fix: async GAV rebuild drops commits during in-flight rebuild

[robfrank]

Summary

• When GraphAnalyticalView uses ASYNCHRONOUS update mode, commit notifications arriving while a rebuild is already in progress were silently discarded, causing the view to miss newly committed data
• Added asyncRebuildNeeded flag so that a follow-up rebuild is automatically triggered after the current one completes
• Updated awaitReady() to loop until no pending rebuilds remain, ensuring callers see the fully settled state

Test plan

• GraphAnalyticalViewTest#asynchronousUpdateModeMultipleCommits — previously failing, now passes
• All 3 async-related tests pass
• Full GraphAnalyticalViewTest suite (121 tests) passes

🤖 Generated with Claude Code

[gemini-code-assist]

Summary of Changes

Hello, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request resolves a critical data consistency issue in GraphAnalyticalView when operating in asynchronous update mode. Previously, concurrent commits during a rebuild could lead to data loss in the view. The changes introduce a mechanism to queue follow-up rebuilds and ensure that clients waiting for the view to be ready will always receive a fully updated and consistent state, preventing data staleness.

Highlights

• Problem Addressed: Fixed an issue where GraphAnalyticalView in ASYNCHRONOUS update mode would silently discard commit notifications during an ongoing rebuild, leading to missed data.
• Follow-up Rebuilds: Introduced an asyncRebuildNeeded flag to ensure that if commits arrive during an asynchronous rebuild, a subsequent rebuild is automatically triggered upon completion of the current one.
• State Settlement: Enhanced the awaitReady() method to loop and wait until all pending rebuilds, including any triggered by asyncRebuildNeeded, have completed, guaranteeing that callers observe a fully settled view state.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Changelog

• engine/src/main/java/com/arcadedb/graph/olap/GraphAnalyticalView.java
  • Introduced a new asyncRebuildNeeded volatile boolean flag to track if a commit occurred during an ongoing asynchronous rebuild.
  • Modified the awaitReady method to repeatedly check the view's status and the asyncRebuildNeeded flag, ensuring all necessary rebuilds complete before returning.
  • Updated the onRelevantCommit method to set asyncRebuildNeeded when a rebuild is already in progress, and to trigger a new rebuild if this flag is set after the current rebuild finishes.

Activity

• The author, robfrank, created this pull request.
• The changes address a previously failing test case: GraphAnalyticalViewTest#asynchronousUpdateModeMultipleCommits.
• All three async-related tests now pass.
• The full GraphAnalyticalViewTest suite (121 tests) passes.
• The PR description indicates it was generated with Claude Code.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[mergify]

🧪 CI Insights

Here's what we observed from your CI run for df1cc4d.

🟢 All jobs passed!

But CI Insights is watching 👀

[claude]

Code Review

The core idea — using an asyncRebuildNeeded boolean flag to coalesce commits-during-rebuild into a single follow-up rebuild — is clean and correct. Good test added. Below are a few issues worth addressing.

---

Bug: awaitReady busy-spins on build failure (STALE status)

The new while loop exits only when status == READY && !asyncRebuildNeeded, but it never exits early for STALE. When a rebuild fails the code path is:

} catch (final Exception e) {
    this.buildError = e;
    this.status = Status.STALE;    // status ≠ READY
    ...
} finally {
    ...
    latch.countDown();             // already-counted-down latch
}

After latch.countDown(), awaitReady wakes up and evaluates:

while (status != Status.READY || asyncRebuildNeeded)   // true: STALE ≠ READY

It then re-reads the latch — which is already at count 0 — so latch.await(...) returns true instantly, and the loop spins until the deadline. That is a regression vs. the original code, which would return false promptly once the latch was counted down.

Suggested fix — add an early exit inside the loop:

if (status == Status.STALE)
    return false;

Or change the loop guard to:

while (status == Status.BUILDING || (status == Status.READY && asyncRebuildNeeded))

---

Concurrency ordering in the finally block

compacting.set(false);
latch.countDown();             // (A) wakes awaitReady
if (asyncRebuildNeeded)
    onRelevantCommit();        // (B) sets new latch

Between (A) and (B) there is a tiny window where awaitReady sees asyncRebuildNeeded == true but readyLatch still points to the already-counted-down old latch. Since latch.await() on a zero-count latch returns immediately, awaitReady will tight-spin for the short time it takes onRelevantCommit() to create the new latch. This is not a correctness bug — just a micro-busy-wait — but it could be avoided by calling onRelevantCommit() before latch.countDown(). However, changing that order requires verifying that onRelevantCommit() under the same lock won't deadlock, so it is a trade-off.

