[![gh-board][kanban-image]][kanban-url]
[![NPM version][npm-image]][npm-url]
[![Downloads][downloads-image]][downloads-url]
[![build status][travis-image]][travis-url]
[![dependency status][dependency-image]][dependency-url]
[![dev dependency status][dev-dependency-image]][dev-dependency-url]
[![code coverage][coverage-image]][coverage-url]
[![greenkeeper][greenkeeper-image]][greenkeeper-url]

Why use CSS to transform the DOM?

Book authors need a way to style their book for a PDF and ePUB. Sometimes they need additional elements to properly style. They also need a way to number and move elements around, like solutions in the back of the book.

Rather than letting them write JavaScript and execute arbitrary code, authors can describe both the styling and the movement of elements in the same file.

So, What's the gist of this thing?

When converting HTML there are 3 basic things one needs to do:

1. lookup attributes or elements (to move)
1. change (or delete) an element (attributes, tag name)
1. add new elements

Each of these are accomplished using different parts of CSS:

1. declaration functions are used to look up information (ie attr(id))
1. declarations are used to modify an element (ie class-add: "foo", "bar"; or tag-name-set: "h3";)
1. pseudoelements are used to create new elements (ie ::before { content: "Example"; })

Motivation/Examples

The Motivation Page has a step-by-step introduction to the CSS features or just play around with the JSFiddle of all the steps combined

The Language Reference Page contains a list of all selectors, rules, functions that are understood.

Table of Contents

- Features
- Precise Error Messages
- HTML SourceMaps
- Exactly 1 Pass
- Parallelizing
- CSS and HTML Coverage
- Debugging CSS Files
- Approximate Sourcemaps
- Screencasts
- Install Notes
- Debugging css-plus
- TODO

Features

This library aims to address several stories that could be helpful.

Note: ap-physics takes a couple minutes to load and then save

Precise Error Messages

This addresses a few use-cases:

1. As a CSS Developer I want to know if I mis-typed a rule/function/selector
1. As a CSS Developer I want to know which CSS and HTML combination caused a crash to occur
1. As a CSS Developer I want to know when something is unexpectedly empty so I do not have to look for the absence in the baked HTML file

Here are some example error messages from various tests. All but one are already implemented:

``js
test/_errors.css:1:20: WARNING: Skipping unrecognized rule 'contentssss:', maybe a typo? (test/_errors.in.html:8:1)
test/specificity.css:30:28: WARNING: Skipping this rule because it was overridden by test/specificity.css:42:13 (test/specificity.in.html:8:1)
test/move.css:14:5: WARNING: Moving 0 items using selector '.exercise'. You can add a :has() guard to prevent this warning (test/move.in.html:10:1)
test/move.css:14:5: ERROR: Tried to move an HTML element that was already marked for moving. The first move rule was matched by test/move.css:6:3. (test/move.in.html:13:6)
(stack trace here)


test/move.css:7:2: ERROR: Exception occurred while manipulating the DOM (test/move.in.html:13:6)
(stack trace here)
`

These are formatted in a way that can be parsed by a linter and therefore can show up in your text editor :smile: and they include both the CSS and HTML source location information.

When coupled with source maps (see below), this vastly reduces debugging time because you know exactly where in the CSS file to look and which HTML element was being processed at the time.

HTML SourceMaps

This addresses a few use-cases:

1. As a CSS Developer I want to see why a specific element, class, or piece of generated text is showing up
1. As a Content Editor I want to review the baked HTML file but when I find a content problem I want to find the source so I can edit it
1. As a GUI Developer I want people to see what the book looks like (collated & numbered) but when I click to edit I want to edit the original content
1. As a CSS Developer I want to see what all is being affected by a CSS selector/rule so I know if it is affecting the right elements

$3

In this animation the left pane contains the baked HTML and the right pane contains the source that generated the code (either HTML content or CSS selectors/rules).

As you click in the left pane, the right pane updates.

!html-sourcemaps-wtf

_(there is no trickery here, this is real working code using the plugin atom-sourcemap)_

Exactly 1 Pass

This addresses a few use-cases:

- As a CSS Developer I want to not have to reason about intermediate state of the DOM
- As a CSS Developer I want the conversion to be fast

This model is inspired by both CSS and virtualdom libraries like React: you describe what you want the result to look like rather than the steps to get to the result.

No more writing the intermediate steps needed to get to the desired result (ie setting temporary attributes, multiple passes to move things around).

$3

Since each element does not depend on the state of other elements (and there is only 1 pass), the conversion can be parallelized, further reducing the conversion time.

CSS and HTML Coverage

css-plus generates an ${OUTPUT_HTML}.lcov file in addition to the ${OUTPUT_HTML}.map file which contains all the CSS covered and the HTML elements that were matched during the conversion.

See the codecov page for examples.

