call-cache.md

Call cache

Reuse login and other setup in the same run, so those HTTP calls are not repeated for every case. The request under test still goes to the network.

Put cache: on an imported type: test file. The duration is yours to choose (5m, 1h, 30s, …). For example, cache: 5m means that calling this test again with the same inputs returns the same results until that TTL expires. The callee body, including its HTTP calls, does not run again. A call with different inputs runs normally and stores its own result. Caller checks still run.

type: test
title: Create session
cache: 5m
inputs:
  user: alice
  pass: secret
outputs:
  token: ''
import:
  login: ./api/login.mmt
steps:
  - call: login
    id: auth
    inputs:
      username: i:user
      password: i:pass
  - js: |
      outputs.token = auth.token

type: test
title: Use cached session
import:
  session: ./create_session.mmt
steps:
  - call: session
    id: first
    inputs:
      user: alice
      pass: secret
  # Same inputs within the TTL: same token, login is not called again.
  - call: session
    id: second
    inputs:
      user: alice
      pass: secret
    expect:
      token: == ${first.token}

This does not apply to type: api files. The cache is in memory for one top-level run (shared across suite siblings until the TTL expires). Direct Run of the cached file always executes the body.

Full rules: cache on tests. A working copy is in examples/intermediate/24_test_call_cache.

See also: call · Parallel runs · import