* It also includes caching mechanism for XPathAPI instances to improve performance.
 * </p>
 * <p>
 * The extracted text is obtained from the nodes selected by the {@code targetNodePath} XPath expression.
 * The default value for {@code targetNodePath} is "//HTML/BODY | //@alt | //@title", which selects the body of the HTML document,
 * as well as the alt and title attributes.
 * </p>
 * <p>

     * Extracts URLs from HTML tag attributes using XPath.
     *
     * @param url the base URL for resolving relative URLs
     * @param document the document to extract URLs from
     * @param xpath the XPath expression to select elements
     * @param attr the attribute name to extract URLs from
     * @param encoding the character encoding to use
     * @return a list of extracted URLs
     */

                } catch (final IOException e) {
                    logger.warn("Failed to delete {}.", file.getAbsolutePath(), e);
                }
            }
        }

    }

    /**
     * Detects the MIME type of the content.
     *
     * @param out the deferred file output stream
     * @return the detected MIME type
     * @throws IOException if an I/O error occurs
     */

            logger.debug("metadata: {}", metadata);
        }

        return metadata;
    }

    // workaround: Tika does not have extention points.
    /**
     * This class is a parser that detects the document type.
     */
    protected class TikaDetectParser extends CompositeParser {
        private static final long serialVersionUID = 1L;

        /**