#18 More reliable way to handle DOCX archive's content

src/docx2jats/DOCXArchive.php

@@ -13,6 +13,19 @@
 
 class DOCXArchive extends \ZipArchive {
 
+	public const CONTENT_TYPES_PATH = '[Content_Types].xml';
+	public const CONTENT_TYPE_DOCUMENT_MAIN = 'application/vnd.openxmlformats-officedocument.wordprocessingml.document.main+xml';
+	public const CONTENT_TYPE_STYLES = 'application/vnd.openxmlformats-officedocument.wordprocessingml.styles+xml';
+	public const CONTENT_TYPE_SETTINGS = 'application/vnd.openxmlformats-officedocument.wordprocessingml.settings+xml';
+	public const CONTENT_TYPE_NUMBERING = 'application/vnd.openxmlformats-officedocument.wordprocessingml.numbering+xml';
+	public const CONTENT_TYPE_CUSTOM_PROP = 'application/vnd.openxmlformats-officedocument.custom-properties+xml';
+	public const CONTENT_TYPE_RELATIONSHIPS = 'application/vnd.openxmlformats-package.relationships+xml';
+
+	/* @var $contentType \DOMDocument */
+	private $contentType;
+
+	static $contentTypeXpath;
+
 	/* @var $filePath string */
 	private $filePath;
 
@@ -29,20 +42,39 @@ public function __construct(string $filepath) {
 		$this->filePath = $filepath;
 
 		if ($this->open($filepath)) {
-			$this->ooxmlDocument = $this->transformToXml("word/document.xml");
-			$relationships = $this->transformToXml("word/_rels/document.xml.rels");
-			$styles = $this->transformToXml("word/styles.xml");
+			$this->contentType = $this->transformToXml(self::CONTENT_TYPES_PATH);
+			self::$contentTypeXpath = new \DOMXPath($this->contentType);
+
+			// Set the Main Document Part
+			$ooxmlDocumentPath = $this->getRealFileDocumentPath('word/document.xml', self::CONTENT_TYPE_DOCUMENT_MAIN);
+			$this->ooxmlDocument = $this->transformToXml($ooxmlDocumentPath);
+
+			// Relationships of the Main Document Part
+			$partRelationshipsPath = $this->getRealFileDocumentPath('word/_rels/document.xml.rels', self::CONTENT_TYPE_RELATIONSHIPS, $ooxmlDocumentPath);
+			$partRelationships = $this->transformToXml($partRelationshipsPath);
+
+			// Style names used in the document, styles should be checked recursively, see docx2jats\objectModel\Document::getBuiltinStyle
+			$stylePath = $this->getRealFileDocumentPath('word/styles.xml', self::CONTENT_TYPE_STYLES);
+			$styles = $this->transformToXml($stylePath);
+
+			// Media files, e.g., images
 			$this->mediaFiles = $this->extractMediaFiles();
-			$numbering = $this->transformToXml("word/numbering.xml");
-			$docPropsCustom = $this->transformToXml("docProps/custom.xml");
+
+			// Description of all numbered content, e.g., lists
+			$numberingPath = $this->getRealFileDocumentPath('word/numbering.xml', self::CONTENT_TYPE_NUMBERING);
+			$numbering = $this->transformToXml($numberingPath);
+
+			// Custom Document properties, this is used by Mendeley plugin export from LibreOffice Writer
+			$docPropsCustom = $this->getRealFileDocumentPath('docProps/custom.xml', self::CONTENT_TYPE_CUSTOM_PROP);
+			$docPropsCustom = $this->transformToXml($docPropsCustom);
 			$this->close();
 
 			// construct as an array
 			$params = array();
 
 			$params["ooxmlDocument"] = $this->ooxmlDocument;
 
-			if ($relationships) $params["relationships"] = $relationships;
+			if ($partRelationships) $params["partRelationships"] = $partRelationships;
 			if ($styles) $params["styles"] = $styles;
 			if ($numbering) $params["numbering"] = $numbering;
 			if ($docPropsCustom) $params["docPropsCustom"] = $docPropsCustom;
@@ -63,7 +95,7 @@ public function getDocument(): Document {
 
 	private function transformToXml(string $path): ?\DOMDocument {
 		$index = $this->locateName($path);
-		if (!$index) return null;
+		if ($index === false) return null;
 		$data = $this->getFromIndex($index);
 		$xml = new \DOMDocument();
 		$xml->loadXML($data, LIBXML_NOENT | LIBXML_XINCLUDE | LIBXML_NOERROR | LIBXML_NOWARNING);
@@ -137,4 +169,58 @@ public function getMediaFilesContent(): array {
 
 		return $filesContent;
 	}
+
+	private function getRealFileDocumentPath(string $defaultPath, string $contentType = null, string $parentPath = null): string {
+		$path = null;
+		if (!is_null($contentType)) {
+			foreach ($this->contentType->getElementsByTagName('Override') as $override) {
+				if ($override->hasAttribute('PartName') &&
+					$override->hasAttribute('ContentType') &&
+					$override->getAttribute('ContentType') == $contentType) {
+					if ($contentType !== self::CONTENT_TYPE_RELATIONSHIPS) {
+						$path = $override->getAttribute('PartName');
+						break;
+					} else {
+						// Find the file associated with relationships, compare by filename
+						$partName = $override->getAttribute('PartName');
+						if (strpos(pathinfo($partName)['basename'], pathinfo($parentPath)['basename']) !== false) {
+							$path = $partName;
+							break;
+						}
+					}
+				}
+			}
+
+			// MS Word may not specify the path to the relationships files trying to guess based on the parent path
+			if ($contentType === self::CONTENT_TYPE_RELATIONSHIPS && is_null($path)) {
+				$path = 'word/_rels/' . pathinfo($parentPath)['basename'] . '.rels';
+			}
+		}
+
+		if (is_null($path)) {
+			$path = $defaultPath;
+		}
+
+		$path = ltrim($path, '/');
+
+		try {
+			$this->findDocumentByPath($path);
+		} catch (\Exception $e) {
+			trigger_error($e->getMessage(), E_USER_ERROR);
+		}
+
+		return $path;
+	}
+
+	/**
+	 * @param string $path
+	 * @return \DOMDocument
+	 * @throws \Exception if the document inside the archive isn't found
+	 */
+	private function findDocumentByPath(string $path): void {
+		$domDocument = $this->transformToXml($path);
+		if (!$domDocument) {
+			throw new \Exception('Cannot find document inside the archive by the path ' . $path);
+		}
+	}
 }

src/docx2jats/objectModel/Document.php

@@ -70,8 +70,8 @@ class Document {
 	public $bookMarks = array();
 
 	public function __construct(array $params) {
-		if (array_key_exists("relationships", $params)) {
-			$this->relationships = $params["relationships"];
+		if (array_key_exists("partRelationships", $params)) {
+			$this->relationships = $params["partRelationships"];
 			self::$relationshipsXpath = new \DOMXPath($this->relationships);
 		}