTableId = 1;
+
+	/**
+	 * @var $parsHaveBookmarks array
+	 * @brief Key numbers of paragraphs that contain bookmarks inside the content
+	 * is used to speed up a search
+	 */
+	private $elsHaveBookmarks = array();
+	private $elsAreTables = array();
+	private $elsAreFigures = array();
+
 	public function __construct(array $params) {
 		if (array_key_exists("relationships", $params)) {
 			$this->relationships = $params["relationships"];
@@ -66,6 +80,7 @@ public function __construct(array $params) {
 		$content = array();
 		$unUsedCaption = null;
 		foreach ($childNodes as $key => $childNode) {
+			// Assign block elements, i.e., Figures, Tables, Paragraphs, depending on the context
 			switch ($childNode->nodeName) {
 				case "w:p":
 					// There can be multiple drawings inside a run and multiple elements inside a drawing
@@ -77,6 +92,10 @@ public function __construct(array $params) {
 							foreach ($imageNodes as $imageNode) {
 								$figure = new Image($imageNode, $this);
 								$content[] = $figure;
+								$this->elsAreFigures[] = count($content)-1;
+
+								// Set unique ID
+								$figure->setFigureId($this->currentFigureId++);
 
 								// Set caption if exists
 								if ($unUsedCaption) {
@@ -103,12 +122,19 @@ public function __construct(array $params) {
 						} else {
 							$content[] = $par;
 						}
+
+						if ($par->hasBookmarks) {
+							$this->elsHaveBookmarks[] = count($content)-1;
+						}
 					}
 					break;
 				case "w:tbl":
 					$table = new Table($childNode, $this);
 					$content[] = $table;
+					$this->elsAreTables[] = count($content)-1;
 
+					// Set unique ID
+					$table->setTableId($this->currentTableId++);
 					// Set caption if exists
 					if ($unUsedCaption) {
 						$table->setCaption($unUsedCaption);
@@ -120,6 +146,7 @@ public function __construct(array $params) {
 
 		$this->content = $this->addSectionMarks($content);
 		self::$minimalHeadingLevel = $this->minimalHeadingLevel();
+		$this->setInternalRefs();
 	}
 
 	/**
@@ -286,4 +313,58 @@ public function getLastReference() : ?Reference {
 		$lastId = array_key_last($this->references);
 		return $this->references[$lastId];
 	}
+
+	/**
+	 * @brief iterate through the content and establish internal links between element
+	 * elsHaveBookmarks holds position in an array of each paragraph that includes a bookmark
+	 * it's slightly faster than looping over the whole content
+	 */
+	private function setInternalRefs(): void {
+		if (empty($this->elsHaveBookmarks)) return;
+
+		// Find and map tables' and figures' bookmarks
+		$refTableMap = $this->getBookmarkCaptionMapping($this->elsAreTables);
+		$refFigureMap = $this->getBookmarkCaptionMapping($this->elsAreFigures);
+
+		// Find bookmark refs
+		foreach ($this->elsHaveBookmarks as $parKeyWithBookmark) {
+			$par = $this->getContent()[$parKeyWithBookmark]; /* @var $par Par */
+			foreach ($par->bookmarkPos as $fieldKeyWithBookmark) {
+				$field = $par->getContent()[$fieldKeyWithBookmark]; /* @var $field \docx2jats\objectModel\body\Field */
+
+				// Set links to tables
+				foreach ($refTableMap as $tableId => $tableRefs) {
+					if (in_array($field->getBookmarkId(), $tableRefs)) {
+						$field->tableIdRef = $tableId;
+					}
+				}
+
+				// Set links to Figures
+				foreach ($refFigureMap as $figureId => $figureRefs) {
+					if (in_array($field->getBookmarkId(), $figureRefs)) {
+						$field->figureIdRef = $tableId;
+					}
+				}
+			}
+		}
+	}
+
+	/**
+	 * @return array
+	 * @brief (or not so brief) Map OOXML bookmark refs inside table and figures with correspondent table/figure IDs.
+	 * In OOXML those bookmarks are stored inside captions
+	 * This is used to set right link to these objects from the text
+	 * Keep in mind that bookmarks also may be stored in an external file, e.g., Mendeley plugin for LibreOffice Writer
+	 * stores links to references this way
+	 */
+	function getBookmarkCaptionMapping(array $keysInContent): array {
+		$refMap = [];
+		foreach ($keysInContent as $tableKey) {
+			$table = $this->content[$tableKey]; /* @var $table Table|Image */
+			if (empty($table->getBookmarkIds())) continue;
+			$refMap[$table->getId()] = $table->getBookmarkIds();
+		}
+
+		return $refMap;
+	}
 }

