Using Test Sets

Why Tests Sets

Many operations such as running, comparing, removing or updating tests all need to somehow select which tests to operate on. To avoid having lots of hard-to-remember options, which may or may not interact well, typst-test offers an expression based set language which is used to select tests. Instead of writing

tt run --regex --mod 'foo-.*' --name 'bar/baz' --no-ignored

typst-test can be invoked like

tt run --expression '(mod(:foo-.*) & !ignored) | name(=bar/baz)'

This comes with quite a few advantages:

  • it's easier to compose multiple identifier filters like mod and name
  • options are ambiguous whether they apply to the next option only or to all options like --regex
  • with options it's unclear how to compose complex relations like and vs or of other options
  • test set expressions are visually close to the filter expressions they describe, their operators are deiberately chosen to feel like witing a predicate which is applied over all tests

Let's first disect what this expression actually means: (mod(:foo-.*) & !ignored) | id(=bar/baz)

  1. We have a top-level binary expression like so a | b, this is a union expression, it includes all tests found in either a or b.
  2. The right expression is id(=bar/baz), this includes all tests who's full identifier matches the given pattern =bar/baz. That's an exact pattern (indicated by =) for the test identifier bar/baz. This means that whatever is on the left of your union, we also include the test bar/baz.
  3. The left expression is itself a binary expression again, this time an intersection. It consists of another pattern test set and a complement.
    1. The name pattern is only applied to modules this time, indicated by mod and uses a regex matcher (indicated by :). It includes all tests who's module identifier matches the given regex.
    2. The complement !ignored includes all tests which are not marked as ignored.

Tying it all together, we can describe what this expression matches in a sentence:

Select all tests which are not marked ignore and are inside a module starting with foo-, include also the test bar/baz.

Trying to describe this relationship using options on the command line would be cumbersome, error prone and, depending on the options present, impossible. 1

Default Test Sets

Many operations take either a set of tests as positional arguments, which are matched exactly, or a test set expression. If neither are given the default test set is used, which is itself a shorthand for !ignored.

This may change in the future, commands my get their own, or even configurable default test sets. See #40.

More concretely given the invocation

tt list test1 test2 ...

is equivalent to the following invocation

tt list --expression 'none & (id(=test1) | id(=test2) | ...)'

An Iterative Example

Suppose you had a project with the following tests:

mod/sub/foo ephemeral  ignored
mod/sub/bar ephemeral
mod/sub/baz persistent
mod/foo     persistent
bar         ephemeral
baz         persistent ignored

and you wanted run only ephemeral tests in mod/sub. You could construct a expression with the following steps:

  1. Firstly, filter out all ignored tests, typst-test does by default, but once we use our own expression we must include this restriction ourselves. Both of the following would work.
    • default & ...
    • !ignored & ... Let's go with default to keep it simple.
  2. Now include only those tests which are ephemeral, to do this, add the ephemeral test set.
    • default & ephemeral
  3. Now finally, restrict it to be only tests which are in mod/sub or it's sub modules. You can do so by adding any of the following identifier matchers:
    • default & ephemeral & mod(~sub)
    • default & ephemeral & mod(=mod/sub)
    • default & ephemeral & id(:^mod/sub)

You can iteratively test your results with typst-test list -e '...' until you're satisfied. Then you can run whatever operation you want with the same expression. If it is a destructive operation, i.e. one that writes chaanges to non-temporary files, then you must also pass --all if your test set contains more than one test.

Scripting

If you build up test set expressions programmatically, consider taking a look at the built-in test set constants. Specifically the all and none test sets can be used as identity sets for certain operators, possibly simplifying the code generating the test sets.

Some of the syntax used in test sets may interfere with your shell, especially the use of whitespace. Use non-interpreting quotes around the test set expression (commonly single quotes '...') to avoid interpreting them as shell specific sequences.

1

To get a more complete look at test sets, take a look at the reference.