There are a few benefits of custom implementation
- csstree instead of css-what parser
- 9 less dependecies
- only standard features are supported (css-select has :parent from jquery)
- new implementation allows to get rid of parentNode
I saw complaints about `extendDefaultPlugins` api
- it cannot be used when svgo is installed globally
- it requires svgo to be installed when using svgo-loader or svgo-jsx
- it prevents using serializable config formats like json
In this diff I introduced the new plugin which is a bundle of all
default plugins.
```js
module.exports = {
plugins: [
'preset_default',
// or
{
name: 'preset_default',
floatPrecision: 4,
overrides: {
convertPathData: {
applyTransforms: false
}
}
}
]
}
```
Ref #1476
The issue is that a path command may have multiple arcs, and the space optimization was only applied to the first one. Modified a test to check it.
Co-authored-by: Josep del Río <joseprio@fb.com>
Co-authored-by: Bogdan Chadkin <trysound@yandex.ru>
Mocha doesn't have a lot of features provided by jest.
There is a great assertion library out of the box.
And the most cool feature is inline snapshots.
Mocha also hides errors which makes debugging a nightmare sometimes.
Wrapping each node with JSAPI class and passing parent to it is not
reliable. Parent may be changed but the reference will stay.
Here I wasn't able to detach comment from parent node for some reason.
Explicit parent node inferred while ast traverse is easier to debug and
work with. Eventually we will not need to wrap each node with JSAPI class.
computeStyle(node) in isolation is quite slow utility because it
collects style elements across whole document, parses and sort them.
In this diff I splitted it into `collectStylesheet(root)` and
`computeStyle(stylesheet, node)` which are easy integrate with new
visitor plugin api.
Remove 5 deps from production install:
`ansi-styles`, `supports-color`, `color-convert`, `has-flag`, `color-name`.
Also reduce deps when using PostCSS and SVGO together, since PostCSS
uses colorette.
Tested with `yarn install svgo`.
Visitor is a simple pattern which helps to avoid many type checks
and provide both "perItem" and "perItemReverse" functionality without
fragmentation.
The most important case is an ability to define state which in many
plugins specified either on module level or by polluting `params`.
In this diff I added visit and detachFromParent utilities and refactored
new mergeStyles plugin with it.
Also fixed bug when cdata content is merged into "text" node which is
not always valid.
https://github.com/svg/svgo/pull/1279 faced some problems with node cloning.
In this diff I moved class/style live objects to JSAPI and made it non
enumerable to avoid deep cloning issues.
class list and style declaration classes are not longer need own clone
methods.
These iterators allows to directly manipulate passed value
which does not let us to get rid from legacy "attrs" field.
Object.entries makes it easier to get an access to both attribute
name and value.
There is a lot of attributes manipulation which is hard to remove at
once. In this diff I added `attributes` object and wrapped it as proxy
for `attrs` field.
Ref https://github.com/syntax-tree/xast
- added type: root | element
- renamed elem to name
- replaced "elem" property checks with check for correct type
Here we add [xast](https://github.com/syntax-tree/xast) support
to three basic nodes: doctype, instruction and comment
Some tests are rewritten instead of checking each field to `.include`
assertion which is able to match shape of object.
Changed the TS config file to include all relevant folders and explicitly exclude those files that fail the type checks. In this way new files will be automatically type checked and we can see the list of files that need types improvements.
In this pass I also fixed some js files.
Added Typescript support via JSDoc comments for some files.
The code has really outdated type signatures, so in order to fully type the codebase the changes need to be incremental.
To check the types run:
npx tsc
Ref https://github.com/svg/svgo/issues/777
Currently a lot of optimisations are attributes specific and may be
broken because of inline or shared styles.
In this diff I'm trying to solve the problem with getComputedStyle
analog.
`computeStyle` collects attributes, shared css rules, inline styles
and inherited styles and checks whether they can be statically optimised
or left as deoptimisation.