src/docx2jats/objectModel/body/Field.php

@@ -14,7 +14,8 @@
 
 class Field extends DataObject {
 
-	const DOCX_FIELD_CSL = 1;
+	const DOCX_FIELD_CSL = 1; // Zotero/Mendeley JSON-CSL reference
+	const DOCX_FIELD_BOOKMARK_REF = 2; // Internal OOXML reference to the bookmark
 	/** @var $type int DOCX_FIELD... const */
 	private $type = 0;
 	private $isZoteroCSL = false;
@@ -25,6 +26,20 @@ class Field extends DataObject {
 	private $rawRuns = array();
 	/** @var array contains instructions to be processed as strings, e.g., CSL citations as a JSON string */
 	private $instructions = array();
+	private $bookmarkId;
+
+	/**
+	 * @var $tableIdRef int
+	 * @brief the reference to table that the field contains
+	 * TODO check if may include several references at once
+	 */
+	public $tableIdRef = 0;
+
+	/**
+	 * @var $figureIdRef int
+	 * @brief the reference to figure that the field contains
+	 */
+	public $figureIdRef = 0;
 
 	public function __construct(\DOMElement $domElement, Document $ownerDocument, Par $parent) {
 		parent::__construct($domElement, $ownerDocument, $parent);
@@ -77,6 +92,7 @@ private function processRuns() {
 				if ($instructionNode) {
 					$instructionString = $instructionNode->nodeValue;
 					$this->instructions[] = $instructionString;
+					// Check if Zotero/Mendeley Citation
 					if (strpos($instructionString, 'CSL_CITATION') !== false) {
 						$this->type = self::DOCX_FIELD_CSL;
 						$rawCSL = $this->extractRawCSL($instructionString);
@@ -92,6 +108,12 @@ private function processRuns() {
 							}
 						}
 					}
+					// Check if Link to the Bookmark (only Tables and Figures are supported)
+					elseif (strpos($instructionString, 'REF') !== false) {
+						$this->getParent()->hasBookmarks = true;
+						$this->type = self::DOCX_FIELD_BOOKMARK_REF;
+						$this->bookmarkId = $this->extractRefID($instructionString);
+					}
 				}
 			} else {
 				$this->content[] = new Text($run);
@@ -119,6 +141,18 @@ private function extractRawCSL(string $instruction) {
 		return $rawCSL;
 	}
 
+	private function extractRefID(string $instruction) {
+		$exploded = explode(' ', trim($instruction));
+		foreach ($exploded as $key => $word) {
+			if ($word == 'REF') {
+				if (array_key_exists($key+1, $exploded)) {
+					return $exploded[$key + 1];
+				}
+			}
+		}
+		return null;
+	}
+
 	public function getPlainCit() {
 		return $this->plainCit;
 	}
@@ -133,4 +167,15 @@ public function getRefIds() {
 	public function isZoteroCSL(): bool {
 		return $this->isZoteroCSL;
 	}
+
+	/**
+	 * @return mixed
+	 */
+	public function getBookmarkId() {
+		if ($this->type === self::DOCX_FIELD_BOOKMARK_REF) {
+			return $this->bookmarkId;
+		}
+
+		return null;
+	}
 }

src/docx2jats/objectModel/body/Image.php

@@ -3,7 +3,7 @@
 /**
 * @file src/docx2jats/objectModel/body/Image.php
 *
-* Copyright (c) 2018-2019 Vitalii Bezsheiko
+* Copyright (c) 2018-2020 Vitalii Bezsheiko
 * Distributed under the GNU GPL v3.
 *
 * @brief parses data from OOXML drawings; supports only pictures
@@ -12,12 +12,16 @@
 use docx2jats\objectModel\DataObject;
 use docx2jats\objectModel\Document;
 
+// TODO create a common parent class for Image and Table
 class Image extends DataObject {
 
 	/* @var $link string */
 	private $link;
-	private ?string $label = null;
-	private ?string $title = null;
+	private $label = null;
+	private $title = null;
+	private $figureId = 0;
+	private $bookmarkIds = array();
+	private $bookmarkText = ''; // TODO Check if there are situation where bookmark text is needed for JATS
 
 	public function __construct(\DOMElement $domElement, $ownerDocument) {
 		parent::__construct($domElement, $ownerDocument);
@@ -81,6 +85,15 @@ public function setCaption(\DOMElement $el): void {
 		if (!empty($title)) {
 			$this->title = trim($title);
 		}
+
+		// Caption may have bookmarks that are pointed from outside the table, retrieve their IDs;
+		// TODO Check if other bookmark types may be inserted in captions
+		$bookmarkStartEls = Document::$xpath->query('w:bookmarkStart', $el);
+		foreach ($bookmarkStartEls as $bookmarkStartEl) { /* @var $bookmarkStartEl \DOMElement */
+			if ($bookmarkStartEl->hasAttribute('w:name')) {
+				$this->bookmarkIds[] = $bookmarkStartEl->getAttribute('w:name');
+			}
+		}
 	}
 
 	/**
@@ -96,4 +109,25 @@ public function getLabel(): ?string {
 	public function getTitle(): ?string {
 		return $this->title;
 	}
+
+	/**
+	 * @param int $currentFigureId
+	 */
+	public function setFigureId(int $currentFigureId): void {
+		$this->figureId = $currentFigureId;
+	}
+
+	/**
+	 * @return int
+	 */
+	public function getId(): int {
+		return $this->figureId;
+	}
+
+	/**
+	 * @return array
+	 */
+	public function getBookmarkIds(): array {
+		return $this->bookmarkIds;
+	}
 }