Unfortunately for backwards compatibility purposes (with `loadConfig` at
least) we can't switch things to use `import(…)` because there's baked
in knowledge that the config is loaded synchronously for v3.
This PR does two things:
- Defers to `require(…)` which allows newer versions that support
`require(esm)` to work natively. This works around the need to switch to
`import(…)` for those versions.
- Allows newer versions of `postcss-load-config` enabling better
ESM+TypeScript support for PostCSS configs in the CLI.
We support v4, v5, and v6 of `postcss-load-config` simultaneously so any
of those versions should work. I've verified that newer node versions
seem to install v6 while earlier ones like Node v14 install v4 of
`postcss-load-config`. So this should be a backwards compatible change.
- [x] needs tests for `import.meta.resolve(…)`
- [x] needs tests for ESM postcss configs
Fixes#14152Fixes#14423
This works around an issue in Chrome where `::selection` does not read
from variables defined on `::selection` thus breaking all uses of color
utilities with the selection variant. e.g. `selection::bg-red-200`.
We now add a fallback value of `1` to all uses of
`var(--tw-bg-opacity)`, `var(--tw-text-opacity)`,
`var(--tw-border-opacity)`, etc… since Chrome treats the variable as if
it did not exist because it's not defined on the parent.
In Chrome 131 (until the change is rolled back) existing utilities like
these will not work:
- `selection:text-opacity-50`
- `selection:[--tw-text-opacity:0.5]`
Fixes#15000
---------
Co-authored-by: Adam Wathan <[email protected]>
We bundle cssnano in our CLI and it's on an older version that does not
handle nested CSS correctly. This PR updates it to the latest version of
v6.x
Fixes#14092
* add `tailwind.config.cts` and `tailwind.config.mts` as default config files
* always use jiti when working with ESM or TS files
* update changelog
* add integration test for `.cts` and `.mts` configuration files
2024-07-03 23:09:35 +02:00
Robin MalfaitGitHubdepfu[bot] <23717796+depfu[bot]@users.noreply.github.com>
* remove all oxide related code
* Update lightningcss to version 1.24.1
* update tests to match bumped Lightning CSS output
---------
Co-authored-by: depfu[bot] <23717796+depfu[bot]@users.noreply.github.com>
* replace `env.OXIDE` with global `__OXIDE__`
This will allow us to replace the `__OXIDE__` at build time, and fully
remove the branches from the final code so that there is not even any
reference to `@tailwindcss/oxide` on the stable engine.
* update changelog
* use `env.ENGINE` in integration tests
* drop oxide branching for the PostCSS plugin for now
This is currently a redirect to the same file, so doesn't hurt.
* Enable better dead-code elimination
* Update CLI tests
Fix indentation
* Fix indentation
---------
Co-authored-by: Jordan Pittman <[email protected]>
* try to use `config.default` before using `config`
* update changelog
* add quick `SHOW_OUTPUT` toggle for integration tests
Setting this to `true` shows the output of the executed commands.
* add integration tests for `tailwind.config.ts` and `tailwind.config.js` with ESM syntax
* add `jiti` and `detective-typescript` dependencies
* use `jiti` and `detective-typescript`
Instead of `detective`, this way we will be able to support
`tailwind.config.ts` files and `ESM` files.
* use `@swc/core` instead of the built-in `babel` form `jiti`
* update changelog
* add `jiti` and `detective-typescript` dependencies to `stable`
* use `sucrase` to transform the configs
* add `sucrase` dependency to `stable` engine
* make loading the config easier
* use abstracted loading config utils
* WIP: make `load` related files public API
* use new config loader in PostCSS plugin
* add list of default config files to look for
* cleanup unused arguments
* find default config path when using CLI
* improve `init` command
* make eslint happy
* keep all files in `stubs` folder
* add `tailwind.config.js` stub file
* Initialize PostCSS config using the same format as Tailwind config
* Rename config content stubs to config.*.js
* Improve option descriptions for init options
* Remove unused code, remove `constants` file
* Fix TS warning
* apply CLI changes to the Oxide version
* update `--help` output in CLI tests
* WIP: make tests work on CI
TODO: Test all combinations of `--full`, `--ts`, `--postcss`, and `--esm`.
* wip
* remove unused `fs`
* Fix init tests
Did you know you could pass an empty args to a command? No? Me neither. ¯\_(ツ)_/¯
* bump `napi-derive`
* list extensions we are interested in
* no-op the `removeFile` if file doesn't exist
* ensure all `init` flags work
* ensure we cleanup the new files
* test ESM/CJS generation based on package.json
* remove unnecessary test
We are not displaying output in the `--help` anymore based on whether
`type: module` is present or not.
