Pre-launch — Gluecron is in final validation. Public signups and git hosting for non-owner users open after launch review.
CodeIssuesPull RequestsActionsSecurityInsightsSettings
✨ AI
More
claude/adoring-hopper-5x74bqclaude/affectionate-feynman-ykrf1hclaude/architecture-audit-design-wxprenclaude/build-status-update-3MXsfclaude/charming-meitner-mllb5rclaude/compare-gate-gluecron-s4mFQclaude/confident-faraday-tikcwbclaude/continue-work-XMTlIclaude/crontech-gluecron-deploy-7MIECclaude/crontech-platform-setup-SeKfwclaude/design-2026claude/ecstatic-ptolemy-jMdigclaude/enhance-github-integration-QNHdGclaude/fix-aa-loop-issue-PonMQclaude/fix-actions-and-processclaude/fix-desktop-errors-XqoW8claude/fix-red-workflowsclaude/fix-website-access-6FKJNclaude/gatetest-integration-hardeningclaude/github-audit-improvements-bDFr9claude/gluecron-launch-status-FoMRlclaude/hopeful-lamport-olfCTclaude/issue-to-pr-and-protectionsclaude/jolly-heisenberg-2sg1Qclaude/launch-preparation-QmTb6claude/new-session-xk1l7claude/plan-platform-architecture-kkN4yclaude/platform-analysis-roadmap-1nUGLclaude/platform-launch-assessment-8dWV8claude/polish-platform-release-AeDrUclaude/resume-previous-work-KzyLwclaude/review-crontech-handoff-qYEVqclaude/review-project-completeness-lHhS2claude/review-readme-docs-ulqPKclaude/serene-edison-rj87weclaude/setup-multi-repo-dev-BCwNQclaude/ship-fixes-and-tests-Jvz1cclaude/site-audit-competitive-pctlwgclaude/site-migration-vercel-XstpKclaude/standalone-product-repos-XHFTDcopilot/feat-smart-empty-states-keyboard-first-enhancementcopilot/feat-smart-morning-digest-review-context-restorecopilot/fix-and-process-workflowscopilot/update-ai-powered-code-reviewfeat/debt-mapfeat/push-policy-codeowners-hardeningfeat/smart-digest-contextfeat/stage-impactfeat/t1-secret-migrationfeat/u-polishfeat/w-self-hostfeat/w2-claude-configfix/agent-journey-orphan-sweepfix/audit-sweep-2026-07-26gatetest/auto-fix-1776586424172gatetest/auto-fix-1776586534814gatetest/auto-fix-1776590685143gatetest/auto-fix-1776590808199mainops/redeploy-retriggerstyle/dxt-cta-themeworktree-agent-a3377aad30d55da26worktree-agent-a7ef607b7ee1d6c74
concurrency.test.ts3.3 KB · 96 lines
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
/**
 * mapWithConcurrency — bounded async fan-out.
 *
 * The /issues and /pulls dashboards enriched their candidate lists with
 * `for (const c of candidates) { await enrich(c) }`, serialising one DB
 * round-trip per item: up to 600 sequential queries per /issues page load
 * (2 per issue × 300 candidates). Latency was N × RTT.
 *
 * A plain Promise.all would fix wall-clock but fire all N at once, which on a
 * pooled managed Postgres just relocates the queue into the connection pool.
 * These pin down both properties that matter: order preservation (the rows
 * are zipped back against `candidates` by index) and the concurrency cap.
 */

import { describe, expect, it } from "bun:test";
import { mapWithConcurrency, DB_FANOUT_LIMIT } from "../lib/concurrency";

const tick = () => new Promise((r) => setTimeout(r, 1));

describe("order preservation", () => {
  it("returns results in INPUT order, not completion order", async () => {
    // Deliberately invert the durations so completion order is the reverse
    // of input order. Zipping by index in the callers depends on this.
    const items = [1, 2, 3, 4, 5];
    const out = await mapWithConcurrency(items, 5, async (n) => {
      await new Promise((r) => setTimeout(r, (6 - n) * 5));
      return n * 10;
    });
    expect(out).toEqual([10, 20, 30, 40, 50]);
  });

  it("passes the index through", async () => {
    const out = await mapWithConcurrency(["a", "b", "c"], 2, async (v, i) => `${i}:${v}`);
    expect(out).toEqual(["0:a", "1:b", "2:c"]);
  });

  it("handles an empty list without spawning workers", async () => {
    let called = 0;
    const out = await mapWithConcurrency([], 8, async () => { called++; return 1; });
    expect(out).toEqual([]);
    expect(called).toBe(0);
  });
});

describe("concurrency is actually bounded", () => {
  it("never exceeds the limit in flight", async () => {
    let inFlight = 0;
    let peak = 0;
    await mapWithConcurrency(Array.from({ length: 50 }, (_, i) => i), 5, async () => {
      inFlight++;
      peak = Math.max(peak, inFlight);
      await tick();
      inFlight--;
      return null;
    });
    expect(peak).toBeLessThanOrEqual(5);
    expect(peak).toBeGreaterThan(1); // ...but it IS parallel, not serial
  });

  it("clamps the limit to the item count", async () => {
    let peak = 0;
    let inFlight = 0;
    await mapWithConcurrency([1, 2], 100, async () => {
      inFlight++;
      peak = Math.max(peak, inFlight);
      await tick();
      inFlight--;
      return null;
    });
    expect(peak).toBeLessThanOrEqual(2);
  });

  it("treats a zero or negative limit as serial rather than deadlocking", async () => {
    // A limit of 0 would spawn no workers and hang forever if unclamped.
    const out = await mapWithConcurrency([1, 2, 3], 0, async (n) => n);
    expect(out).toEqual([1, 2, 3]);
  });

  it("processes every item exactly once", async () => {
    const seen: number[] = [];
    await mapWithConcurrency(Array.from({ length: 30 }, (_, i) => i), 7, async (n) => {
      await tick();
      seen.push(n);
      return n;
    });
    expect(seen.length).toBe(30);
    expect(new Set(seen).size).toBe(30);
  });
});

describe("the shared limit", () => {
  it("is conservative enough for a pooled managed Postgres", () => {
    expect(DB_FANOUT_LIMIT).toBeGreaterThan(1);
    expect(DB_FANOUT_LIMIT).toBeLessThanOrEqual(20);
  });
});