Skip to Content
ReferenceLimitations & Trade-offs

Limitations & Trade-offs

Every tool makes trade-offs. Here are the ones Glubean makes — and why.

Scratch Mode

  • .test.ts files show TypeScript errors — no node_modules means no type definitions. Use .test.js for scratch files. In a full project, .test.ts works normally with full type safety.
  • No .env / .env.secrets — environment variables and secrets aren’t loaded in scratch mode. Hardcode values or use process.env directly.
  • No configure() — shared HTTP clients need a project setup. Each test writes its own URL and headers.

Shared State & Execution Order

Tests in the same file share the module scope and run sequentially in export order. This is by design — it enables workflow-style tests where one test creates a resource and the next verifies it.

let userId: number; // ✅ declare at module level
 
export const createUser = test("create-user", async (ctx) => {
  const res = await ctx.http.post("https://api.example.com/users", {
    json: { name: "Test" },
  });
  userId = (await res.json()).id; // ✅ assign inside test
});
 
export const verifyUser = test("verify-user", async (ctx) => {
  const res = await ctx.http.get(`https://api.example.com/users/${userId}`);
  ctx.expect(res).toHaveStatus(200);
});

Rules:

  • Declare shared variables at module top level
  • Assign them only inside test callbacks, never at the top level
  • Export order = execution order — put dependencies first
  • Running a single test skips earlier tests, so shared variables keep their initial value

Why not parallel? Parallel execution within a file would break shared state. Files are the isolation boundary — different files can run in parallel (planned).

CodeLens Static Analysis

The extension uses regex-based static analysis (no runtime imports) to show CodeLens buttons. Some patterns can’t be resolved statically:

  • Nested generics lose pick key resolution: fromDir.merge<Record<string, T>>() may not show individual pick buttons — fromDir.merge<T>() works
  • Dynamic keys — computed variable names or conditional data loading won’t show pick buttons
  • require() syntax — only ES import is detected; CommonJS require is not

When static analysis fails, tests still run correctly — only the inline CodeLens buttons are affected. Use glubean run --pick <name> from the CLI as a fallback.

Template Resolution ({{KEY}})

In configure() HTTP options, {{KEY}} templates resolve from both vars and secrets (secrets checked first). There is no strict mode that limits resolution to secrets only.

export const api = configure({
  vars: { baseUrl: "{{BASE_URL}}" },           // reads from .env
  secrets: { apiKey: "{{API_KEY}}" },           // reads from .env.secrets
  http: {
    prefixUrl: "{{BASE_URL}}",
    headers: {
      Authorization: "Bearer {{API_KEY}}",      // checks .env.secrets first, then .env
    },
  },
});

If you put a secret in .env instead of .env.secrets, it will still work — but you shouldn’t. Secrets in .env are committed to git and visible in Cloud uploads. Always use .env.secrets for sensitive values.

Plugins Require Separate Install

Browser automation, auth strategies, GraphQL, and gRPC each require their own package:

npm install @glubean/auth      # bearer, apiKey, OAuth, basicAuth
npm install @glubean/browser   # Puppeteer-based browser tests
npm install @glubean/graphql   # GraphQL query/mutation helpers
npm install @glubean/grpc      # gRPC client

This keeps the core SDK lightweight — you only install what you use.

CLI Defaults to tests/ Only

glubean run without arguments only scans the tests/ directory. To run explore files:

glubean run --explore              # run all files in explore/
glubean run explore/smoke.test.ts  # run a specific explore file

This is intentional — explore/ is your interactive API collection, not something CI should run automatically.

Last updated on
Artifacts