Therefore this test is unneeded.
* only look for `TypeScript` files when the entryFile is `TypeScript` as well
* refactor `load` to be `loadConfig`
This will allow you to use:
```js
import loadConfig from 'tailwindcss/loadConfig'
let config = loadConfig("/Users/xyz/projects/my-app/tailwind.config.ts")
```
The `loadConfig` function will return the configuration object based on
the given absolute path of a tailwind configuration file.
The given path can be a CJS, an ESM or a TS file.
* use the `config.full.js` stub instead of the `defaultConfig.stub.js` file
The root `defaultConfig` is still there for backwards compatibilty
reasons. But the `module.exports = requrie('./config.full.js')` was
causing some problems when actually using tailwindcss.
So dropped it instead.
* apply `load` -> `loadConfig` changes to `Oxide` engine CLI
* ensure we write the config file in the Oxide engine
* improve type in Oxide engine CLI
* catch errors instead of checking if the file exists
A little smaller but just for tests so doesn't matter too much here 👍
* ensure we publish the correct stub files
---------
Co-authored-by: Adam Wathan <[email protected]>
Co-authored-by: Jordan Pittman <[email protected]>
Co-authored-by: Nate Moore <[email protected]>
Co-authored-by: Enzo Innocenzi <[email protected]>
* disable color opacity plugins by default for the `oxide` engine
* update tests to reflect this change in the `oxide` engine
* update changelog
* reflect changes in integration tests
* bump lightningcss
* use `lightningcss` in the main PostCss Plugin
* use lightningcss in our custom matchers
Now that we are using `lightningcss` and nesting in the new `oxide`
engine, the generated output _will_ be different in the majority of test
cases.
Using a combination of `prettier` and `lightningcss` will make the
output consistent.
The moment we are fully using the `oxide` engine, we can drop
`lightningcss` or `prettier` again to improve the performance of the
tests.
* update tests to apply `lightningcss` related changes
* update changelog
* add `lightningcss` and `browserslist` as dev dependencies to stable package.json
* only use `lightningcss` in tests (without prettier)
We will only fallback to prettier if lightningcss fails somehow.
* apply side effect chagnes due to only using lightningcss for tests
* make CI happy (integration tests)
Apply changes to integration tests now that we are using lightningcss
* transform `lightningcss` for Node 12 when running tests
* run prettier on failing tests for `toMatchFormattedCss`
This will result in better diffs because diffs are typically per block
and/or per line. But lightningcss will simplify certain selectors and
the diff won't be as clear.
We will only apply the prettier formatting for failing tests in the diff
view so that diffs are cleaner and we don't pay for the additional
prettier calls when tests pass.
* Record and watch PostCSS dependencies in the CLI
* ensure `changedContent` gets cleared
Otherwise this list gets bigger and bigger, not only that there is a
subtle bug. The moment you save a `.css` file we want to create a new
context and start from scratch. However, since the list was never
cleared, it meant that every subsequent save to *any* file (not only
config / css files) creates a new context...
By clearing the least we should work around this problem.
* add test that verifies an odd bug
The story goes like this:
1. add `underline` to html file
-> css contains `underline` rule
2. add `font-bold` to html file
-> css contains `underline` and `font-bold`
3. remove `underline` from html file
-> css still contains `underline` and `font-bold` for performance reasons
4. Save a css file (! RED FLAG)
-> css contains `font-bold` because we started from scratch
5. add `underline` to html file
-> css contains `underline` and `font-bold`
6. remove `underline` from html file
-> css only contains `font-bold`... (UH OH)
This is because the moment we did step 4, every single save in any file created a new context. Every. Single. Time.
* use a property that doesn't require `autoprefixer`
* update changelog
Co-authored-by: Jordan Pittman <[email protected]>
* update changelog
* ensure `--content` is taken into account
* cleanup tests
- Use `rm` instead of deprecated `rmdir`
- Type the returnType correctly
* use a file not included in `content` of your tailwind.config.js file
* Add support for postcss-import in watch mode
* Add regression test
* Extract shared logic
* restructure test a little bit
Instead of relying on a arbitrary setTimout value, let's wait for the
file to be created instead.
* update changelog
Co-authored-by: Adam Bergman <[email protected]>
* Add content glob integration tests
* Use tagged version in parcel integration tests
* Upgrade postcss-cli integration tests
* upgrade integration test deps
* Fix CS
* Fix tests
* Update lockfile
* Fix vite test in CI
* add generate-types script
This script will generate the full list of core plugins, which will
allow you to get code completion for the `corePlugins` section.
It will also generate all the colors (and deprecated colors) which is
used in multiple places in the config.
