Skip to content

Why class-recipe

class-recipe solves the same problem as class-variance-authority and tailwind-variants: turning component props into class names. Its config looks like theirs, so moving over is mostly mechanical. It differs in four ways:

  • It styles components of several elements with slots, which class-variance-authority does not have.
  • A variant without a default is required by its type, so a component cannot forget to choose it.
  • It resolves conflicting classes with any join function, such as twMerge, and caches the result, so the join runs once per selection.
  • It is faster: it prepares each config once and caches the class name of each selection, so most calls are a lookup.
class-recipe 1.7.2 class-variance-authority 0.7.1 tailwind-variants 3.3.1
Slots Yes No Yes
Variant without a default Required by its type Optional Optional
Conflict resolution Any join function, cached Call twMerge on the result tailwind-merge, built in
Extending a recipe composes No extend
Recipe calls per second, without merging 13.4 million 1.4 million 1.0 million

tailwind-variants has two builds. Its main entry point merges classes with tailwind-merge. Its lite build, tailwind-variants/lite, does not merge classes, and does not need tailwind-merge. The calls per second above measure the lite build, so that every library does the same work. They come from the benchmarks, which also measure each library with tailwind-merge.

One recipe call is fast in every one of these libraries. For a page with a few buttons, you will not notice the difference. The cost adds up when a recipe runs many times per second:

  • Large lists. A table, a feed, or a grid with hundreds of rows calls several recipes for each row, on every render.
  • Frequent re-renders. A component that renders on each keystroke, scroll event, or animation frame calls its recipes each time.
  • Merging classes on every call. A cn(button(props)) helper, which passes each result to twMerge, runs tailwind-merge on every render. tailwind-merge does much more work than joining strings. A recipe of class-recipe created with createRecipes({ join: twMerge }) runs it once for each selection, then reads its cache. In the benchmarks, it runs 4× as many calls per second as class-variance-authority with twMerge on every result.

A faster recipe leaves more of each frame to the rest of your app. It cannot make a slow component fast, though: measure your app before you optimize it.