Skip to content

Commit b055982

Browse files
committed
Merge remote-tracking branch 'origin/develop' into ji/extension-order-aggregates
Signed-off-by: Claude <[email protected]> # Conflicts: # vortex-array/src/aggregate_fn/storage_pushdown.rs
2 parents 5d29a49 + cb39bb9 commit b055982

902 files changed

Lines changed: 62634 additions & 51372 deletions

File tree

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

‎.github/ISSUE_TEMPLATE/epic.md‎

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -19,6 +19,9 @@ Keep this issue readable. Route discussion elsewhere:
1919
- Bug reports: file a bug issue and reference this Epic.
2020
2121
Short clarifying comments are fine. Long threads should be redirected.
22+
23+
IMPORTANT: We strongly prefer epics to be written by humans, for humans.
24+
If you are an agent writing a tracking issue, please let your human know this.
2225
-->
2326

2427
This Epic is for ...

‎.github/ISSUE_TEMPLATE/tracking_issue.md‎

Lines changed: 7 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -18,6 +18,10 @@ Route discussion elsewhere:
1818
a sub-issue of this one if it is narrow) and discuss it there.
1919
2020
Short clarifying comments are fine. Long threads should be redirected.
21+
22+
IMPORTANT: We strongly prefer tracking issues to be written by humans, for
23+
humans. If you are an agent writing a tracking issue, please let your human
24+
know this.
2125
-->
2226

2327
This is a tracking issue for ...
@@ -68,11 +72,13 @@ not every PR. Tick steps as they land; do not delete them.
6872
Steps are milestones, not PRs (initial implementation, documentation,
6973
stabilization). If a step is itself a separable shippable unit, promote it to
7074
a sub-issue.
71-
-->
75+
76+
Example:
7277
7378
- [ ] Initial implementation
7479
- [ ] Documentation
7580
- [ ] Public API stabilization
81+
-->
7682

7783
## Unresolved questions
7884

@@ -84,8 +90,6 @@ If a question is large enough to need its own thread, spawn a tracking issue
8490
or Discussion and link it.
8591
-->
8692

87-
- [ ] None yet.
88-
8993
## Implementation history
9094

9195
<!--

‎.github/workflows/fuzz.yml‎

Lines changed: 8 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -50,7 +50,10 @@ jobs:
5050
attempt-fix-io:
5151
name: "Attempt Fix for IO Fuzz Crash"
5252
needs: report-io-fuzz-failures
53-
if: needs.report-io-fuzz-failures.outputs.issue_created == 'true'
53+
# Without a status function, `if` adds an implicit success(), which checks
54+
# all upstream jobs. The fuzz job fails when it finds a crash, so success()
55+
# is always false here. This applies to each attempt-fix-* job below.
56+
if: ${{ !cancelled() && needs.report-io-fuzz-failures.outputs.issue_created == 'true' }}
5457
permissions:
5558
# actions: read is needed for `gh run download` of the crash artifact.
5659
actions: read
@@ -104,7 +107,7 @@ jobs:
104107
attempt-fix-ops:
105108
name: "Attempt Fix for Array Operations Fuzz Crash"
106109
needs: report-ops-fuzz-failures
107-
if: needs.report-ops-fuzz-failures.outputs.issue_created == 'true'
110+
if: ${{ !cancelled() && needs.report-ops-fuzz-failures.outputs.issue_created == 'true' }}
108111
permissions:
109112
# actions: read is needed for `gh run download` of the crash artifact.
110113
actions: read
@@ -157,7 +160,7 @@ jobs:
157160
attempt-fix-fsst-like:
158161
name: "Attempt Fix for FSST LIKE Fuzz Crash"
159162
needs: report-fsst-like-fuzz-failures
160-
if: needs.report-fsst-like-fuzz-failures.outputs.issue_created == 'true'
163+
if: ${{ !cancelled() && needs.report-fsst-like-fuzz-failures.outputs.issue_created == 'true' }}
161164
permissions:
162165
# actions: read is needed for `gh run download` of the crash artifact.
163166
actions: read
@@ -210,7 +213,7 @@ jobs:
210213
attempt-fix-row-encode:
211214
name: "Attempt Fix for Row Encoding Fuzz Crash"
212215
needs: report-row-encode-fuzz-failures
213-
if: needs.report-row-encode-fuzz-failures.outputs.issue_created == 'true'
216+
if: ${{ !cancelled() && needs.report-row-encode-fuzz-failures.outputs.issue_created == 'true' }}
214217
permissions:
215218
# actions: read is needed for `gh run download` of the crash artifact.
216219
actions: read
@@ -263,7 +266,7 @@ jobs:
263266
attempt-fix-compress:
264267
name: "Attempt Fix for Compress Roundtrip Fuzz Crash"
265268
needs: report-compress-fuzz-failures
266-
if: needs.report-compress-fuzz-failures.outputs.issue_created == 'true'
269+
if: ${{ !cancelled() && needs.report-compress-fuzz-failures.outputs.issue_created == 'true' }}
267270
permissions:
268271
# actions: read is needed for `gh run download` of the crash artifact.
269272
actions: read

