Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance. Project price only 1 $
You can buy this project and download/modify it how often you want.
/*
* Copyright 2012 The Netty Project
*
* The Netty Project licenses this file to you under the Apache License,
* version 2.0 (the "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at:
*
* https://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
* License for the specific language governing permissions and limitations
* under the License.
*/
package io.netty.handler.codec.http.multipart;
import com.github.netty.protocol.servlet.util.ServletUtil;
import io.netty.buffer.ByteBuf;
import io.netty.buffer.Unpooled;
import io.netty.handler.codec.http.HttpConstants;
import io.netty.handler.codec.http.HttpContent;
import io.netty.handler.codec.http.HttpRequest;
import io.netty.handler.codec.http.LastHttpContent;
import io.netty.handler.codec.http.multipart.HttpPostBodyUtil.SeekAheadOptimize;
import io.netty.handler.codec.http.multipart.HttpPostRequestDecoder.EndOfDataDecoderException;
import io.netty.handler.codec.http.multipart.HttpPostRequestDecoder.ErrorDataDecoderException;
import io.netty.handler.codec.http.multipart.HttpPostRequestDecoder.MultiPartStatus;
import io.netty.handler.codec.http.multipart.HttpPostRequestDecoder.NotEnoughDataDecoderException;
import io.netty.util.ByteProcessor;
import io.netty.util.internal.PlatformDependent;
import io.netty.util.internal.StringUtil;
import java.io.IOException;
import java.nio.charset.Charset;
import java.util.ArrayList;
import java.util.List;
import java.util.Map;
import java.util.TreeMap;
import static io.netty.util.internal.ObjectUtil.checkPositiveOrZero;
/**
* This decoder will decode Body and can handle POST BODY.
*
* You MUST call {@link #destroy()} after completion to release all resources.
*
*/
public class CompatibleHttpPostStandardRequestDecoder implements InterfaceHttpPostRequestDecoder {
private static final UrlEncodedDetector urlEncodedDetector = new UrlEncodedDetector();
/**
* Factory used to create InterfaceHttpData
*/
private final HttpDataFactory factory;
/**
* Request to decode
*/
private final HttpRequest request;
/**
* Default charset to use
*/
private final Charset charset;
/**
* Does the last chunk already received
*/
private boolean isLastChunk;
/**
* HttpDatas from Body
*/
private final List bodyListHttpData = new ArrayList();
/**
* HttpDatas as Map from Body
*/
private final Map> bodyMapHttpData = new TreeMap>(
CaseIgnoringComparator.INSTANCE);
/**
* The current channelBuffer
*/
private ByteBuf undecodedChunk;
/**
* Body HttpDatas current position
*/
private int bodyListHttpDataRank;
/**
* Current getStatus
*/
private MultiPartStatus currentStatus = MultiPartStatus.NOTSTARTED;
/**
* The current Attribute that is currently in decode process
*/
private Attribute currentAttribute;
private boolean destroyed;
private int discardThreshold = HttpPostRequestDecoder.DEFAULT_DISCARD_THRESHOLD;
/**
*
* @param request
* the request to decode
* @throws NullPointerException
* for request
* @throws ErrorDataDecoderException
* if the default charset was wrong when decoding or other
* errors
*/
public CompatibleHttpPostStandardRequestDecoder(HttpRequest request) {
this(new DefaultHttpDataFactory(DefaultHttpDataFactory.MINSIZE), request, HttpConstants.DEFAULT_CHARSET);
}
/**
*
* @param factory
* the factory used to create InterfaceHttpData
* @param request
* the request to decode
* @throws NullPointerException
* for request or factory
* @throws ErrorDataDecoderException
* if the default charset was wrong when decoding or other
* errors
*/
public CompatibleHttpPostStandardRequestDecoder(HttpDataFactory factory, HttpRequest request) {
this(factory, request, HttpConstants.DEFAULT_CHARSET);
}
/**
*
* @param factory
* the factory used to create InterfaceHttpData
* @param request
* the request to decode
* @param charset
* the charset to use as default
* @throws NullPointerException
* for request or charset or factory
* @throws ErrorDataDecoderException
* if the default charset was wrong when decoding or other
* errors
*/
public CompatibleHttpPostStandardRequestDecoder(HttpDataFactory factory, HttpRequest request, Charset charset) {
this.request = request;
this.charset = charset;
this.factory = factory;
try {
if (request instanceof HttpContent) {
// Offer automatically if the given request is as type of HttpContent
// See #1089
offer((HttpContent) request);
} else {
parseBody();
}
} catch (Throwable e) {
destroy();
PlatformDependent.throwException(e);
}
}
private void checkDestroyed() {
if (destroyed) {
throw new IllegalStateException("CompatibleHttpPostStandardRequestDecoder was destroyed already");
}
}
/**
* True if this request is a Multipart request
*
* @return True if this request is a Multipart request
*/
@Override
public boolean isMultipart() {
checkDestroyed();
return false;
}
/**
* Set the amount of bytes after which read bytes in the buffer should be discarded.