---

Minor: asyncRebuildNeeded read without the lock in the async thread

// finally block — NOT inside synchronized(GraphAnalyticalView.this)
if (asyncRebuildNeeded)
    onRelevantCommit();

asyncRebuildNeeded is volatile, so visibility is guaranteed. The synchronized keyword on onRelevantCommit() ensures proper locking when it actually executes. This is fine as-is, but a quick comment explaining the deliberate lock-free read would help readers.

---

Test: consider asserting getNodeCount defensively

The test hard-codes a result of exactly 6 nodes after 5 rapid commits. Because the loop uses sequential commits in the same thread and awaitReady returns only after all rebuilds settle, the count should be deterministic, so this is fine. Just noting it explicitly in the test comment (e.g. "exactly 6: Alice + P0..P4") already present — good.

---

Summary

Severity	Item
Bug	awaitReady busy-spins until timeout when a rebuild fails (STALE status)
Minor	Micro-busy-wait window between latch.countDown() and new latch creation
Nit	Missing comment on intentional volatile read of asyncRebuildNeeded outside lock

The STALE spin-until-timeout is the only change I'd consider required before merging. Everything else looks solid.

[gemini-code-assist]

Code Review

This pull request correctly addresses an issue where commits were dropped during an asynchronous GAV rebuild by introducing a flag to queue a subsequent rebuild. The logic for chaining rebuilds appears sound. However, the updated awaitReady() method introduces a potential concurrency issue: it can enter a busy-wait loop, consuming high CPU, while waiting for the next rebuild to be scheduled. I've suggested a fix to address this by yielding the thread in that specific transient state.

[gemini-code-assist]

The updated awaitReady method has a potential for a busy-wait spin. If a rebuild completes and another one is needed (asyncRebuildNeeded is true), this method can loop continuously on an already counted-down latch until the next rebuild is scheduled by another thread. This can lead to high CPU usage.

To fix this, we can check if the latch is already open (getCount() == 0). If it is, and the waiting condition is still met, we should Thread.yield() to prevent a tight loop and allow the other thread to schedule the next build.

  public boolean awaitReady(final long timeout, final TimeUnit unit) {
    final long deadlineNanos = System.nanoTime() + unit.toNanos(timeout);
    try {
      // Loop to handle follow-up rebuilds triggered by asyncRebuildNeeded.
      // Each rebuild creates a new latch, so we re-read the field after each wait.
      while (status != Status.READY || asyncRebuildNeeded) {
        final long remainingNanos = deadlineNanos - System.nanoTime();
        if (remainingNanos <= 0)
          return false;

        final CountDownLatch latch = readyLatch;
        if (latch == null)
          return status == Status.READY && !asyncRebuildNeeded;

        // If latch is already open, we are in a transient state waiting for the next build.
        // Yield to avoid a busy-wait spin while the other thread sets up the new latch.
        if (latch.getCount() == 0) {
          Thread.yield();
          continue;
        }

        if (!latch.await(remainingNanos, TimeUnit.NANOSECONDS))
          return false;
      }
    } catch (final InterruptedException e) {
      Thread.currentThread().interrupt();
      return false;
    }
    return true;
  }

[robfrank]

Addressed review feedback in df1cc4d:

Fixed:

• STALE busy-spin bug (flagged by both reviewers): awaitReady now exits immediately with false when status == STALE, preventing spin on a zero-count latch after a failed rebuild.
• Volatile read comment (Claude bot nit): Added comment explaining the intentional lock-free read of asyncRebuildNeeded in the finally block.

Not addressed (by design):

• Micro-busy-wait between latch.countDown() and new latch creation: This is a transient nanosecond-scale window that resolves itself once onRelevantCommit() creates the new latch. Thread.yield() (Gemini's suggestion) adds complexity for no practical benefit. Reordering to call onRelevantCommit() before latch.countDown() (Claude bot's alternative) risks deadlock since onRelevantCommit() is synchronized.

All 121 GraphAnalyticalViewTest tests pass.

[codacy-production]

Coverage summary from Codacy

See diff coverage on Codacy

Coverage variation	Diff coverage
✅ -9.67%	✅ 88.89%

Coverage variation details

	Coverable lines	Covered lines	Coverage
Common ancestor commit (024b303)	109700	81927	74.68%
Head commit (5120447)	140655 (+30955)	91443 (+9516)	65.01% (-9.67%)

Coverage variation is the difference between the coverage for the head and common ancestor commits of the pull request branch: <coverage of head commit> - <coverage of common ancestor commit>

Diff coverage details

	Coverable lines	Covered lines	Diff coverage
Pull request (#3670)	18	16	88.89%

Diff coverage is the percentage of lines that are covered by tests out of the coverable lines that the pull request added or modified: <covered lines added or modified>/<coverable lines added or modified> * 100%

See your quality gate settings    Change summary preferences