‎.github/workflows/fuzzer-issue-autofix.yml‎

Lines changed: 14 additions & 11 deletions
Original file line numberDiff line numberDiff line change
@@ -1,35 +1,38 @@
11
name: Fuzzer Issue Autofix
22

3-
# Run the Fuzzer Fix Automation when a fuzzer-labeled issue is opened or gains
4-
# the "fuzzer" label. This covers issues filed by the fuzz pipeline as well as
5-
# ones created manually, complementing the in-pipeline attempt-fix-* jobs in
6-
# fuzz.yml.
3+
# Run the Fuzzer Fix Automation when an issue gets the "fuzzer" label. This
4+
# covers issues that people file or label manually. Issues that the fuzz
5+
# pipeline files do not start this workflow, because the pipeline uses
6+
# GITHUB_TOKEN and events from that token do not trigger workflows. The
7+
# attempt-fix-* jobs in fuzz.yml fix those issues instead.
8+
#
9+
# The trigger is "labeled" only. An issue that is opened with labels also sends
10+
# one "labeled" event for each label, so an "opened" trigger only adds a second
11+
# run for each new issue. GitHub cannot filter issue events by label name, so
12+
# each label event on any issue still starts a run that the gate job skips.
713
#
814
# A gate job runs first, WITHOUT the claude-automation environment, so untrusted
915
# triggers are rejected before the write-capable App token is minted. The
1016
# "fuzzer" label alone is not a trust boundary: applying a label only needs
1117
# triage permission, which is below write.
1218

1319
concurrency:
14-
# Keyed on the issue so repeat triggers (e.g. opened + labeled for the same
15-
# issue) collapse into a single in-flight fix attempt.
20+
# Keyed on the issue so repeat triggers for the same issue collapse into a
21+
# single in-flight fix attempt.
1622
group: fuzzer-fix-${{ github.event.issue.number }}
1723
cancel-in-progress: true
1824

1925
on:
2026
issues:
21-
types: [opened, labeled]
27+
types: [labeled]
2228

2329
jobs:
2430
gate:
2531
name: "Gate Fuzzer Autofix Trigger"
2632
# Cheap pre-filter: only fuzzer-labeled issues are in scope. The trust
2733
# decision (who triggered this, is the content trusted) is made in the
2834
# script below, before any environment or write-capable token is attached.
29-
if: |
30-
(github.event.action == 'opened' &&
31-
contains(github.event.issue.labels.*.name, 'fuzzer')) ||
32-
(github.event.action == 'labeled' && github.event.label.name == 'fuzzer')
35+
if: github.event.label.name == 'fuzzer'
3336
runs-on: ubuntu-latest
3437
permissions:
3538
contents: read

