Add tests for pattern selection #863
Conversation
| - `never` (default) | ||
| - `select` | ||
| - `format` | ||
| - `always` |
There was a problem hiding this comment.
are these sufficiently self-documenting?
There was a problem hiding this comment.
They're somewhat documented by the behaviour section below. Not explicitly documenting these is also in line with what we've done for all the other functions' option values.
| @@ -209,173 +209,6 @@ | |||
| } | |||
| ] | |||
| }, | |||
| { | |||
| "src": ".match {$foo :number} one {{one}} * {{other}}", | |||
There was a problem hiding this comment.
This file tests the function :number (and its friend :integer, which are specified as part of the standard function set*. Won't we need these tests for that purpose? The file is named appropriately for that.
There was a problem hiding this comment.
The tests removed from here rely on the category selection behaviour of :number, which is not strictly speaking required of an implementation, as we only say that "selection should be based on CLDR plural rule data". If that wiggle room is not intentional, then we should (separately from this PR) firm that up, and possibly introduce more specific tests for :number that check a wider range of locales.
These tests also mostly exercise the MF2 spec side of pattern selection, for which each case is now included in the :test:select tests.
There was a problem hiding this comment.
We probably should fix these tests to use CLDR data, if that's what we want. My primary concern is not to lose tests for something that is part of our collection of specifications, even if it isn't part of MF2's core spec.
There was a problem hiding this comment.
If we keep these :number tests in as they are, we're requiring that a "should" in the spec is treated the same as a "MUST". If we do want to test this behaviour, the tests for it ought to be separate and somehow marked so that an implementation that chooses to do something else can ignore them.
The spec-required parts of these tests is completely retained in the :test:select tests added in this PR.
I added #868 to continue the conversation about the spec language for :number and :integer selection.
Co-authored-by: Addison Phillips <[email protected]>
Currently, the tests we have for pattern selection rely on the behaviour of
:number, and are somewhat incomplete. We should instead use:test:selectand:test:formatfor this, and work to ensure more aspects are covered.A new
failsoption is added to the the test functions, to allow for the changes made in #828 to be testable (i.e. what happens if a custom selector function throws an error during MatchSelectorKeys).The
.matchtests on:numberare all moved to the new test suite.