/*
 * Hibernate, Relational Persistence for Idiomatic Java
 *
 * Copyright (c) 2013, Red Hat Inc. or third-party contributors as
 * indicated by the @author tags or express copyright attribution
 * statements applied by the authors.  All third-party contributions are
 * distributed under license by Red Hat Inc.
 *
 * This copyrighted material is made available to anyone wishing to use, modify,
 * copy, or redistribute it subject to the terms and conditions of the GNU
 * Lesser General Public License, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
 * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public License
 * for more details.
 *
 * You should have received a copy of the GNU Lesser General Public License
 * along with this distribution; if not, write to:
 * Free Software Foundation, Inc.
 * 51 Franklin Street, Fifth Floor
 * Boston, MA  02110-1301  USA
 */
package org.hibernate.jpa.boot.archive.internal;

import java.io.File;
import java.io.IOException;
import java.io.InputStream;
import java.net.MalformedURLException;
import java.net.URL;
import java.util.LinkedList;
import java.util.List;

import org.hibernate.jpa.boot.archive.spi.ArchiveException;

import org.jboss.logging.Logger;

Helper for dealing with archives
Author:Emmanuel Bernard, Steve Ebersole
/** * Helper for dealing with archives * * @author Emmanuel Bernard * @author Steve Ebersole */
public class ArchiveHelper { private static final Logger log = Logger.getLogger( ArchiveHelper.class );
Get the JAR URL of the JAR containing the given entry Method used in a non managed environment
Params:
  • url – URL pointing to the known file in the JAR
  • entry – file known to be in the JAR
Throws:
Returns:the JAR URL
/** * Get the JAR URL of the JAR containing the given entry * Method used in a non managed environment * * @param url URL pointing to the known file in the JAR * @param entry file known to be in the JAR * @return the JAR URL * @throws IllegalArgumentException if none URL is found */
public static URL getJarURLFromURLEntry(URL url, String entry) throws IllegalArgumentException { URL jarUrl; String file = url.getFile(); if ( ! entry.startsWith( "/" ) ) { entry = "/" + entry; } file = file.substring( 0, file.length() - entry.length() ); if ( file.endsWith( "!" ) ) { file = file.substring( 0, file.length() - 1 ); } try { final String protocol = url.getProtocol(); if ( "jar".equals( protocol ) || "wsjar".equals( protocol ) ) { //Original URL is like jar:protocol //WebSphere has it's own way jarUrl = new URL( file ); if ( "file".equals( jarUrl.getProtocol() ) ) { if ( file.indexOf( ' ' ) != -1 ) { //not escaped, need to voodoo; goes by toURI to escape the path jarUrl = new File( jarUrl.getFile() ).toURI().toURL(); } } } else if ( "zip".equals( protocol ) //OC4J prevent ejb.jar access (ie everything without path) || "code-source".equals( url.getProtocol() ) //if no wrapping is done || "file".equals( protocol ) ) { //we have extracted the zip file, so it should be read as a file if ( file.indexOf( ' ' ) != -1 ) { //not escaped, need to voodoo; goes by toURI to escape the path jarUrl = new File(file).toURI().toURL(); } else { jarUrl = new File(file).toURL(); } } else { try { //We reconstruct the URL probably to make it work in some specific environments //Forgot the exact details, sorry (and the Git history does not help) jarUrl = new URL( protocol, url.getHost(), url.getPort(), file ); } //HHH-6442: Arquilian catch ( final MalformedURLException e ) { //Just use the provided URL as-is, likely it has a URLStreamHandler //associated w/ the instance jarUrl = url; } } } catch (MalformedURLException e) { throw new IllegalArgumentException( "Unable to determine JAR Url from " + url + ". Cause: " + e.getMessage() ); } log.trace( "JAR URL from URL Entry: " + url + " >> " + jarUrl ); return jarUrl; }
get the URL from a given path string
Params:
  • jarPath – The path that represents a URL
Throws:
Returns:The resolved URL reference
/** * get the URL from a given path string * * @param jarPath The path that represents a URL * * @return The resolved URL reference * * @throws IllegalArgumentException is something goes wrong */
public static URL getURLFromPath(String jarPath) { URL jarUrl; try { //is it an url jarUrl = new URL( jarPath ); } catch ( MalformedURLException e) { try { //consider it as a file path jarUrl = new URL( "file:" + jarPath ); } catch (MalformedURLException ee) { throw new IllegalArgumentException( "Unable to find jar:" + jarPath, ee ); } } return jarUrl; }
Extracts the bytes out of an InputStream. This form is the same as getBytesFromInputStream except that any IOException are wrapped as (runtime) ArchiveException
Params:
  • inputStream – The stream from which to extract bytes.
Throws:
Returns:The bytes
/** * Extracts the bytes out of an InputStream. This form is the same as {@link #getBytesFromInputStream} * except that any {@link IOException} are wrapped as (runtime) {@link ArchiveException} * * @param inputStream The stream from which to extract bytes. * * @return The bytes * * @throws ArchiveException Indicates a problem accessing the stream */
public static byte[] getBytesFromInputStreamSafely(InputStream inputStream) throws ArchiveException { try { return getBytesFromInputStream( inputStream ); } catch (IOException e) { throw new ArchiveException( "Unable to extract bytes from InputStream", e ); } }
Extracts the bytes out of an InputStream.
Params:
  • inputStream – The stream from which to extract bytes.
Throws:
  • IOException – Indicates a problem accessing the stream
See Also:
Returns:The bytes
/** * Extracts the bytes out of an InputStream. * * @param inputStream The stream from which to extract bytes. * * @return The bytes * * @throws IOException Indicates a problem accessing the stream * * @see #getBytesFromInputStreamSafely(java.io.InputStream) */
public static byte[] getBytesFromInputStream(InputStream inputStream) throws IOException { // Optimized by HHH-7835 int size; final List<byte[]> data = new LinkedList<byte[]>(); final int bufferSize = 4096; byte[] tmpByte = new byte[bufferSize]; int offset = 0; int total = 0; for ( ;; ) { size = inputStream.read( tmpByte, offset, bufferSize - offset ); if ( size == -1 ) { break; } offset += size; if ( offset == tmpByte.length ) { data.add( tmpByte ); tmpByte = new byte[bufferSize]; offset = 0; total += tmpByte.length; } } final byte[] result = new byte[total + offset]; int count = 0; for ( byte[] arr : data ) { System.arraycopy( arr, 0, result, count * arr.length, arr.length ); count++; } System.arraycopy( tmpByte, 0, result, count * tmpByte.length, offset ); return result; } private ArchiveHelper() { } }