/**
* @see
* org.alfresco.repo.content.transform.ContentTransformer#transform(org.alfresco.service.cmr.repository.ContentReader,
* org.alfresco.service.cmr.repository.ContentWriter,
* org.alfresco.service.cmr.repository.TransformationOptions)
*/
public final void transform(
ContentReader reader, ContentWriter writer, TransformationOptions options)
throws ContentIOException {
try {
depth.set(depth.get() + 1);
// begin timing
long before = System.currentTimeMillis();
String sourceMimetype = reader.getMimetype();
String targetMimetype = writer.getMimetype();
// check options map
if (options == null) {
options = new TransformationOptions();
}
try {
if (transformerDebug.isEnabled()) {
transformerDebug.pushTransform(
this,
reader.getContentUrl(),
sourceMimetype,
targetMimetype,
reader.getSize(),
options);
}
// Check the transformability
checkTransformable(reader, writer, options);
// Pass on any limits to the reader
setReaderLimits(reader, writer, options);
// Transform
// MNT-12238: CLONE - CLONE - Upload of PPTX causes very high memory usage leading to system
// instability
// Limiting transformation up to configured amount of milliseconds to avoid very high RAM
// consumption
// and OOM during transforming problematic documents
TransformationOptionLimits limits =
getLimits(reader.getMimetype(), writer.getMimetype(), options);
long timeoutMs = limits.getTimeoutMs();
if (!useTimeoutThread || (null == limits) || (-1 == timeoutMs)) {
transformInternal(reader, writer, options);
} else {
Future<?> submittedTask = null;
StreamAwareContentReaderProxy proxiedReader = new StreamAwareContentReaderProxy(reader);
StreamAwareContentWriterProxy proxiedWriter = new StreamAwareContentWriterProxy(writer);
try {
submittedTask =
getExecutorService()
.submit(new TransformInternalCallable(proxiedReader, proxiedWriter, options));
submittedTask.get(timeoutMs + additionalThreadTimout, TimeUnit.MILLISECONDS);
} catch (TimeoutException e) {
releaseResources(submittedTask, proxiedReader, proxiedWriter);
throw new TimeoutException("Transformation failed due to timeout limit");
} catch (InterruptedException e) {
releaseResources(submittedTask, proxiedReader, proxiedWriter);
throw new InterruptedException(
"Transformation failed, because the thread of the transformation was interrupted");
} catch (ExecutionException e) {
Throwable cause = e.getCause();
if (cause instanceof TransformInternalCallableException) {
cause = ((TransformInternalCallableException) cause).getCause();
}
throw cause;
}
}
// record time
long after = System.currentTimeMillis();
recordTime(sourceMimetype, targetMimetype, after - before);
} catch (ContentServiceTransientException cste) {
// A transient failure has occurred within the content transformer.
// This should not be interpreted as a failure and therefore we should not
// update the transformer's average time.
if (logger.isDebugEnabled()) {
logger.debug(
"Transformation has been transiently declined: \n"
+ " reader: "
+ reader
+ "\n"
+ " writer: "
+ writer
+ "\n"
+ " options: "
+ options
+ "\n"
+ " transformer: "
+ this);
}
// the finally block below will still perform tidyup. Otherwise we're done.
// We rethrow the exception
throw cste;
} catch (UnsupportedTransformationException e) {
// Don't record an error or even the time, as this is normal in compound transformations.
transformerDebug.debug(" Failed", e);
throw e;
} catch (Throwable e) {
// Make sure that this transformation gets set back i.t.o. time taken.
// This will ensure that transformers that compete for the same transformation
// will be prejudiced against transformers that tend to fail
long after = System.currentTimeMillis();
recordError(sourceMimetype, targetMimetype, after - before);
// Ask Tika to detect the document, and report back on if
// the current mime type is plausible
String differentType = getMimetypeService().getMimetypeIfNotMatches(reader.getReader());
// Report the error
if (differentType == null) {
transformerDebug.debug(" Failed", e);
throw new ContentIOException(
"Content conversion failed: \n"
+ " reader: "
+ reader
+ "\n"
+ " writer: "
+ writer
+ "\n"
+ " options: "
+ options.toString(false)
+ "\n"
+ " limits: "
+ getLimits(reader, writer, options),
e);
} else {
transformerDebug.debug(" Failed: Mime type was '" + differentType + "'", e);
if (this.retryTransformOnDifferentMimeType) {
// MNT-11015 fix.