* add types for the `tailwind.config.js` config file
* annotate stubs with a JSDoc pointing to the types
* add types to package.json
- Updated the files to make sure that the types are being published
- Add a `types` section in the `package.json`, otherwise your editor by
default will look for the `DefinitelyTyped` types which got me really
confused for a second.
- Added some scripts to make sure that the generation of types happens
when needed (before tests and before building). This way you never
ever have to think about generating them when working on Tailwind CSS
internals.
* re-export types top-level
Having a `colors.d.ts` next to the `colors.js` file allows us to type
the `colors.js` file and your editor will pickup the types from
`colors.d.ts`.
* also publish generated types
* update changelog
* enable TypeScript only when using `init --types` for now
* update tests to verify that `--types` works
* add prettier-plugin-tailwindcss
This will use the prettier plugin in our tests as well, yay consistency!
* ensure that both `group` and `peer` can't be used in `@apply`
This was only configured for `group`
* expose `sortClassList` on the context
This function will be used by the `prettier-plugin-tailwindcss` plugin,
this way the sorting happens within Tailwind CSS itself adn the
`prettier-plugin-tailwindcss` plugin doesn't have to use internal /
private APIs.
The signature looks like this:
```ts
function sortClassList(classes: string[]): string[]
```
E.g.:
```js
let sortedClasses = context.sortClassList(['p-1', 'm-1', 'container'])
```
* update changelog
* add sort test for utilities with the important modifier e.g.: `!p-4`
* use outputFile instead of direct writeFile
This is an improvement we introduced earlier but forgot this part.
* allow to pipe in data to the CLI
* add integration tests to validate piping to the CLI
* update changelog
* use String.raw for css escapes
This will allow us to write code like:
```css
.mobile\:font-bold {}
```
Instead of
```css
.mobile\\:font-bold {}
```
Which resembles "real" css way better in our tests.
* use String.raw in integration tests as well
* enabled `optimizeUniversalDefaults` by default
This PR is done in a way so that the default is set to `true`, but you
can still disable it if it causes issues. In this case we do appreciate
an issue in that case 😅.
* update tests to use optimized universal selector
* update integration tests
* add dedicated tests for the optimized universal selector
* improve minimumImpactSelector algorithm
I think I cracked the algorithm, but I will probably need another pair
of eyes on the subject.
The current implementation works like this:
Prerequisites:
- The selector should already have been parsed using the selectorParser
from 'postcss-selector-parser'.
Algorithm:
1. Remove all of the pseudo classes from the list of nodes.
1.1. We do want to keep pseudo elements (E.g.: `::before`, `::first-line`, ...)
1.2. We do want to keep pseudo classes that contain nodes (E.g.:
`:not(...)`)
2. Reverse the list of nodes.
This will make it easier to search from the end to the start. For
example `.group:hover .group-hover` should result in `.group-hover`
not `.group`.
2.1. Find the index of the best match (class, id, attribute), and
convert the node if required. (E.g.: `span#app` -> `#app` => `[id="app"]`)
2.2. Remove the rest of the selector that is not important anymore
2.3. Re-join the left-over nodes together
* update tests using new algorithm
* also look for `tag` types
* take `tag` into account
* simplify logic
* add test to prove `rest.reverse()` in first case is required
In case we don't find a match (idx === -1), we use `rest.reverse()`.
However, it looks like you can just use `nodes` instead.
This is not entirely true, because the `rest` variable will contain only
the nodes that are not pseudo elements.
`*:hover` would result in `*:hover` instead of just `*`
* replace all nodes after > with a single universal selector
* immediately take the `safelist` values into account
Currently we had to manually add them in the `setupTrackingContext`,
`setupWatchingContext` and the `cli`.
This was a bit cumbersome, because the `safelist` function (to resolve
regex patterns) was implemented on the context. This means that we had
to do something like this:
```js
let changedContent = []
let context = createContext(config, changedContent)
for (let content of context.safelist()) {
changedContent.push(content)
}
```
This just feels wrong in general, so now it is handled internally for
you which means that we can't mess it up anymore in those 3 spots.
* drop the dot from the extension
Our transformers and extractors are implemented for `html` for example.
However the `path.extname()` returns `.html`.
This isn't an issue by default, but it could be for with custom
extractors / transformers.
* normalize the configuration
* make shared cache local per extractor
* ensure we always have an `extension`
Defaults to `html`
* splitup custom-extractors test
* update old config structure to new structure
* ensure we validate the "old" structure, and warn if invalid
* add tests with "old" config, to ensure it keeps working
* add missing `content` object
* inline unnecessary function abstraction
* setup integration tests
* fix rgb color syntax
* ensure integration tests always exit
If for any reason the integration tests fail, then it will run forever
on CI (~2hours or something). The `--forceExit` is not ideal but it will
prevent long running processes.