Debugging CSS Files

When building an interpreter for a language there are a few useful features that make development much easier:

1. Logging within the language
- Add x-log: "My message"; in CSS
1. When the interpreter crashes, show which line of code was being executed (and with which data)
- Something like foo.scss:13:2 [error message] (foo.html#id123) followed by a stack trace
1. A way to pause in the code and see what data is being operated on
- You can add a data-debugger="true" attribute to the HTML (or debugger: true; in the CSS)


Adding the data-debugger="true" attribute to an element is like setting a breakpoint in other programming languages.

It will output the following and if css-plus is run in debugging mode, pause the debugger.

Example Output below contains the following:

- the HTML element and its location in the source code
- which selectors were matched and their location in the source code
- which declarations were applied and their location in the source code

`
/----------------------------------------------------
| Debugging data for <<./data/statistics-raw.html:div#098e1a26-e612-4449-a45e-80fa23feba02@12>>
| Matched Selectors:
| ../rulesets/books/statistics/book.scss:32:8 body > [data-type="chapter"]:has(section.summary) {...}
| Applied Declarations:
| ../rulesets/books/statistics/book.scss:18:2 content: 1 "" "." "" 2 "" " ";
| ../rulesets/books/statistics/book.scss:35:10 attrs-add: "href" "" "#" "" "098e1a26-e612-4449-a45e-80fa23feba02@12";
| ../rulesets/books/statistics/book.scss:34:10 tag-name-set: "a";
\----------------------------------------------------
`

Approximate SourceMaps

This is just a script that takes 2 HTML files (raw and baked HTML) and builds a sourcemap file by looking at the id attributes of elements. It is not precise but may be "good enough" to use in the short term for things like:

- Content Editors fixing typos or broken links
- WYSIWYG Editor that renders the baked HTML file but edits the raw when you click on something

It is also useful for XSLT passes (like CNXML -> HTML) that have no way to otherwise generate a sourcemap

!approximate-sourcemaps

_This shows a full baked textbook in Atom in the left pane and the raw book file in the right pane. It does not open up the CSS file in the right pane because the sourcemap was not generated using css-plus_

To run it yourself:

`
node ./sourcemap-approximator.js ${SOURCE_FILE} ${GENERATED_FILE}
`

$3

There are 3 phases (annotate, build a work tree, manipulate):

1. The entire HTML DOM is annotated with rulesets that match each element
1. The DOM is traversed (in-order but that does not matter).
- Selectors/declarations that require looking at DOM nodes are evaluated at this point and any manipulations that need to happen are added to a work-tree (as Closures/Promises)
- [ ] elements that will move are marked to see if 2 rules are moving the same element (aka Mark-Sweep in Garbage-Collection terms)
1. The DOM is manipulated by evaluating the closures in the work-tree

Screencasts

Debugging Example:

![](https://asciinema.org/a/1xtpezj5j9mmce8pxfpq7djs9)

Example of converting an entire book (takes a few minutes but you get progress bars and status updates):

![](https://asciinema.org/a/9ml3wzqzdide7r4bagigujdip)

Running from the commandline

You can install this globally or as a package in an existing project.

To install and run globally, run npm install --global css-plus

Then you can run:

`sh
css-plus --css ${INPUT_CSS} --html ${INPUT_HTML} --output ${OUTPUT_HTML}

Or if you are lazy:

To install locally and run, type npm install --save css-plus

Then you can run the previous examples but replace css-plus with $(npm bin)/css-plus

$3

This takes 2 XML/HTML files (ie raw & baked) and builds a rough sourcemap using the id attribute of elements.

`
node ./sourcemap-approximator.js ${SOURCE_FILE} ${GENERATED_FILE}
`

and it generates a file at ${GENERATED_FILE}.map

Debugging css-plus

run ./script/test-debug to start up a debugger and run the tests.

To test the commandline:

- $(npm bin)/inspect ./bin/css-plus ${CSS_PATH} ${HTML_PATH} ${OUTPUT_HTML_PATH}
- $(npm bin)/inspect ./bin/css-plus-debug ${CSS_PATH} ${HTML_PATH} ${OUTPUT_HTML_PATH} starts up the debugger (using node --inspect-brk)

To pause when evaluating an element, add data-debugger="true" to the element.

TODO

- [x] Support tag-name-set:
- [x] Support selector specificity
- [x] Support !important
- [x] Add ::for-each(1, descendant, ${SELECTOR})
- [x] Add :target(${ATTRIBUTE_NAME}, ${MATCH_SELECTORS...})
- [x] Use Promises to defer DOM Manipulation to all be after selector/rule/attribute evaluation is done
- [x] Add a command-line script to run the conversion
- [ ] Add Examples
- [x] Add example showing complex numbering (only counting some exercises)
- [ ] may need to introduce :is(${SELECTOR}) to add exceptions
- [ ] Add example showing how to only transform certain chapters (or any selector)
- May require adding support for multiple CSS files and an env(NAME, DEFAULT) function
- [x] Add example showing how to build a glossary
- [x] add additional args to move-here() for sorting a glossary or answers to exercises
- [ ] add contents() which does a deep clone but removes id's
- [ ] support tag-name-set: none; which unwraps the element (useful for