* Setting this lower gives lower memory usage but with the overhead of more memory copies.
* Use {@code 0} to disable it.
*/
@Override
public void setDiscardThreshold(int discardThreshold) {
this.discardThreshold = checkPositiveOrZero(discardThreshold, "discardThreshold");
}
/**
* Return the threshold in bytes after which read data in the buffer should be discarded.
*/
@Override
public int getDiscardThreshold() {
return discardThreshold;
}
/**
* This getMethod returns a List of all HttpDatas from body.
*
* If chunked, all chunks must have been offered using offer() getMethod. If
* not, NotEnoughDataDecoderException will be raised.
*
* @return the list of HttpDatas from Body part for POST getMethod
* @throws NotEnoughDataDecoderException
* Need more chunks
*/
@Override
public List getBodyHttpDatas() {
checkDestroyed();
if (!isLastChunk) {
throw new NotEnoughDataDecoderException();
}
return bodyListHttpData;
}
/**
* This getMethod returns a List of all HttpDatas with the given name from
* body.
*
* If chunked, all chunks must have been offered using offer() getMethod. If
* not, NotEnoughDataDecoderException will be raised.
*
* @return All Body HttpDatas with the given name (ignore case)
* @throws NotEnoughDataDecoderException
* need more chunks
*/
@Override
public List getBodyHttpDatas(String name) {
checkDestroyed();
if (!isLastChunk) {
throw new NotEnoughDataDecoderException();
}
return bodyMapHttpData.get(name);
}
/**
* This getMethod returns the first InterfaceHttpData with the given name from
* body.
*
* If chunked, all chunks must have been offered using offer() getMethod. If
* not, NotEnoughDataDecoderException will be raised.
*
* @return The first Body InterfaceHttpData with the given name (ignore
* case)
* @throws NotEnoughDataDecoderException
* need more chunks
*/
@Override
public InterfaceHttpData getBodyHttpData(String name) {
checkDestroyed();
if (!isLastChunk) {
throw new NotEnoughDataDecoderException();
}
List list = bodyMapHttpData.get(name);
if (list != null) {
return list.get(0);
}
return null;
}
/**
* Initialized the internals from a new chunk
*
* @param content
* the new received chunk
* @throws ErrorDataDecoderException
* if there is a problem with the charset decoding or other
* errors
*/
@Override
public CompatibleHttpPostStandardRequestDecoder offer(HttpContent content) {
checkDestroyed();
if (content instanceof LastHttpContent) {
isLastChunk = true;
}
ByteBuf buf = content.content();
if (undecodedChunk == null) {
undecodedChunk =
// Since the Handler will release the incoming later on, we need to copy it
//
// We are explicit allocate a buffer and NOT calling copy() as otherwise it may set a maxCapacity
// which is not really usable for us as we may exceed it once we add more bytes.
buf.alloc().buffer(buf.readableBytes()).writeBytes(buf);
} else {
undecodedChunk.writeBytes(buf);
}
parseBody();
if (undecodedChunk != null && undecodedChunk.writerIndex() > discardThreshold) {
if (undecodedChunk.refCnt() == 1) {
// It's safe to call discardBytes() as we are the only owner of the buffer.
undecodedChunk.discardReadBytes();
} else {
// There seems to be multiple references of the buffer. Let's copy the data and release the buffer to
// ensure we can give back memory to the system.
ByteBuf buffer = undecodedChunk.alloc().buffer(undecodedChunk.readableBytes());
buffer.writeBytes(undecodedChunk);
undecodedChunk.release();
undecodedChunk = buffer;
}
}
return this;
}
/**
* True if at current getStatus, there is an available decoded
* InterfaceHttpData from the Body.
*
* This getMethod works for chunked and not chunked request.
*
* @return True if at current getStatus, there is a decoded InterfaceHttpData
* @throws EndOfDataDecoderException
* No more data will be available
*/
@Override
public boolean hasNext() {
checkDestroyed();
if (currentStatus == MultiPartStatus.EPILOGUE) {
// OK except if end of list
if (bodyListHttpDataRank >= bodyListHttpData.size()) {
return false;
}
}
return !bodyListHttpData.isEmpty() && bodyListHttpDataRank < bodyListHttpData.size();
}
/**
* Returns the next available InterfaceHttpData or null if, at the time it
* is called, there is no more available InterfaceHttpData. A subsequent
* call to offer(httpChunk) could enable more data.
