When we add a new feature to Sass, we want to make sure the feature is well-designed, clearly specified, feasible to implement, and that it meets the use-cases it's designed for.
The process for adding a new feature works as follows:
The feature is informally discussed on the issue tracker. Most new features come directly from use-cases brought up by Sass users, or new CSS syntax that Sass needs to support. Once the Sass team has agreed that a feature is desirable, it's marked as Planned and can move to step 2.
A formal proposal is written for the feature, following the format outlined below. This proposal is sent as a pull request, where the Sass team will discuss its specifics with the author. If/when everyone agrees on a first draft, the pull request will be accepted and the feature moves to step 3.
Step 2 is also where issues are opened for each individual implementation to add the feature. These issues should link to the feature's main issue in the sass/sass issue tracker, and that issue should link back to the implementation issues.
Public comments are solicited for the feature, usually via a tweet from @SassCSS. If the feature is big enough, a blog post soliciting feedback may also be written. Then we await comments and iterate on feedback for an amount of time that varies based on the size of the feature and the amount of feedback received.
As the proposal is updated based on feedback, its draft number should be increased according to the versioning policy and changes should be logged in a changelog file named
<proposal>.changes.md. Once enough time has elapsed and the Sass team is satisfied that all feedback is addressed, the feature moves to step 4.
The proposal is marked as accepted and moved into the
accepted/directory. This doesn't mean that the proposal is immutable, but it does mean that no major changes to its semantics are expected. At this point, it's time to write specs for the new feature, in tandem with implementing it in Dart Sass (since it's the reference implementation). Writing the specs alongside an implementation helps ensure that the specs are accurate and sensible, and that the implementation is correct.
The new specs should have an
options.ymlfile that marks them as TODO for LibSass, with a reference to its issue for the new feature, and marks them as ignored for Ruby Sass. For example:
--- :todo: - libsass # sass/libsass#2701 :ignore_for: - ruby-sass
Once the specs and the implementation are complete, they're sent as pull requests to sass-spec and Dart Sass, respectively. They need to have special lines in their pull request messages in order to build properly:
The sass-spec pull request message should include
[skip dart-sass]. This will cause it not to run Dart Sass tests, which would otherwise fail because the implementation of the new feature hasn't landed yet.
The Dart Sass pull request's message should link to the sass-spec pull request (for example,
See sass/sass-spec#1293). This will cause it to run against the specs in that pull request and so test your new feature.
Once these pull requests land, the feature moves to step 5.
The feature is eventually implemented in LibSass. Once this happens, the original issue can be closed, and the feature is considered fully implemented.
A good feature proposal should make it possible for an average Sass user to understand and discuss the feature and the context around it, and possible for Sass maintainers to implement consistent and well-defined behavior. The following outline is designed to make satisfy these needs.
A proposal must include at minimum a Summary and a Syntax or a Semantics section. Everything else is optional. Proposals may include additional sections, or divide a section into sub-sections, as necessary to make it clear and readable. All proposals should also include tables of contents that link to all their sections.
Everything in sections that aren't explicitly marked as non-normative should be construed as part of the specification of the feature. Non-normative notes can be included inline in normative sections using blockquotes.
accepted/ directory for examples of proposals that have been
This non-normative section describes the broader context for the feature. This is particularly relevant for changes to existing syntax, and especially for backwards-incompatible changes. It should explain Sass's current behavior, the original reasoning behind that behavior, and why it's insufficient.
See Plain CSS
max()for a good example of a Background section.
This non-normative section provides a concise, user-friendly summary of the behavior being proposed. It doesn't need to be fully explicit about every corner of the feature, it just needs to give users an idea of how it works and what use-cases it addresses. Code examples are encouraged.
See Escapes in Identifiers for a good example of a Summary section.
This sub-section goes into detail about decisions that were made during the design of the feature. It should describe alternatives that were considered, and explain why the final decision was made the way it was.
See Plain CSS
max()for a good example of a Design Decisions section.
This section describes the syntax of the feature being added, if it adds new syntax to the language. The syntax should be written in Backus-Naur form, with regular expression-style operators and the convention that nonterminals are written in capitalized camel-case form. For example:
MinMaxExpression ::= CssMinMax | FunctionExpression CssMinMax ::= ('min(' | 'max(') CalcValue (',' CalcValue)* ')' CalcValue ::= CalcValue (('+' | '-' | '*' | '/') CalcValue)+ | '(' CalcValue ')' | ('calc(' | 'env(' | 'var(') InterpolatedDeclarationValue ')' | CssMinMax | Interpolation | Number
Syntax definitions can also refer to productions from CSS specs. The proposal should link to the specs in question.
See Range-Context Media Features for an good example of a Syntax section.
This section describes the runtime behavior of the new feature. It may be omitted if the feature only has to do with how the stylesheet is parsed. The semantics section covers everything about how a stylesheet is evaluated, including how imports are resolved and the behavior of built-in functions.
See CSS Imports for a good example of a Semantics section.
All backwards-incompatible features should go through a deprecation process if at all possible (see Dart Sass's compatibility policy). This section describes the details of that process, including what code will produce deprecation warnings and how those warnings will indicate what the user should do to make their stylesheet forwards-compatible.
See CSS Imports for a good example of a Deprecation Process section.