Skip to main content

Chain

A chain is what a test handler returns: an ordered list of steps, written in the order a person would explain them, executed in whatever order their data needs allow. blong-chain is the executor behind group(name)([...]), and it is the reason a test file reads as a sequence while the run behaves as a graph.

Key behaviours​

  • Dependencies are observed, not declared. A step's context is a thenable proxy; reading a property records an edge to the step of that name and hands back a promise that resolves with its result. A step that touches no context runs immediately.
  • Everything is dispatched; the queue decides. All steps at a level are started, and the executor's concurrency limit (ten by default) bounds how many run at once. A step waiting on a dependency is waiting, not failing — the edge is a promise, not a schedule.
  • Groups nest. An inner array is a sub-test with its own name; an empty array is a barrier that waits for everything before it.
  • Names are checked. A duplicate step name, or a reference to a step that does not exist, fails the run and names the offender rather than producing a confusing timeout.
  • The run reports its own graph. After execution the executor can show the dependency graph, each step's queue time and execution time, the critical path and the parallel efficiency of the test.

See the test patterns for the step and context syntax, Cucumber for chains derived from .feature files, and the unified handler test rationale for why handlers and tests share one step model.