heroui-inc/tailwind-variants v4.0.0

on GitHub

✨ Features

• compiled merge tables; default entry exports twMerge, twJoin, and clsx
• tailwind-variants/merge: merger without tv
• tailwind-variants/config: twMergeConfig and extendTailwindMerge / createTailwindMerge / createTwMerge / fromTheme / validators / mergeConfigs / getDefaultConfig
• multi-extend via extend: [a, b, c] (#297)
• empty class lists return "" (#289)
• array compound / variant classes apply to the base slot (#263)
• experimental debug mode
• inject twMerge from tailwind-variants/lite (#307)
• cacheSize and prefix on the default tables

⚡ Performance

• hot tv() about 3.5× vs 3.3.1, over 6× with compound variants
• cn() about 2× vs 3.3.1 on join-and-merge, over 7× on interpolated arguments, about 2× on unique arbitrary values
• pre-merged static core; class / className cached by text
• compound variants kept out of cache keys (#317)

🐛 Fixes

• unlabeled font-[Inter] is family, font-[700] is weight
• contain-*, bg-gradient-to-*, bg-conic, columns-auto, max-h-none, extra size scales
• debug reporter stripped when NODE_ENV is production
• compoundSlots no longer widens variant inference (#316)

🚨 Breaking Changes

• twMergeConfig: tailwind-variants/config only; default entry throws
• experimentalParseClassName: removed
• cn: cn("a", 0) → "a"; cn(0) → "0"
• empty lists: "" instead of undefined
• types: unknown variant / slot keys error
• peers: tailwind-merge and tailwindcss removed
• lite: function twMerge is called (3.3.1 treated it as true)
• twMerge: false: cx join; whitespace normalized at compile

🛠 Chores

• pnpm compile-tables
• isolated benchmark suites
• pkg.pr.new
• toolchain on Vitest 5, tsup 9, Biome 2.6, pnpm 12; audit clean (#315)

See the v3 to v4 migration guide.

Thanks @danielroe for #317 and @wingkwong for #315.

View changes on GitHub