
com.whaleal.icefrog.extra.compress.extractor.StreamExtractor Maven / Gradle / Ivy
package com.whaleal.icefrog.extra.compress.extractor;
import com.whaleal.icefrog.core.io.FileUtil;
import com.whaleal.icefrog.core.io.IORuntimeException;
import com.whaleal.icefrog.core.io.IoUtil;
import com.whaleal.icefrog.core.lang.Filter;
import com.whaleal.icefrog.core.lang.Precondition;
import com.whaleal.icefrog.core.util.StrUtil;
import com.whaleal.icefrog.extra.compress.CompressException;
import org.apache.commons.compress.archivers.ArchiveEntry;
import org.apache.commons.compress.archivers.ArchiveException;
import org.apache.commons.compress.archivers.ArchiveInputStream;
import org.apache.commons.compress.archivers.ArchiveStreamFactory;
import java.io.File;
import java.io.IOException;
import java.io.InputStream;
import java.nio.charset.Charset;
/**
* 数据解压器,即将归档打包的数据释放
*
* @author looly
*/
public class StreamExtractor implements Extractor {
private final ArchiveInputStream in;
/**
* 构造
*
* @param charset 编码
* @param file 包文件
*/
public StreamExtractor( Charset charset, File file ) {
this(charset, null, file);
}
/**
* 构造
*
* @param charset 编码
* @param archiverName 归档包格式,null表示自动检测
* @param file 包文件
*/
public StreamExtractor( Charset charset, String archiverName, File file ) {
this(charset, archiverName, FileUtil.getInputStream(file));
}
/**
* 构造
*
* @param charset 编码
* @param in 包流
*/
public StreamExtractor( Charset charset, InputStream in ) {
this(charset, null, in);
}
/**
* 构造
*
* @param charset 编码
* @param archiverName 归档包格式,null表示自动检测
* @param in 包流
*/
public StreamExtractor( Charset charset, String archiverName, InputStream in ) {
final ArchiveStreamFactory factory = new ArchiveStreamFactory(charset.name());
try {
in = IoUtil.toBuffered(in);
if (StrUtil.isBlank(archiverName)) {
this.in = factory.createArchiveInputStream(in);
} else {
this.in = factory.createArchiveInputStream(archiverName, in);
}
} catch (ArchiveException e) {
throw new CompressException(e);
}
}
/**
* 释放(解压)到指定目录,结束后自动关闭流,此方法只能调用一次
*
* @param targetDir 目标目录
* @param filter 解压文件过滤器,用于指定需要释放的文件,null表示不过滤。当{@link Filter#accept(Object)}为true时释放。
*/
@Override
public void extract( File targetDir, Filter filter ) {
try {
extractInternal(targetDir, filter);
} catch (IOException e) {
throw new IORuntimeException(e);
} finally {
close();
}
}
/**
* 释放(解压)到指定目录
*
* @param targetDir 目标目录
* @param filter 解压文件过滤器,用于指定需要释放的文件,null表示不过滤。当{@link Filter#accept(Object)}为true时释放。
* @throws IOException IO异常
*/
private void extractInternal( File targetDir, Filter filter ) throws IOException {
Precondition.isTrue(null != targetDir && ((false == targetDir.exists()) || targetDir.isDirectory()), "target must be dir.");
final ArchiveInputStream in = this.in;
ArchiveEntry entry;
File outItemFile;
while (null != (entry = in.getNextEntry())) {
if (null != filter && false == filter.accept(entry)) {
continue;
}
if (false == in.canReadEntryData(entry)) {
// 无法读取的文件直接跳过
continue;
}
outItemFile = FileUtil.file(targetDir, entry.getName());
if (entry.isDirectory()) {
// 创建对应目录
//noinspection ResultOfMethodCallIgnored
outItemFile.mkdirs();
} else {
FileUtil.writeFromStream(in, outItemFile, false);
}
}
}
@Override
public void close() {
IoUtil.close(this.in);
}
}
© 2015 - 2025 Weber Informatics LLC | Privacy Policy