UNPKG

postcss-merge-longhand

Version:

Merge longhand properties into shorthand with PostCSS.

24 lines • 929 B
export declare const closingTokens: Set<import("@csstools/css-tokenizer").TokenType>; export type Component = { raw: string; tokens: import('@csstools/css-tokenizer').CSSToken[]; }; /** * @typedef {{raw: string, tokens: import('@csstools/css-tokenizer').CSSToken[]}} * Component */ /** * Split raw CSS into top-level components and comma-separated parts. This is * deliberately lexical: property grammar is applied by the caller after * brackets and source ranges have been preserved. * * @param {string} value * @param {',' | '/'} [separator] - the top-level token that divides the value * into parts; a comma is otherwise a failure, and a slash an ordinary token * @return {{components: Component[], raw: string}[] | null} */ export declare function splitValue(value: string, separator?: ',' | '/'): { components: Component[]; raw: string; }[] | null; //# sourceMappingURL=valueComponents.d.ts.map