*
* Be sure to call {@link InterfaceHttpData#release()} after you are done
* with processing to make sure to not leak any resources
*
* @return the next available InterfaceHttpData or null if none
* @throws EndOfDataDecoderException
* No more data will be available
*/
@Override
public InterfaceHttpData next() {
checkDestroyed();
if (hasNext()) {
return bodyListHttpData.get(bodyListHttpDataRank++);
}
return null;
}
@Override
public InterfaceHttpData currentPartialHttpData() {
return currentAttribute;
}
/**
* This getMethod will parse as much as possible data and fill the list and map
*
* @throws ErrorDataDecoderException
* if there is a problem with the charset decoding or other
* errors
*/
private void parseBody() {
if (currentStatus == MultiPartStatus.PREEPILOGUE || currentStatus == MultiPartStatus.EPILOGUE) {
if (isLastChunk) {
currentStatus = MultiPartStatus.EPILOGUE;
}
return;
}
parseBodyAttributes();
}
/**
* Utility function to add a new decoded data
*/
protected void addHttpData(InterfaceHttpData data) {
if (data == null) {
return;
}
List datas = bodyMapHttpData.get(data.getName());
if (datas == null) {
datas = new ArrayList(1);
bodyMapHttpData.put(data.getName(), datas);
}
datas.add(data);
bodyListHttpData.add(data);
}
/**
* This getMethod fill the map and list with as much Attribute as possible from
* Body in not Multipart mode.
*
* @throws ErrorDataDecoderException
* if there is a problem with the charset decoding or other
* errors
*/
private void parseBodyAttributesStandard() {
int firstpos = undecodedChunk.readerIndex();
int currentpos = firstpos;
int equalpos;
int ampersandpos;
if (currentStatus == MultiPartStatus.NOTSTARTED) {
currentStatus = MultiPartStatus.DISPOSITION;
}
boolean contRead = true;
try {
while (undecodedChunk.isReadable() && contRead) {
char read = (char) undecodedChunk.readUnsignedByte();
currentpos++;
switch (currentStatus) {
case DISPOSITION:// search '='
if (read == '=') {
currentStatus = MultiPartStatus.FIELD;
equalpos = currentpos - 1;
String key = decodeAttribute(undecodedChunk.toString(firstpos, equalpos - firstpos, charset),
charset);
currentAttribute = factory.createAttribute(request, key);
firstpos = currentpos;
} else if (read == '&') { // special empty FIELD
currentStatus = MultiPartStatus.DISPOSITION;
ampersandpos = currentpos - 1;
String key = decodeAttribute(
undecodedChunk.toString(firstpos, ampersandpos - firstpos, charset), charset);
currentAttribute = factory.createAttribute(request, key);
currentAttribute.setValue(""); // empty
addHttpData(currentAttribute);
currentAttribute = null;
firstpos = currentpos;
contRead = true;
}
break;
case FIELD:// search '&' or end of line
if (read == '&') {
currentStatus = MultiPartStatus.DISPOSITION;
ampersandpos = currentpos - 1;
setFinalBuffer(undecodedChunk.retainedSlice(firstpos, ampersandpos - firstpos));
firstpos = currentpos;
contRead = true;
} else if (read == HttpConstants.CR) {
if (undecodedChunk.isReadable()) {
read = (char) undecodedChunk.readUnsignedByte();
currentpos++;
if (read == HttpConstants.LF) {
currentStatus = MultiPartStatus.PREEPILOGUE;
ampersandpos = currentpos - 2;
setFinalBuffer(undecodedChunk.retainedSlice(firstpos, ampersandpos - firstpos));
firstpos = currentpos;
contRead = false;
} else {
// Error
throw new ErrorDataDecoderException("Bad end of line");
}
} else {
currentpos--;
}
} else if (read == HttpConstants.LF) {
currentStatus = MultiPartStatus.PREEPILOGUE;
ampersandpos = currentpos - 1;
setFinalBuffer(undecodedChunk.retainedSlice(firstpos, ampersandpos - firstpos));
firstpos = currentpos;
contRead = false;
}
break;
default:
// just stop
contRead = false;
}
}
if (isLastChunk && currentAttribute != null) {
// special case
ampersandpos = currentpos;
if (ampersandpos > firstpos) {
setFinalBuffer(undecodedChunk.retainedSlice(firstpos, ampersandpos - firstpos));
} else if (!currentAttribute.isCompleted()) {
setFinalBuffer(Unpooled.EMPTY_BUFFER);
}
firstpos = currentpos;
currentStatus = MultiPartStatus.EPILOGUE;
} else if (contRead && currentAttribute != null && currentStatus == MultiPartStatus.FIELD) {
// reset index except if to continue in case of FIELD getStatus
currentAttribute.addContent(undecodedChunk.retainedSlice(firstpos, currentpos - firstpos),
false);
firstpos = currentpos;
}
undecodedChunk.readerIndex(firstpos);
} catch (ErrorDataDecoderException e) {
// error while decoding
undecodedChunk.readerIndex(firstpos);
throw e;
} catch (IOException e) {
// error while decoding
undecodedChunk.readerIndex(firstpos);
throw new ErrorDataDecoderException(e);
} catch (IllegalArgumentException e) {
// error while decoding
undecodedChunk.readerIndex(firstpos);
throw new ErrorDataDecoderException(e);
}
}
/**
* This getMethod fill the map and list with as much Attribute as possible from
* Body in not Multipart mode.
