Benchmarks

Every push to main measures what ResXLocalization costs an application, on the same GitHub-hosted Windows runner, .NET 10 and the just-in-time compiler:

• Latest results: the tables of the last commit on main, with the hardware and runtime they were measured on.
• History: the time and allocations of every benchmark over the last 200 commits on main.

Every benchmark is compared with the code an application writes without the library: the .resx designer class and hand-written code over it.

Benchmark	What it measures	Baseline
CodeLookupBenchmarks	Localizer.Current.Get with a typed key, a scoped key, a search-all key and an enum value	The designer class's property, BenchmarkStrings.Greeting
SearchAllBenchmarks	A search-all lookup whose key is in the last of 1, 5 or 20 registered resource files	A hand-written loop that asks the same resource managers, in the same order, until one has the key
ArgumentBenchmarks	Get with MessageArguments.FromValues, FromTuples and FromObject	string.Format over the designer class's property
MessageFormatBenchmarks	Get with two named arguments and UseMessageFormat() installed	The MessageFormat library's FormatMessage over the designer class's property
AvaloniaFormsBenchmarks, WpfFormsBenchmarks	Opening 100 elements that all use {l:Localize}, {l:Localize} with two <l:Argument> children, or {l:LocalizeEnum}	The same views with {x:Static} of the designer class's properties
AvaloniaScalingBenchmarks, WpfScalingBenchmarks, Open	Opening a window with 100 or 10,000 {l:Localize} elements	The same views with {x:Static}
AvaloniaScalingBenchmarks, WpfScalingBenchmarks, Switch	Switching the culture while 100 or 10,000 {l:Localize} elements are live	The same views bound to a hand-written TranslationSource with a string indexer

The Ratio column is a benchmark's time divided by its baseline's, so 1.20 means 20% slower than the code without the library; Allocated is the memory one operation allocates, and Alloc Ratio compares it with the baseline the same way.