jgehrcke
Follow
Pinned
359 contributions in the last year
Less
More
Activity overview
Contributed to
conbench/conbench,
jgehrcke/github-repo-stats,
jgehrcke/gipc
and 11 other
repositories
Contribution activity
December 2022
Created 45 commits in 5 repositories
Created 2 repositories
- jgehrcke/arrow C++
- jgehrcke/vdlabs-benchmarks Python
Created a pull request in voltrondata-labs/benchmarks that received 42 comments
benchmarks: add dataset_serialize_benchmark.py
This is work in progress for #123.
+408
−4
•
42
comments
Opened 37 other pull requests in 5 repositories
conbench/conbench
32
merged
1
open
- get_github_commit(): clarify interface, expect rate limiting errs in tests
- ci: take screenshots of UI, upload as artifacts
- GitHub actions: concurrent, independent steps
- api/index.py: fix wipe-db filter
- README: improvements w.r.t. state management
- makefile: add teardown-app target
- devprod: mount local checkout into container
- db population helper: consolidate, test
- devprod: add /api/wipe-db endpoint
- conbench cli: show result timing overview
- duration measurements: use monotonic time source
- access control: remove qmark in redirect
- makefile, readme: small tweaks/fixes
- access control: app: use authorize_or_terminate decorator
- access control: retain target URL as query parm when redirecting to the login page
- makefile: remove CI=true
-
ci, readme, makefile: use Compose V2, i.e.
docker compose - cover tests.populate_local_conbench in CI
- README and Makefile: first-class common developer workflows
- tests_docs: do not look at CI var
- makefile: cmd to recreate expected api docs
- tests: speedup: empty tables instead of drop_all/create_all
- Introduce Makefile, clarify that Conbench wants Python 3.10
- Remove Buildkite from per-PR CI
- oidc: introduce parse_state_from_authorization_response(), fix state parser
- Some pull requests not shown.
apache/arrow
1
merged
voltrondata-labs/arrow-benchmarks-ci
1
merged
voltrondata-labs/benchmarks
1
merged
jgehrcke/github-repo-stats
1
merged
Reviewed 18 pull requests in 3 repositories
conbench/conbench
16 pull requests
- UI updates : conbench index page
- get_github_commit(): clarify interface, expect rate limiting errs in tests
- GitHub actions: concurrent, independent steps
- Recommend GITHUB_API_TOKEN
- devprod: add /api/wipe-db endpoint
- conbench cli: show result timing overview
- redact SQLALCHEMY_DATABASE_URI
- Reworked queries to use git history
- makefile, readme: small tweaks/fixes
- access control: retain target URL as query parm when redirecting to the login page
- oidc: introduce parse_state_from_authorization_response(), fix state parser
- tests_docs: do not look at CI var
- config: make INTENDED_BASE_URL required
- Pass CI=true to GHA tests
- README: document current recommended test invocation method
- OIDC: carry target URL across flow
apache/arrow
1 pull request
voltrondata-labs/benchmarks
1 pull request
Created an issue in conbench/conbench that received 6 comments
ci instability: test_get_github_commit... might run into github api rate limit
Got this when running the test suite locally:
============================================================== FAILURES =============================…
6
comments
Opened 39 other issues in 4 repositories
conbench/conbench
24
open
12
closed
- commit author prominence: rethink?
- UI: should work nicely on the mobile phone
- rethink terminology for main primitives
- allow for searching by benchmark name
- Finding all commits to the master branch of apache/arrow: may take longer than WORKER TIMEOUT
- ci: db-populate: RemoteDisconnected upon first benchmark POST
- CI: split independent categories into concurrently running jobs
- batches view: slider: do not slide automatically
- api/compare/commits/foo...bar may take too long, leading to worker timeout
- python -m conbench.tests.populate_local_conbench may fail with worker timeout
- result page: show mean value and uncertainty in 'standard notation'
-
detect multimodal distribution in sample submission, then do not use the
mean - BenchmarkCreate schema: timestamp should be tz-aware or tz-agnostic
-
input validation: require consistency between
data,iterations, and aggregates (stdevetc) - input validation: reject zero samples?
- plotting and analysis: derive per-sample uncertainty from N repetitions, show error bars/distr
- Any duration measurement: use monotonic time source: time.monotonic(), time.perf_counter()
- redirect to target after login: tidy up query section -- stray question mark
- Cover tests/populate_local_conbench.py in CI
-
GET /benchmarks/<id> might fail with
TypeError: object of type 'NoneType' has no len() - debuggability: have better logging around failing github api interaction
- explore: require GITHUB_API_TOKEN for local dev?
-
use
docker composeinstead ofdocker-compose - developer experience: provide make/build targets for most common scenarios
- Define CPython version recommeded to run Conbench with, use this across tooling
- Some issues not shown.
voltrondata-labs/arrow-benchmarks-ci
1
open
apache/arrow
1
closed
voltrondata-labs/benchmarks
1
closed
11
contributions
in private repositories
Dec 6 – Dec 24