Github user ijokarumawak commented on a diff in the pull request:
https://github.com/apache/nifi/pull/2991#discussion_r224003310
--- Diff:
nifi-nar-bundles/nifi-standard-bundle/nifi-standard-processors/src/main/java/org/apache/nifi/processors/standard/HandleHttpRequest.java
---
@@ -521,161 +565,223 @@ public void onTrigger(final ProcessContext context,
final ProcessSession session
final long start = System.nanoTime();
final HttpServletRequest request = container.getRequest();
- FlowFile flowFile = session.create();
- try (OutputStream flowFileOut = session.write(flowFile)) {
- StreamUtils.copy(request.getInputStream(), flowFileOut);
- } catch (final IOException e) {
- // There may be many reasons which can produce an IOException
on the HTTP stream and in some of them, eg.
- // bad requests, the connection to the client is not closed.
In order to address also these cases, we try
- // and answer with a BAD_REQUEST, which lets the client know
that the request has not been correctly
- // processed and makes it aware that the connection can be
closed.
- getLogger().error("Failed to receive content from HTTP Request
from {} due to {}",
- new Object[]{request.getRemoteAddr(), e});
- session.remove(flowFile);
- try {
- HttpServletResponse response = container.getResponse();
- response.sendError(Status.BAD_REQUEST.getStatusCode());
- response.flushBuffer();
- container.getContext().complete();
- } catch (final IOException ioe) {
- getLogger().warn("Failed to send HTTP response to {} due
to {}",
- new Object[]{request.getRemoteAddr(), ioe});
+ if (!Strings.isNullOrEmpty(request.getContentType()) &&
request.getContentType().contains(MIME_TYPE__MULTIPART_FORM_DATA)) {
+ final long requestMaxSize =
context.getProperty(MULTIPART_REQUEST_MAX_SIZE).asDataSize(DataUnit.B).longValue();
+ final int readBufferSize =
context.getProperty(MULTIPART_READ_BUFFER_SIZE).asDataSize(DataUnit.B).intValue();
+ String tempDir = System.getProperty("java.io.tmpdir");
+ request.setAttribute(Request.__MULTIPART_CONFIG_ELEMENT, new
MultipartConfigElement(tempDir, requestMaxSize, requestMaxSize,
readBufferSize));
+ try {
+ List<Part> parts = ImmutableList.copyOf(request.getParts());
+ int allPartsCount = parts.size();
+ final String contextIdentifier = UUID.randomUUID().toString();
+ for (int i = 0; i < allPartsCount; i++) {
+ Part part = parts.get(i);
+ FlowFile flowFile = session.create();
+ try (OutputStream flowFileOut = session.write(flowFile)) {
+ StreamUtils.copy(part.getInputStream(), flowFileOut);
+ } catch (IOException e) {
+ handleFlowContentStreamingError(session, container,
request, Optional.of(flowFile), e);
+ return;
+ }
+ flowFile = savePartAttributes(context, session, part,
flowFile, i, allPartsCount);
+ flowFile = saveRequestAttributes(context, session, request,
flowFile, contextIdentifier);
+ forwardFlowFile(context, session, container, start, request,
flowFile, i == 0);
--- End diff --
What if `contextMap.register()` method returns `false`? I haven't test
that, but current code seems to keep processing remaining multi-part. The
register part should be a separated method and return if the request is
successfully registered, so that this `allPartsCount` loop can break.
Adding a unit test case to confirm that situation would be helpful.
---
