cache TailwindMerge instance

[thedumbtechguy]

TailwindMerge operates a cache and undertakes a lot of operations that require that an instance be reused rather than constantly instantiating one.

I added a simple benchmark

################ Simple ################* Single run: user system total realCached: 0.000234 0.000006 0.000240 ( 0.000237)Uncached: 0.004120 0.000448 0.004568 ( 0.004769)== Cached is 20.12x faster than Uncached.* 1000 runs: user system total realCached: 0.001120 0.000013 0.001133 ( 0.001332)Uncached: 2.199289 0.036016 2.235305 ( 2.511064)== Cached is 1885.18x faster than Uncached.* Iterations per second (5 second run):Cached: 877208Uncached: 452== Cached performs 1940.73x more iterations per second than Uncached.-------------------------------------------------------------------------------------

Benchmark Code

require"benchmark"require"tailwind_merge"classCachedMergerMERGER=TailwindMerge::Merger.new.freezedefcall(classes)MERGER.merge(classes)endendclassUncachedMergerdefcall(classes)TailwindMerge::Merger.new.merge(classes)endenddefrun_benchmark(name,classes)puts"\n################ #{name} ################\n"puts"\n* Single run:"single_results=Benchmark.bm(20)do |x|
x.report("Cached:"){CachedMerger.new.call(classes)}x.report("Uncached:"){UncachedMerger.new.call(classes)}endcompare_results(single_results,"Single run")puts"\n* 1000 runs:"bulk_results=Benchmark.bm(20)do |x|
x.report("Cached:"){1000.times{CachedMerger.new.call(classes)}}x.report("Uncached:"){1000.times{UncachedMerger.new.call(classes)}}endcompare_results(bulk_results,"1000 runs")# Simple IPS-like benchmarkduration=5# Run for 5 secondsputs"\n* Iterations per second (5 second run):"ips_results=[["Cached",->{CachedMerger.new.call(classes)}],["Uncached",->{UncachedMerger.new.call(classes)}]].mapdo |name,proc|
iterations=0start_time=Time.nowwhileTime.now - start_time < durationproc.calliterations += 1endips=iterations / durationputs"#{name}: #{ips.round(2)}"[name,ips]endcompare_ips(ips_results)puts"-------------------------------------------------------------------------------------"enddefcompare_results(results,label)our_time=results[0].realphlex_time=results[1].realfaster,slower=(our_time < phlex_time) ? ["Cached","Uncached"] : ["Uncached","Cached"]multiplier=[our_time,phlex_time].max / [our_time,phlex_time].minputs"== #{faster} is #{multiplier.round(2)}x faster than #{slower}."enddefcompare_ips(results)our_ips,phlex_ips=results.map{ |_,ips| ips.to_f}faster,slower=(our_ips > phlex_ips) ? ["Cached","Uncached"] : ["Uncached","Cached"]multiplier=[our_ips,phlex_ips].max / [our_ips,phlex_ips].minputs"== #{faster} performs #{multiplier.round(2)}x more iterations per second than #{slower}."end# Run benchmarksrun_benchmark("Simple","inline-block h-12 w-12 rounded-full ring-2 ring-white block font-bold px-3 py-2 text-slate-700 rounded-lg hover:bg-slate-100 hover:text-slate-900")

[cirdes]

@thedumbtechguy, wow! That is incredible.

Right now, we've decided to freeze the main branch and work on v1 until we complete the migration. We explained it here.

Would you mind sending the PR to the v1 branch or resolving the conflicts?

[thedumbtechguy]

Sure thing. I'll resolve the issues and resubmit.