Skip to content

Tags: Query-farm/vgi-java

Tags

v0.27.0

Toggle v0.27.0's commit message
v0.27.0 — protocol 1.4.0 (splits)

v0.26.1

Toggle v0.26.1's commit message
release: 0.26.1 — vgirpc 0.22.0, HTTP client and worker-log fix

vgirpc 0.22.0 adds HttpRpcConnection, so a Java client built on this SDK can
finally reach a worker on another host — every transport before it was
single-host. It also fixes a server-side defect that affects anyone already
serving streams over HTTP: worker logs were dropped after each stream's
first turn, and a failing turn's logs were discarded with it.

No change to this library's own code. Pin and CI refs move together.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

v0.26.0

Toggle v0.26.0's commit message
release: 0.26.0 — a Java client can build the payloads it sends

e63f297 is the release: `farm.query.vgi.client`, the encoder half this SDK
never had. Minor rather than patch — a new public package.

Picks up vgirpc 0.21.6, without which an exchange stream cannot be
cancelled and a worker error poisons the connection.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

v0.25.1

Toggle v0.25.1's commit message
release: 0.25.1 — the Java client no longer loses externalized batches

Picks up vgirpc 0.21.5. 0.25.0's headline was that a Java client can call a
VGI worker; it could, but a worker that externalised an oversized batch had
those rows silently discarded on the client. No change to this library's own
code — the fix is entirely in the RPC layer this pins.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

v0.25.0

Toggle v0.25.0's commit message
release: 0.25.0 — a Java client can call a VGI worker

4040555 is the release. `Worker.rpcServer()` lets a worker be served over a
caller-supplied transport, which is what an in-process client — or any
embedding — needs; the `run*` methods each own their transport and block.

Minor rather than patch: that method is new public API. It also picks up
vgirpc 0.21.4, without which a Java client's `init` stream mis-reads its
header and hangs at end of stream.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

v0.24.1

Toggle v0.24.1's commit message
release: 0.24.1 — dependency bumps

v0.24.0

Toggle v0.24.0's commit message
release: 0.24.0 — per-catalog attach options, and a gated ATTACH that…

… answers

`attach/attach_options_required.test` was skipping here: upstream 1f1d7b1 split
the `required` assertions out of the shared attach_options_echo.test behind
VGI_ATTACH_OPTIONS_REQUIRED_WORKER, because the catalog they need existed only
in the Python fixture. This grows the same catalog.

- `Worker.ExtraCatalog` gains a 5th component, `attachOptions`: an auxiliary
  catalog advertises its own specs on its `catalog_catalogs()` row and enforces
  them at its own attach. A 4-arg convenience constructor keeps every existing
  call site source-compatible. Deliberately per-catalog rather than folded into
  `Worker.attachOptions(...)` — one worker serves a catalog that requires an
  option next to one that takes none, which is exactly what the test asserts.
- Fixture: `AttachOptionsFixture.requiredSpecs()` (`api_key` required, `region`
  defaulting) as the `attach_options_required` catalog of the attach-options
  worker, mirroring vgi-python's `RequiredAttachOptions`. ci/run-integration.sh
  points the env var at that same worker on both lanes, so the test runs.
- The refusal path had never been exercised in Java, and it HUNG.
  `AttachOptionRequirements` throws `RpcError("ValueError", …)`, whose
  `errorKind()` is null as documented, and vgirpc wrote that null into the error
  batch's metadata — throwing in the flatbuffer writer, so the batch never
  reached the wire and the client waited on a response that would never come.
  Fixed in vgirpc 0.21.1; pinned here (build.gradle.kts + VGI_RPC_JAVA_REF).

Global functions needed no work: they shipped in 0.22.0 and both
global_functions tests pass against an extension carrying vgi 21fdd4e. CI's red
on them is the community extension lagging the consuming side — see CLAUDE.md.

Local: launch 296 files / 268 pass / 28 skip / 0 fail (attach_options_required
now among the passes); the two attach tests also verified under shm and over
HTTP. `:vgi:test` green with a new ExtraCatalogAttachOptionsTest; javadoc clean.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

v0.23.0

Toggle v0.23.0's commit message
release: 0.23.0 — the landing page reads the catalog over the protocol

Drops the describe.json producer (650 lines) and its conformance CI job: the
shared landing page now speaks the VGI protocol through the browser client
build vgirpc serves at {prefix}/vgi-client.js. WorkerLandingInfo.of(worker)
supplies the identity that used to be baked into the describe document.

Adds attach-option `required`, enforced at catalog_attach.

Requires farm.query:vgirpc 0.21.0; the integration workflow's VGI_RPC_JAVA_REF
moves with it.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

v0.22.0

Toggle v0.22.0's commit message
fix(ci): pick up vgirpc 0.20.1, which unbreaks the http integration lane

vgirpc 0.19.0 started stamping X-VGI-RPC-Error on a producer's in-band
error, where the exception rides a second concatenated IPC stream and the
reply is still present. The extension's error branch assumes a unary body,
so it read the header stream, never reached the exception, and reported a
generic "VGI HTTP RPC error" -- which sqllogictest ignores, so four
statement-error files silently SKIPPED rather than ran, and
ci/run-integration.sh's unexpected-skip guard failed the lane.

The four: copy_from/options, table_in_out/exceptions,
table_in_out/table_buffering_finalize_crash, and
table/partition_columns_contract.

Nothing in this repo could work around it. Bisected by holding vgi,
haybarn and the extension constant and varying only VGI_RPC_JAVA_REF:
typed errors propagate at v0.17.0 and v0.18.0, break at v0.19.0 and
v0.20.0, fixed in v0.20.1.

Bumps the floor and both workflow refs together -- leaving the refs behind
would have CI building vgirpc from source at a version consumers of the
published artifact don't get.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

v0.21.0

Toggle v0.21.0's commit message
release: bump 0.20.0 -> 0.21.0

Protocol 1.2.0 (schema-qualified (schema, name) dispatch across bind + the unary
aggregate/table-buffering RPCs) + result-cache key carries the owning schema so
same-name functions in different schemas no longer cross-serve. Breaking: needs
the vgi 1.2.0 extension.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>