Joe Witt created NIFI-16313:
-------------------------------
Summary: SegmentContent documents a filename rewrite it does not
perform
Key: NIFI-16313
URL: https://issues.apache.org/jira/browse/NIFI-16313
Project: Apache NiFi
Issue Type: Task
Reporter: Joe Witt
SegmentContent's generated documentation disagrees with what onTrigger actually
writes.
@WritesAttributes currently has two entries, both named
"segment.original.filename " (trailing space):
1. "The filename of the parent FlowFile"
2. "The filename will be updated to include the parent's filename, the segment
index, and the segment count"
Neither is accurate as written. The runtime attribute is
FragmentAttributes.SEGMENT_ORIGINAL_FILENAME ("segment.original.filename", no
trailing space). The processor copies CoreAttributes.FILENAME onto that
attribute and leaves filename itself unchanged. MergeContent reconstitutes the
original FlowFile using segment.original.filename, so rewriting filename would
be the wrong contract.
This is a documentation defect, not a behavior change. The trailing-space paste
also appears on other Split* processors; those belong in a follow-up so this
ticket stays about SegmentContent's false rewrite claim.
Proposed change:
- Keep a single @WritesAttribute for segment.original.filename (no trailing
space) describing the parent's filename.
- Do not document a filename rewrite.
- Extend TestSegmentContent so segments keep the parent filename and set
segment.original.filename to that same value, for both the split path and the
"small enough not to segment" path.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