* fix incorrect test
We were never properly waiting for the command to finish.
* handle AbortError properly
In CI, when an AbortController gets aborted an error is thrown
(AbortError). If we don't catch it properly then it will "leak" and the
test will fail.
* improve IO functions
* quit integration tests after 10seconds
* only test a few integrations
* test all integrations using matrix
This will cancel other builds when one fails, it will also separate the
output per integration which can be useful especially now that we are
still figuring things out.
* rename `build` to `test`
* add --verbose flag to receive output in the console
* when reading stdout or stderr, wait a certain about to ensure stability
Debouncing for 200ms means that if another message comes in within those
200ms we delay the execution of the callback.
* simplify workflow
* use terminal output instead of disk events
* cache node_modules for integrations
* empty commit, to test cache hits
* make `jit` mode the default when no mode is specified
* unify JIT and AOT codepaths
* ensure `Object.entries` on undefined doesn't break
It could be that sometimes you don't have values in your config (e.g.: `presets: []`), this in turn will break some plugins where we assume we have a value.
* drop AOT specific tests
These tests are all covered by JIT mode already and were AOT specific.
* simplify tests, and add a few
Some of the tests were written for AOT specifically, some were missing. We also updated the way we write those tests, essentially making Tailwind a blackbox, by testing against the final output.
Now that JIT mode is the default, this is super fast because we only generate what is used, instead of partially testing in a 3MB file or building it all, then purging.
* add some todo's to make sure we warn in a few cases
* make `darkMode: 'media'`, the default
This also includes moving dark mode tests to its own dedicated file.
* remove PostCSS 7 compat mode
* update CLI to be JIT-first
* fix integration tests
This is not a _real_ fix, but it does solve the broken test for now.
* warn when using @responsive or @variants
* remove the JIT preview warning
* remove AOT-only code paths
* remove all `mode: 'jit'` blocks
Also remove `variants: {}` since they are not useful in `JIT` mode
anymore.
* drop unused dependencies
* rename `purge` to `content`
* remove static CDN builds
* mark `--purge` as deprecated in the CLI
This will still work, but a warning will be printed and it won't show up
in the `--help` output.
* cleanup nesting plugin
We don't have to duplicate it anymore since there is no PostCSS 7
version anymore.
* make sure integration tests run in band
* cleanup folder structure
* make sure nesting folder is available
* simplify resolving of purge/content information
* update (old) incorrect snapshots
The version changed, so it is save to update these.
* add `shadow` so that we can test `@tailwind base`
We were only using `font-bold`, but that doesn't require any base rules.
We try to only inject rules that are required, even for `base` related
rules.
Adding `shadow` will ensure that we can test that `base` rules have been
added.
* use `~/.cache/tailwindcss` as default touch dir
When more tools catch up with the latest postcss dependency tracking,
then we probably don't even need this anymore. However, it is a good
practice to put (global) caches in ~/.cache and (global) configs in
~/.config
Fixes: #4569
* change test to verify @tailwind base slightly
Because of some changes in tailwind itself, the base selector contains a
ton of new --tw-* custom properties. Keeping this up to date might be a
bit of an annoying challenge. So instead we now check for the selector
to be present.
* add `tmp` package
* use a (simple) temporary file, instead of a full cache dir
* add tests for the --postcss option in the new CLI
* add `oneOf` ability to the `arg()` functionality
By default, `arg()` doesn't have a way to define multiple types. We want
the possibility of using `--postcss` (Boolean) or `--postcss
./custom-path.js`. But by default this is not possible.
This commit will allow us to do a few things, mainly:
- Keep the same API using the `{ type: oneOf(String, Boolean), description: '...' }`
- Keep the `--help` output similar
What we did behind the scenes is make sure to put the non recognized
flags in the `_` arguments list. This is possible by doing `permissive:
true`. We then manually parse those and resolve the correct value.
* ensure that we can use a custom `--postcss ./with-custom-path.js`
* fix --help output in tests
* add tests to ensure we can use `purge.safelist`
* implement the `purge.safelist` for strings
* proxy `purge.safelist` to `purge.options.safelist`
This allows us to have a similar API in `AOT` and `JIT` mode.
* only proxy `purge.safelist` to `purge.options.safelist` if
`purge.options.safelist` doesn't exists yet.
* remove unused file
using syntax.js instead
* add a way to remove files that are generated from the tests
Essentially setting the restore cache contents to `null` as a special
value.
* combine the stdout and stderr output
* add jest-diff instead of relying on it existing *somewhere*
This also came with a small API change.
* add `--runInBand` to the test script
* add dedicated cli tests for the new Tailwind CLI
