Legacy Pkg.test Integration

TestItemRunner.jl integrates test items with the standard Pkg.test() workflow, so your @testitems run from the usual test/runtests.jl entry point.

Which should I use?

For running tests yourself — locally or in CI — prefer the juliati command line runner. It is faster to invoke, runs test items in parallel by default, and gives you per-item results.

Pkg.test integration still matters when something else insists on calling Pkg.test(): package registries and their automated checks, downstream integration testing, ] test habits, and any CI you have not migrated. Keeping a test/runtests.jl around costs you nothing and makes your package behave like every other Julia package.

The two approaches are not exclusive. The same @testitem blocks work under both, so a package can ship a runtests.jl for compatibility while you use juliati day to day. They also build the same kind of test environment — see Environments for what test items run in outside of Pkg.test.

Setup

1. Add TestItemRunner.jl as a test dependency:

   # In your package directory
   using Pkg
   Pkg.activate("test")
   Pkg.add("TestItemRunner")

2. Create or update test/runtests.jl:

   using TestItemRunner
   
   @run_package_tests

Now Pkg.test() will discover and run all @testitems in your package.

Filtering

Pass a filter function to @run_package_tests:

using TestItemRunner

# Skip tests tagged :skipci
@run_package_tests filter=ti->!(:skipci in ti.tags)

The filter function receives a named tuple with metadata about each test item:

• filename — full path of the file where the test item is defined
• name — the name of the test item
• tags — a Vector{Symbol} of tags

You can write arbitrarily complex filter conditions:

# Run only tests tagged :fast
@run_package_tests filter=ti->(:fast in ti.tags)

# Run tests from a specific file
@run_package_tests filter=ti->(endswith(ti.filename, "test_foo.jl"))

# Combine conditions
@run_package_tests filter=ti->(!(:skipci in ti.tags) && endswith(ti.filename, "test_foo.jl"))

Note that this is a Julia function taking a named tuple, whereas juliati --filter takes an expression over the same variables — the concepts match, but the syntax differs.