Noodle 0.9.3 brings response tests, request chaining, and repeated data runs to the same YAML collections you use for manual requests. Scripts and tests can now live at collection or folder level, so common preparation and checks do not need to be copied into every request.

Check more than one value

The new inline tests field runs after declarative assertions. Use test to name a check and expect to describe what should hold:

tests: |-
  test("user response matches its contract", () => {
    expect(noodle.response.status).toBe(200);
    expect(noodle.response.json()).toMatchSchema({
      type: "object",
      required: ["id", "email"],
      properties: {
        id: { type: "integer" },
        email: { type: "string", format: "email" }
      }
    });
  });

Matchers cover equality, containment, regular expressions, numeric comparisons, literal properties, lengths, types, partial objects, and JSON Schema draft-07. Each supports .not. Schema compilation and validation stay inside the sandbox, with standard formats and local references. They do not coerce values, apply defaults, or fetch external schemas.

Callbacks can be synchronous or async. Results retain declaration order, and a failed check does not prevent later tests from running. Tests read the final request, response, and run variables; they cannot mutate state or send HTTP. Failed tests and top-level test errors make automation exit nonzero and count toward fail-fast. Simple contracts can keep using declarative assert.

Await another request

Pre and post scripts now support top-level await. Call a saved request with noodle.runRequest, then use its response to prepare the current request:

scripts:
  pre: |-
    const login = await noodle.runRequest("auth/login");
    const token = login.json().token;
    noodle.run.set("TOKEN", token);
    noodle.request.headers.set("Authorization", `Bearer ${token}`);

This example expects an existing auth/login.yml that returns a token. Saved calls run their own scripts, captures, assertions, and tests. For a literal HTTP call, use noodle.sendRequest with a URL and optional method, headers, body, or timeout.

Successful child variables reach the parent, and the combined changes commit only when the enclosing script succeeds. Child persistence stays transient. HTTP effects and received cookies cannot be undone by variable rollback. The current request has already passed substitution, so apply newly obtained values with request setters as shown above.

Await each call before starting another. Calls share a limit of ten per top-level request and four child levels, with cancellation and ancestor wall deadlines. The 500 ms VM execution budget excludes network waiting; each invocation has a 30-second wall limit. Raw fetch, timers, imports, and background work remain unavailable.

Share setup and checks

Put common scripts.pre, scripts.post, or tests in collection settings.yml or a nested folder.yml. Every phase runs collection blocks, then outer-to-inner folder blocks, then the request block.

Each block has a fresh sandbox. Share values through noodle.run, since JavaScript locals do not cross blocks. Empty blocks do not disable inherited ones. A pre failure stops the send; post and top-level test errors retain later blocks, with each error linked to its originating file.

Repeat a selection for each data row

Pass a CSV or JSON dataset to the CLI:

noodle collection run ./my-api users/ --data ./data/users.json --fail-fast --json

A JSON file contains a non-empty array of objects:

[{ "user_id": 1, "active": true }, { "user_id": 2, "active": false }]

Each row runs the selected requests in collection order. Row values override environment variables for substitution and remain typed in JSON; CSV cells are strings. Scripts and tests can read the original row through the frozen noodle.iteration object, alongside its zero-based index and total count.

Every row starts with fresh variables and an in-memory copy of the initial cookie jar. Changes flow between requests within that row and disappear before the next row starts. The whole file is validated before HTTP, with limits of 5 MiB, 1,000 rows, and 10,000 selected request executions.

In the F5 Runner, use the Data file field, path completion, and iteration preview before choosing Run. CLI relative paths start at your current directory; Runner relative paths start at the collection root. Results group requests by iteration, and delay and fail-fast apply across row boundaries.

Follow the result back to its source

Results, Runner details, CLI output, and manual history now include test outcomes, inherited source labels, and child-call summaries. Diagnostics stay bounded and redact known secrets; history excludes child bodies, script source, and runtime values.

Test names and durations stay aligned as the pane resizes. Top-level test errors remain visible even when every completed test passed, with all block errors available in expanded details. URL credential and sensitive-query redaction also preserves valueless query parameter names.

The updated noodle-use skill documents authoring and automation, while noodle-dev covers the shared lifecycle and regression checks. Start with Scripted Tests, Inline Scripting, or the Collection Runner guide for complete examples and limits.