feat: add bazel-diff serve HTTP query service (#29)#393
Open
tinder-maxwellelliott wants to merge 2 commits into
Open
feat: add bazel-diff serve HTTP query service (#29)#393tinder-maxwellelliott wants to merge 2 commits into
bazel-diff serve HTTP query service (#29)#393tinder-maxwellelliott wants to merge 2 commits into
Conversation
Implements the "bazel-diff as a service" model from the BazelCon talk that inspired this repo (issue #29), as a first self-contained increment: core service + per-SHA hash cache. New `com.bazel_diff.server` package + `serve` subcommand: - HTTP/JSON server on the JDK's com.sun.net.httpserver (no new dependency) with GET /health and GET /impacted_targets?from&to[&targetType]. - Per-SHA hash cache behind a byte-oriented HashCacheStorage interface (LocalDiskHashCacheStorage impl, atomic writes); cache key folds in a fingerprint of the query-affecting flags. The interface is ready for a remote (e.g. S3) backend. - GitClient/ProcessGitClient: subprocess git fetch/resolve/checkout. - HashService (per-SHA generate+cache, single workspace lock) and ImpactedTargetsService, which reuse CalculateImpactedTargetsInteractor so results match `get-impacted-targets`. - Initial `git fetch` gates readiness; a fetch failure "lame-ducks" the instance (health stays 503) rather than self-repairing. Query-affecting flags mirror `generate-hashes`, so hashes the service produces match a cold CLI run. Tests: unit tests per component plus a real end-to-end test (E2ETest#testServeEndToEnd) that drives the live server over HTTP against a two-commit git repo with real `bazel query`. New code is ~97% line-covered. Docs: README "Query Service (experimental)" section, with the `serve` help wired into the generated CLI reference. Deferred to follow-ups: remote (S3) cache backend, OCI image, k8s manifests, load balancing / multi-instance, dynamic scaling, git auth config. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds a JGit-backed GitClient so the query service performs git fetch / resolve / checkout in-process, without forking a `git` subprocess or requiring a `git` binary on the PATH. - JGitClient implements the existing GitClient interface (drop-in). - `--gitEngine` selects the backend: `jgit` (default, in-process) or `subprocess` (shells out to --gitPath, exact C-git behavior, useful for workspaces relying on checkout filters/hooks JGit doesn't run). - Pins org.eclipse.jgit 5.13.x (the last Java-8-compatible line, matching bazel-diff's documented minimum) and repins maven_install.json. Note: "in process" removes the subprocess only; the working tree is still materialized on disk because `bazel query` reads real files. JGit's InMemoryRepository can't back a bazel workspace. Tests: JGitClientTest (resolve/checkout/fetch incl. remote loop and error paths against a real repo) + engine-selection tests; the serve E2E now exercises JGit by default. New code remains ~97% line-covered. Docs: README + generated serve help document --gitEngine and the on-disk working-tree caveat. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Implements #29 (RFC: Bazel Query Service) — the "bazel-diff as a service" model from the BazelCon talk that inspired this repo — as a first, self-contained increment: core service + per-SHA hash cache.
A new
bazel-diff servesubcommand runs the affectedness calculation as a long-running HTTP service. Hashes for each commit SHA are generated once and cached, so repeat queries are near-instant.What's included
New
com.bazel_diff.serverpackage +cli/ServeCommand.kt:BazelDiffServer) built on the JDK'scom.sun.net.httpserver— no new Maven dependency:GET /health→200once warm,503otherwise (load-balancer signal)GET /impacted_targets?from=<rev>&to=<rev>[&targetType=Rule,SourceFile]→ JSONHashCacheStoragebyte-interface +LocalDiskHashCacheStorage, atomic writes). Cache key folds in a fingerprint of the query-affecting flags, so a server started with different flags never serves another configuration's hashes. The interface is built to drop in a remote (S3) backend later.GitClient/ProcessGitClient): subprocessgitfetch / resolve / checkout, reusing the existingprocess()helper.HashService(per-SHA generate + cache, single workspace lock since one clone is shared) andImpactedTargetsService, which reusesCalculateImpactedTargetsInteractor— the exact logic behindget-impacted-targets— so results are identical to the CLI.git fetchgates readiness; a fetch failure "lame-ducks" the instance (health stays503) rather than attempting an in-place repair, per the RFC.Query-affecting flags (
--useCquery,--fineGrainedHashExternalRepos, etc.) mirrorgenerate-hashes, so the hashes the service produces match a cold CLI run.Tests & coverage
E2ETest#testServeEndToEnd) builds a two-commit git repo from the shell-onlydistance_metricsworkspace, runsservein a background thread, and drives/health+/impacted_targetsover real HTTP with realbazel query.ServeCommand99%); the change raises overall coverage above the 91% baseline and clears the 90% gate.Docs
README gains a "Query Service (experimental)" section, and the
servehelp is wired into the generated CLI reference (tools/generate_readme.py/readme_template.md/tools/BUILD).Deliberately deferred (follow-ups)
Remote (S3) cache backend, OCI/Docker image, k8s manifests, load balancing / multi-instance, dynamic scaling, and git auth-key config. These are the natural next increments of the RFC.
🤖 Generated with Claude Code