*
* @throws ErrorDataDecoderException
* if there is a problem with the charset decoding or other
* errors
*/
private void parseBodyAttributes() {
if (undecodedChunk == null) {
return;
}
if (!undecodedChunk.hasArray()) {
parseBodyAttributesStandard();
return;
}
SeekAheadOptimize sao = new SeekAheadOptimize(undecodedChunk);
int firstpos = undecodedChunk.readerIndex();
int currentpos = firstpos;
int equalpos;
int ampersandpos;
if (currentStatus == MultiPartStatus.NOTSTARTED) {
currentStatus = MultiPartStatus.DISPOSITION;
}
boolean contRead = true;
try {
loop: while (sao.pos < sao.limit) {
char read = (char) (sao.bytes[sao.pos++] & 0xFF);
currentpos++;
switch (currentStatus) {
case DISPOSITION:// search '='
if (read == '=') {
currentStatus = MultiPartStatus.FIELD;
equalpos = currentpos - 1;
String key = decodeAttribute(undecodedChunk.toString(firstpos, equalpos - firstpos, charset),
charset);
currentAttribute = factory.createAttribute(request, key);
firstpos = currentpos;
} else if (read == '&') { // special empty FIELD
currentStatus = MultiPartStatus.DISPOSITION;
ampersandpos = currentpos - 1;
String key = decodeAttribute(
undecodedChunk.toString(firstpos, ampersandpos - firstpos, charset), charset);
currentAttribute = factory.createAttribute(request, key);
currentAttribute.setValue(""); // empty
addHttpData(currentAttribute);
currentAttribute = null;
firstpos = currentpos;
contRead = true;
}
break;
case FIELD:// search '&' or end of line
if (read == '&') {
currentStatus = MultiPartStatus.DISPOSITION;
ampersandpos = currentpos - 1;
setFinalBuffer(undecodedChunk.retainedSlice(firstpos, ampersandpos - firstpos));
firstpos = currentpos;
contRead = true;
} else if (read == HttpConstants.CR) {
if (sao.pos < sao.limit) {
read = (char) (sao.bytes[sao.pos++] & 0xFF);
currentpos++;
if (read == HttpConstants.LF) {
currentStatus = MultiPartStatus.PREEPILOGUE;
ampersandpos = currentpos - 2;
sao.setReadPosition(0);
setFinalBuffer(undecodedChunk.retainedSlice(firstpos, ampersandpos - firstpos));
firstpos = currentpos;
contRead = false;
break loop;
} else {
// Error
sao.setReadPosition(0);
throw new ErrorDataDecoderException("Bad end of line");
}
} else {
if (sao.limit > 0) {
currentpos--;
}
}
} else if (read == HttpConstants.LF) {
currentStatus = MultiPartStatus.PREEPILOGUE;
ampersandpos = currentpos - 1;
sao.setReadPosition(0);
setFinalBuffer(undecodedChunk.retainedSlice(firstpos, ampersandpos - firstpos));
firstpos = currentpos;
contRead = false;
break loop;
}
break;
default:
// just stop
sao.setReadPosition(0);
contRead = false;
break loop;
}
}
if (isLastChunk && currentAttribute != null) {
// special case
ampersandpos = currentpos;
if (ampersandpos > firstpos) {
setFinalBuffer(undecodedChunk.retainedSlice(firstpos, ampersandpos - firstpos));
} else if (!currentAttribute.isCompleted()) {
setFinalBuffer(Unpooled.EMPTY_BUFFER);
}
firstpos = currentpos;
currentStatus = MultiPartStatus.EPILOGUE;
} else if (contRead && currentAttribute != null && currentStatus == MultiPartStatus.FIELD) {
// reset index except if to continue in case of FIELD getStatus
currentAttribute.addContent(undecodedChunk.retainedSlice(firstpos, currentpos - firstpos),
false);
firstpos = currentpos;
}
undecodedChunk.readerIndex(firstpos);
} catch (ErrorDataDecoderException e) {
// error while decoding
undecodedChunk.readerIndex(firstpos);
throw e;
} catch (IOException e) {
