* Start of cli docs
* Document the CLI
Also adds support for the `--config` option and `--port` option for the dev server.
* Add tests for --config and --port flags
* Add port to validateConfig
Svelte component resolution wasn't handled correctly during the build.
Note that in the future we need to consolidate a "framework" API, so this stuff is not sprinkled throughout the codebase.
Another change in snowpack@3 caused this bug. It's not actually a bug in snowpack. Previously snowpack was keeping its list of installed packages in a global cache. In 3.3 it stopped doing so. We were accidentally relying on that global cache to be able to resolve dynamic components.
This fixes it so that we use the frontend snowpack instance to resolve dynamic components. Doing so means they are available when we try to load them.
* Pass the `mode` through to snowpack
This allows the production packages to be prepared.
* Use snowpack 3.3.1
* Update path to prism loadComponents external ref
* Upgrade to snowpack 3.3.2
* add dep domhandler imported in in src/build/static
* lint and jsDoc error
* move domhandler to devDep
* chore: add package lock
* escape string jsDoc
* chore: add astro dep in until prism import is refactored
* chore: add snowpack example package lock
* Parse inner JSX as Astro
This completes the compiler changes, updating the parser so that it parses inner "JSX" as Astro. It does this by finding the start and end of HTML tags and feeds that back into the parser.
The result is a structure like this:
```
{
type: 'MustacheTag',
expression: [
{
type: 'Expression',
codeStart: 'colors.map(color => (',
codeEnd: '}}'
children: [ {
type: 'Fragment',
children: [ {
type: 'Element',
name: 'div'
} ]
} ]
}
]
}
```
There is a new Node type, `Expression`. Note that `MustacheTag` remains in the tree, all it contains is an Expression though. I could spend some time trying to remove it, there's just a few places that expect it to exist.
* Update import to the transform
* Transform prism components into expressions
* Compiler cleanup
This is general compiler cleanup, especially around the codegen part. Goals here were too:
1. Make it possible to compile HTML recursively (needed for future astro-in-expressions work) by moving that work into its own function.
1. Get rid of collectionItems and have compiling the HTML return just a source string.
Also not planned, this change gets rid of the different between components and pages. All Astro components compile to the same JavaScript.
* Remove unused node types
This improves the algorithm for searching for pages. It now works like:
1. If pathname ends with /
1. Look for PATHNAME/index.astro
1. Look for PATHNAME/index.md
1. else
1. Look for PATHNAME.astro
1. Look for PATHNAME.md
1. Look for PATHNAME/index.astro
1. 301
1. Look for PATHNAME/index.md
1. 301
1. 404
* Fix complex MDX parsing
This allows fully MDX support using the micromark MDX extension. One caveat is that if you do something like use the less than sign, you need to escape it because the parser expects these to be tags otherwise.
* Move micromark definition
* Implement fallback capability
This makes it possible for a dynamic component to render fallback content on the server.
The mechanism is a special `static` prop passed to the component. If `static` is true then the component knows it can render static content.
Putting aside the word `static`, is this the right approach? I think giving components the flexibility to make the decision themselves *is* the right approach.
However in this case we have a special property that is passed in non-explicitly. I think we have to do it this way because if the caller passes in a prop it will get serialized and appear on the client. By making this something we *add* during rendering, it only happens on the server (and only when using `:load`).
Assuming this is the right approach, is `static` the right name for this prop? Other candidates:
* `server`
That's all I have!
* Use `import.meta.env.astro` to tell if running in SSR mode.
* Run formatter
* Convert CSS Modules to scoped styles
* Update README
* Move class scoping into HTML walker
* Fix SSR styles test
* Fix mustache tags
* Update PostCSS plugin name
* Add JSDoc comment
* Update test
* Allow HMX components in markdown
This adds support for HMX components in markdown. The mechanism for importing is via frontmatter. We could do this differently (setup script maybe?) but since this was the easiest to implement I thought it was a good first-pass option.
* Remove node-fetch from snowpack config
* Assert that the runtime is created successfully
* Add back in the micromark extension for encoding entities
* Encode both codeTextData and codeFlowValue
* Install snowpack app's deps
* Add support for React components.
This adds support for react components via a new `extensions` config in astro.config.mjs. In the future we can extend this to do things like look at the import statements, as Snowpack does.
* Fix the tests
* Initial tests set up
This adds tests using uvu (we can switch if people want) and restructures things a bit so that it's easier to test.
Like in snowpack you set up a little project. In our tests you can say:
```js
const result = await runtime.load('/blog/hello-world')
```
And analyze the result. I included a `test-helpers.js` which has a function that will turn HTML into a cheerio instance, for inspecting the result HTML.
* Bring snowpack example in
* Formatting
* Begin debugging
* Initial tests set up
This adds tests using uvu (we can switch if people want) and restructures things a bit so that it's easier to test.
Like in snowpack you set up a little project. In our tests you can say:
```js
const result = await runtime.load('/blog/hello-world')
```
And analyze the result. I included a `test-helpers.js` which has a function that will turn HTML into a cheerio instance, for inspecting the result HTML.
* Add CI
* Remove extra console logs
* Formatting
