Add tsblock tsfile writer

[shuwenwei]

No description provided.

[Copilot]

Pull request overview

Adds a V4 writer to convert table-model TsBlock data into a TsFile, with accompanying tests, plus some refactoring to enable subclass access in aligned chunk-group writers.

Changes:

• Introduce TableTsBlock2TsFileWriter to split a table-model TsBlock by device and write per-device aligned chunks (with optional synthetic time generation).
• Add unit tests covering multiple mappings (existing time column vs generated time; tag-driven device vs single-device).
• Extend chunk writer implementations (new TableChunkGroupWriterImpl.write(...) and make some AlignedChunkGroupWriterImpl members protected) to support the new writer flow.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
java/tsfile/src/test/java/org/apache/tsfile/write/writer/TableTsBlock2TsFileWriterTest.java	Adds tests validating the new TsBlock→TsFile writing behavior across different column mappings.
java/tsfile/src/main/java/org/apache/tsfile/write/v4/TableTsBlock2TsFileWriter.java	New writer that partitions TsBlock rows by device and writes aligned chunk groups, optionally generating per-device timestamps.
java/tsfile/src/main/java/org/apache/tsfile/write/chunk/TableChunkGroupWriterImpl.java	Adds a convenience write(...) API for table-style writes (currently contains correctness issues).
java/tsfile/src/main/java/org/apache/tsfile/write/chunk/AlignedChunkGroupWriterImpl.java	Changes access modifiers to protected to enable reuse/extension by table-model writers.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

getDeviceId assumes every tag column is a non-null STRING/BINARY and unconditionally calls getBinary(rowIdx).getStringValue(...). If a tag value is null (or the tag column type isn’t binary-compatible), this will throw at runtime. Consider following the Tablet#getDeviceID approach: check isNull(rowIdx) and use getObject(rowIdx)/toString() (or otherwise handle non-string tag types) so device ID generation is robust to nulls and supported tag types.

Suggested change

	segments[i + 1] =
	tsBlock
	.getValueColumns()[tagColumnIndexInTsBlock[i]]
	.getBinary(rowIdx)
	.getStringValue(TSFileConfig.STRING_CHARSET);
	Column tagColumn = tsBlock.getValueColumns()[tagColumnIndexInTsBlock[i]];
	if (tagColumn.isNull(rowIdx)) {
	segments[i + 1] = null;
	} else {
	Object value = tagColumn.getObject(rowIdx);
	segments[i + 1] = value == null ? null : value.toString();
	}

[Copilot]

The test method names test1, test2, test3 don’t communicate what scenarios are being validated (e.g., existing time column vs generated time column, with/without tag columns). Renaming them to reflect the specific behavior under test will make failures much easier to triage.

Suggested change

	public void test1() throws IOException, WriteProcessException {
	public void testWriteWithExistingTimeColumnAndTagColumns() throws IOException, WriteProcessException {

[Copilot]

In write(...), the ValueChunkWriter is indexed by rowIndex (valueChunkWriters[rowIndex]) instead of by valueColumnIndex. This will throw ArrayIndexOutOfBoundsException as soon as endRowIndex exceeds the number of columns, and it also writes data into the wrong series. Use the value-column index when selecting the writer, and consider validating measurementSchemas.size() matches valueColumns.length before initializing/iterating to avoid AIOOB or null writers. Also, this method currently skips the out-of-order check (checkIsHistoryData(time)) that other chunk writers perform, so out-of-order writes may silently corrupt state; it should apply the same check (and adjust the throws signature accordingly).

[JackieTien97]

it may be better to check timeColumn.isNull, you can throw exception if so.