// error while decoding
undecodedChunk.readerIndex(firstpos);
throw new ErrorDataDecoderException(e);
} catch (IllegalArgumentException e) {
// error while decoding
undecodedChunk.readerIndex(firstpos);
throw new ErrorDataDecoderException(e);
}
}
private void setFinalBuffer(ByteBuf buffer) throws IOException {
currentAttribute.addContent(buffer, true);
ByteBuf decodedBuf = decodeAttribute(currentAttribute.getByteBuf(), charset);
if (decodedBuf != null) { // override content only when ByteBuf needed decoding
currentAttribute.setContent(decodedBuf);
}
addHttpData(currentAttribute);
currentAttribute = null;
}
/**
* Decode component
*
* @return the decoded component
*/
private static String decodeAttribute(String s, Charset charset) {
try {
return ServletUtil.decodeComponent(s, 0, s.length(), charset);
} catch (IllegalArgumentException e) {
return s;
// throw new ErrorDataDecoderException("Bad string: '" + s + '\'', e);
}
}
private static ByteBuf decodeAttribute(ByteBuf b, Charset charset) {
int firstEscaped = b.forEachByte(urlEncodedDetector);
if (firstEscaped == -1) {
return null; // nothing to decode
}
ByteBuf buf = b.alloc().buffer(b.readableBytes());
UrlDecoder urlDecode = new UrlDecoder(buf);
b.forEachByte(urlDecode);
if (urlDecode.nextEscapedIdx != 0) { // incomplete hex byte
// if (idx == -1) {
// idx = b.readableBytes() - 1;
// }
// idx -= urlDecode.nextEscapedIdx - 1;
buf.release();
return null;
// throw new ErrorDataDecoderException(
// String.format("Invalid hex byte at index '%d' in string: '%s'", idx, b.toString(charset)));
}
return buf;
}
/**
* Destroy the and release all it resources. After this method
* was called it is not possible to operate on it anymore.
*/
@Override
public void destroy() {
// Release all data items, including those not yet pulled, only file based items
cleanFiles();
// Clean Memory based data
for (InterfaceHttpData httpData : bodyListHttpData) {
// Might have been already released by the user
if (httpData.refCnt() > 0) {
httpData.release();
}
}
destroyed = true;
if (undecodedChunk != null && undecodedChunk.refCnt() > 0) {
undecodedChunk.release();
undecodedChunk = null;
}
}
/**
* Clean all {@link HttpData}s for the current request.
*/
@Override
public void cleanFiles() {
checkDestroyed();
factory.cleanRequestHttpData(request);
}
/**
* Remove the given FileUpload from the list of FileUploads to clean
*/
@Override
public void removeHttpDataFromClean(InterfaceHttpData data) {
checkDestroyed();
factory.removeHttpDataFromClean(request, data);
}
private static final class UrlEncodedDetector implements ByteProcessor {
@Override
public boolean process(byte value) throws Exception {
return value != '%' && value != '+';
}
}
private static final class UrlDecoder implements ByteProcessor {
private final ByteBuf output;
private int nextEscapedIdx;
private byte hiByte;
UrlDecoder(ByteBuf output) {
this.output = output;
}
@Override
public boolean process(byte value) {
if (nextEscapedIdx != 0) {
if (nextEscapedIdx == 1) {
hiByte = value;
++nextEscapedIdx;
} else {
int hi = StringUtil.decodeHexNibble((char) hiByte);
int lo = StringUtil.decodeHexNibble((char) value);
if (hi == -1 || lo == -1) {
++nextEscapedIdx;
return false;
}
output.writeByte((hi << 4) + lo);
nextEscapedIdx = 0;
}
} else if (value == '%') {
nextEscapedIdx = 1;
} else if (value == '+') {
output.writeByte(' ');
} else {
output.writeByte(value);
}
return true;
}
}
}