In its current form, luceneutil expects a directory structure with baseline repo present in $BENCH_DIR/lucene_baseline and candidate repo in $BENCH_DIR/lucene_candidate.
While a lot of existing users already work with this setup, it adds some additional steps in certain scenarios like when your candidate is already checkout out in a different location, or when candidate is a different upstream branch on the baseline repo (e.g. fetching and benchmarking a PR locally).
This change leverages the argparser support added in #230 and adds arguments for baseline and candidate repo paths to the benchmark script. Both arguments are optional. If skipped, the benchmark script picks code from $BENCH_DIR/lucene_baseline and $BENCH_DIR/lucene_candidate respectively.
It also adds a -r / --reindex flag that recreates the candidate index when passed, which is useful when benchmarking an indexing side change.
The existing behavior is retained by default.
usage: Local Benchmark Run [-h] [-s SOURCE] [-concurrentSearches] [-b BASELINE] [-c CANDIDATE] [-r]
Run a local benchmark on provided source dataset.
options:
-h, --help show this help message and exit
-s SOURCE, -source SOURCE, --source SOURCE
Data source to run the benchmark on.
-concurrentSearches, --concurrentSearches
Run concurrent searches
-b BASELINE, --baseline BASELINE
Path to lucene repo to be used for baseline
-c CANDIDATE, --candidate CANDIDATE
Path to lucene repo to be used for candidate
-r, --reindex Reindex data for candidate run
In its current form,
luceneutilexpects a directory structure with baseline repo present in$BENCH_DIR/lucene_baselineand candidate repo in$BENCH_DIR/lucene_candidate.While a lot of existing users already work with this setup, it adds some additional steps in certain scenarios like when your candidate is already checkout out in a different location, or when candidate is a different upstream branch on the baseline repo (e.g. fetching and benchmarking a PR locally).
This change leverages the
argparsersupport added in #230 and adds arguments for baseline and candidate repo paths to the benchmark script. Both arguments are optional. If skipped, the benchmark script picks code from$BENCH_DIR/lucene_baselineand$BENCH_DIR/lucene_candidaterespectively.It also adds a
-r / --reindexflag that recreates the candidate index when passed, which is useful when benchmarking an indexing side change.The existing behavior is retained by default.
Testing
With new params
With default params