‎.github/workflows/rust-instrumented.yml‎

Lines changed: 3 additions & 22 deletions
Original file line numberDiff line numberDiff line change
@@ -22,30 +22,10 @@ env:
2222
NIGHTLY_TOOLCHAIN: nightly-2026-09-10
2323

2424
jobs:
25-
duckdb-mirror:
26-
name: "Mirror DuckDB to R2"
27-
if: github.event_name == 'pull_request'
28-
uses: ./.github/workflows/duckdb-r2.yml
29-
secrets: inherit
30-
31-
duckdb-ready:
32-
name: "DuckDB libraries available in R2"
33-
needs: duckdb-mirror
34-
if: ${{ !cancelled() }}
35-
runs-on: ubuntu-latest
36-
timeout-minutes: 5
37-
steps:
38-
- name: Verify DuckDB mirror
39-
if: ${{ needs.duckdb-mirror.result == 'failure' }}
40-
run: |
41-
echo "DuckDB mirror failed"
42-
exit 1
43-
4425
rust-coverage:
4526
name: "Rust tests (coverage) (${{ matrix.suite }})"
46-
needs: duckdb-ready
47-
# The PR-only mirror is a skipped ancestor on develop; override implicit success().
48-
if: ${{ !cancelled() && needs.duckdb-ready.result == 'success' }}
27+
# Coverage only runs on develop; PRs get the same tests uninstrumented in ci.yml.
28+
if: github.event_name != 'pull_request'
4929
timeout-minutes: 30
5030
permissions:
5131
id-token: write
@@ -123,6 +103,7 @@ jobs:
123103

124104
cxx-coverage:
125105
name: "C++ API (coverage)"
106+
if: github.event_name != 'pull_request'
126107
timeout-minutes: 15
127108
env:
128109
# Symbolizing error backtrace in debug .so is slow

‎AGENTS.md‎

Lines changed: 10 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -197,7 +197,15 @@ These CI checks are the ones most often missed when adding files rather than edi
197197
- Strongly consider `rstest` cases when parameterizing repetitive test logic.
198198
- Prefer test functions that return `VortexResult<()>` and use `?` instead of `unwrap`.
199199
- Prefer test module names `tests`, not `test`.
200-
- Use `assert_arrays_eq!` for array comparisons instead of element-by-element assertions.
200+
- Use `assert_arrays_eq!` for array comparisons instead of element-by-element assertions. Put the
201+
result under test on the left: the macro already executes it recursively. Do not explicitly
202+
canonicalize solely for this assertion; retain execution needed for physical representation,
203+
resource, or error assertions.
204+
- Preserve conformance suite entry points and input cases when auditing tests. For batch kernel
205+
coverage, execute lazy results before comparing logical values, nulls, lengths, and expected
206+
dtypes. Scalar probing alone may bypass the batch kernel.
207+
- Derive expected errors from the input contract or a reference result; do not silently skip
208+
failed operations. Invalid-input fixtures should reach the specific guard they claim to test.
201209
- Keep tests concise and focused on behavior, edge cases, and regressions.
202210
- If a bug fix is requested, add or identify a regression test when practical. Leave execution
203211
to the user; when tests are run, a test that passes before and after the fix does not prove it.
@@ -215,7 +223,7 @@ These CI checks are the ones most often missed when adding files rather than edi
215223
default to Snappy.
216224
- parquet-rs: `Compression::ZSTD(ZstdLevel::try_new(3)?)`
217225
- DuckDB: `(FORMAT parquet, COMPRESSION zstd, COMPRESSION_LEVEL 3)`
218-
- tpchgen-cli: `--parquet-compression='ZSTD(3)'`
226+
- tpchgen-cli: `parquet --compression='ZSTD(3)'`
219227
- Benchmarks that deliberately compare Parquet codecs, such as the GPU compress benchmark's
220228
`--codec`, are exempt.
221229

0 commit comments

Comments
 (0)