// Set a new reader to refresh the input stream.
reader = reader.getReader();
// set the actual file MIME type detected by Tika for content reader
reader.setMimetype(differentType);
// Get correct transformer according actual file MIME type and try to transform file
// with
// actual transformer
ContentTransformer transformer =
this.registry.getTransformer(
differentType, reader.getSize(), targetMimetype, options);
if (null != transformer) {
transformer.transform(reader, writer, options);
} else {
transformerDebug.debug(" Failed", e);
throw new ContentIOException(
"Content conversion failed: \n"
+ " reader: "
+ reader
+ "\n"
+ " writer: "
+ writer
+ "\n"
+ " options: "
+ options.toString(false)
+ "\n"
+ " limits: "
+ getLimits(reader, writer, options)
+ "\n"
+ " claimed mime type: "
+ reader.getMimetype()
+ "\n"
+ " detected mime type: "
+ differentType
+ "\n"
+ " transformer not found"
+ "\n",
e);
}
} else {
throw new ContentIOException(
"Content conversion failed: \n"
+ " reader: "
+ reader
+ "\n"
+ " writer: "
+ writer
+ "\n"
+ " options: "
+ options.toString(false)
+ "\n"
+ " limits: "
+ getLimits(reader, writer, options)
+ "\n"
+ " claimed mime type: "
+ reader.getMimetype()
+ "\n"
+ " detected mime type: "
+ differentType,
e);
}
}
} finally {
transformerDebug.popTransform();
// check that the reader and writer are both closed
if (reader.isChannelOpen()) {
logger.error(
"Content reader not closed by transformer: \n"
+ " reader: "
+ reader
+ "\n"
+ " transformer: "
+ this);
}
if (writer.isChannelOpen()) {
logger.error(
"Content writer not closed by transformer: \n"
+ " writer: "
+ writer
+ "\n"
+ " transformer: "
+ this);
}
}
// done
if (logger.isDebugEnabled()) {
logger.debug(
"Completed transformation: \n"
+ " reader: "
+ reader
+ "\n"
+ " writer: "
+ writer
+ "\n"
+ " options: "
+ options
+ "\n"
+ " transformer: "
+ this);
}
} finally {
depth.set(depth.get() - 1);
}
}
/**
* @see
* org.alfresco.repo.content.transform.AbstractContentTransformer2#transformInternal(org.alfresco.service.cmr.repository.ContentReader,
* org.alfresco.service.cmr.repository.ContentWriter,
* org.alfresco.service.cmr.repository.TransformationOptions)
*/
@Override
public void transformInternal(
ContentReader reader, ContentWriter writer, TransformationOptions options) throws Exception {
final String outputMimetype = writer.getMimetype();
final String outputFileExt = getMimetypeService().getExtension(outputMimetype);
// We need to keep a reference to thrown exceptions as we're going to catch them and
// then move on to the next transformer. In the event that they all fail, we will throw
// the final exception.
Exception transformationException = null;
for (int i = 0; i < transformers.size(); i++) {
int oneBasedCount = i + 1;
ContentTransformer transf = transformers.get(i);
ContentWriter currentWriter = null;
File tempFile = null;
try {
if (logger.isDebugEnabled()) {
logger.debug(
"Transformation attempt "
+ oneBasedCount
+ " of "
+ transformers.size()
+ ": "
+ transf);
}
if (!transf.isTransformable(
reader.getMimetype(), reader.getSize(), outputMimetype, options)) {
throw new UnsupportedTransformationException(
"Unsupported transformation: " + reader.getMimetype() + " to " + outputMimetype);
}
// We can't know in advance which transformer in the sequence will work - if any.
// Therefore we can't write into the ContentWriter stream.
// So make a temporary file writer with the current transformer name.
tempFile =
TempFileProvider.createTempFile(
"FailoverTransformer_intermediate_" + transf.getClass().getSimpleName() + "_",
"." + outputFileExt);
currentWriter = new FileContentWriter(tempFile);
currentWriter.setMimetype(outputMimetype);
currentWriter.setEncoding(writer.getEncoding());
// attempt to transform
transf.transform(reader, currentWriter, options);
// TODO Could add a check for zero-length output and treat that as a failure
// final long writtenSize = currentWriter.getSize();
} catch (Exception are) {
if (transformationException == null) {
transformationException = are;
}
if (logger.isDebugEnabled()) {
logger.debug("Transformation " + oneBasedCount + " was unsuccessful.");
if (i != transformers.size() - 1) {
// We don't log the last exception as we're going to throw it.
logger.debug("The below exception is provided for information purposes only.", are);
}
}
// Set a new reader to refresh the input stream.
reader = reader.getReader();
// and move to the next transformer
continue;
}
// No need to close input or output streams
// At this point the current transformation was successful i.e. it did not throw an exception.
// Now we must copy the content from the temporary file into the ContentWriter stream.
if (tempFile != null) {
writer.putContent(tempFile);
}
if (logger.isInfoEnabled()) {
logger.info("Transformation was successful");
}
return;
}
// At this point we have tried all transformers in the sequence without apparent success.
if (transformationException != null) {
transformerDebug.debug(" No more transformations to failover to");
if (logger.isDebugEnabled()) {
logger.debug(
"All transformations were unsuccessful. Throwing first exception.",
transformationException);
}
throw transformationException;
}
}