Chu Cheng Li created HDDS-16620:
-----------------------------------
Summary: Expose BlockReadCursor as an iterator of read ranges
Key: HDDS-16620
URL: https://issues.apache.org/jira/browse/HDDS-16620
Project: Apache Ozone
Issue Type: Improvement
Reporter: Chu Cheng Li
Follow-up to HDDS-16258 ([PR
#11302|https://github.com/apache/ozone/pull/11302]).
The patch for HDDS-16258 moves chunk lookup and checksum-boundary calculations
into {{BlockReadCursor}}, but {{KeyValueHandler#readBlockImpl}} still obtains
the offset, length, and overlapping chunks separately and explicitly advances
the cursor.
Expose the cursor through {{Iterator<ReadRange>}}, where each range contains
its block offset, length, and overlapping chunk metadata. The handler can
consume each range with {{hasNext()}}/{{next()}}, then read, verify, and send
its data. File I/O and stream lifecycle stay in the handler.
Define advancement as handing out a planned range, and preserve accounting for
bytes actually read. The handler currently aborts on incomplete reads; retain
that behavior. Keep fallible range preparation outside {{Iterator.next()}},
which cannot declare a checked {{IOException}}.
Reuse the existing cursor and tests. Preserve checksum alignment for uneven
chunks, short final intervals, {{NONE}}, response metadata controlled by
{{includeChecksums}}, and stream error/cleanup behavior. Add coverage for
repeated {{hasNext()}} calls without advancement, complete range iteration, and
{{NoSuchElementException}} after exhaustion. This follow-up does not require a
protocol or storage-format change.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
