Commit graph

20 commits

Author SHA1 Message Date
Robin Malfait
a1246de6a5
WIP: New @apply logic
use `lightningcss` and `prettier` for `toMatchFormattedCss`

Lightning CSS applies some optimizations and prettier ensures that the
output is always formatted in the same way.

In a future version, we will be using Lightning CSS to apply
optimizations like handling the nesting, optimizing selectors, sorting
declarations, simplifying colors and so on.

To guarantee that we have the same output in both the normal version and
OXIDE version, we can apply lightningcss in our tests so that it always
produces correct results in both version with the same optimizations
applied.

This also aliases `toMatchCss` to `toMatchFormattedCss` so that we can
guarantee the same output. This has a small caveat where `lightningcss`
is pretty strict. It will complain about `@media screen(md) {}` for
example. This is correct because it is not valid CSS, but we eventually
take care of it.
In this case, we will catch the error and use prettier instead (since
this should only happen in tests that doesn't use end-to-end tests, but
uses a more specialized test instead). If prettier still finds an error,
we will ensure to throw the original error.

Eventually, all the "lower" level tests, should probably become higher
level end-to-end tests. But let's do everything step by step!
2023-01-05 16:20:24 +01:00
Jordan Pittman
2882d4c9ee
Fix ring color utility generation when using respectDefaultRingColorOpacity (#9070)
* Correct ring color list when using `respectDefaultRingColorOpacity`

* Update changelog
2022-08-10 10:36:04 -04:00
Jordan Pittman
6b1eb19079
Split ::backdrop into separate defaults group (#8567)
* Split `::backdrop` into separate defaults group

* Update tests

* Update changelog
2022-06-09 16:26:18 -04:00
Adam Wathan
4f400767a8
Add ::backdrop to universal defaults (#8526)
* Add `::backdrop` to universal defaults

* Update changelog
2022-06-06 15:53:08 -04:00
Jordan Pittman
3f4005e833 Disable filterDefault for ring opacity when using respectDefaultRingColorOpacity 2022-05-27 09:59:52 -04:00
Jordan Pittman
cab1fcecc5
Add future flag to preserve custom, default ring color opacity (#8448)
* Add future flag to preserve ring color opacity

* Update changelog
2022-05-27 09:48:27 -04:00
Jordan Pittman
d676086a75
Rewrite default class extractor (#8204)
* Rewrite default extractor

* Eliminate lookbehind assertions in expand apply at rules

* Update changelog
2022-05-04 16:08:25 -04:00
Jordan Pittman
7ed46b6f52 Re-enable new features
This reverts commits 1456ed9021, d0269c24b3, and 57699a0403.
2022-04-12 14:13:04 -04:00
Jordan Pittman
57699a0403
Prep for patch release (#8098)
* Disable backdrop variant

* Disable enabled variant

* Disable border spacing utilities

* Disable text-start/end utilities

* Disable poll option in CLI

* Disable ring color default fn

* Disable dark mode class name customization

* Disable support for `Document` node types

* Disable rgb/hsl fns

* Update tests

* Temporarily disable type generation

* Update changelog
2022-04-12 13:58:57 -04:00
Robin Malfait
48728ed5d3
Fix generation of div:not(.foo) if .foo is never defined (#7815)
* fix little typo

* ensure that `div:not(.unknown-class)` gets generated

* update changelog
2022-03-17 16:35:40 +01:00
Juan Martín Seery
dbb5b1d8f7
Added border-spacing utility (#7102)
* Added `border-spacing` utility

* Update to base CSS variable approach

Co-authored-by: Jonathan Reinink <jonathan@reinink.ca>
2022-03-04 11:20:23 -05:00
Jordan Pittman
04686b8a05
Split box shadows on top-level commas only (#7479)
* Split box shadows on top-level commas only

* Update changelog
2022-02-23 11:31:55 -05:00
Jordan Pittman
d72b277ba6
Allow default ring color to be a function (#7587)
* Allow default ring color to be a function

* Update changelog
2022-02-22 16:51:04 -05:00
Jordan Pittman
be5d5c9e66
Fix wildcard duplication issue (#7478)
This would be better as a symbol but the stringy-ness of class candidates is fairly well baked into assumptions across the codebase. Using `new String` with a well placed check seems to solve the problem.
2022-02-15 13:44:19 -05:00
Jordan Pittman
dc6644e0c8
Eliminate recursion from candidatePermutations (#7331) 2022-02-07 09:52:54 -05:00
Jordan Pittman
50802e1aed
Correctly parse shadow lengths without a leading zero (#7289)
* Correctly parse shadow lengths without a leading zero

* Update changelog

* Fix code style
2022-02-01 11:54:03 -05:00
Jordan Pittman
82f163d425
Fix use of falsy values in theme config (#6917) 2022-01-05 16:37:07 -05:00
Jordan Pittman
6cf3e3e55f
Don't mutate custom color palette when overriding per-plugin colors (#6546) 2021-12-16 16:19:54 -05:00
Robin Malfait
deb68d5816
Ensure all plugins are executed for a given candidate (#6540)
* remove early return so that all plugins are handled

We had an early return so that once a plugin was matched, that we could
stop running the code through the other plugins. However, in this case
we have an issue that user defined css is technically also a plugin.

This means that:
  - `bg-green-light`
Would check for:
  - `bg-green-light` (no hit, continue)
  - `bg-green` (Hit! Don't execute next plugins)
  - `bg` (This is the one that would have generated `bg-green-light`)

We tested this change and it doesn't seem to have an impact functionally
and also not really performance wise.

* update changelog
2021-12-15 18:07:26 +01:00
Robin Malfait
e37931ba65
JIT by default - move tests and make them consistent (#5374)
* move `./tests/jit` to `./tests`

* make tests consistent

Abstracted a `run` function and some syntax highlighting helpers for
`html`, `css` and `javascript`.
2021-09-03 13:48:16 +02:00
Renamed from tests/jit/basic-usage.test.js (Browse further)