Simple benchmarks in both node and browser
benchee
requires that Promise
is available globally. If using an environment that does not support it, you should polyfill prior to importing benchee
.
import { benchmark, createSuite } from "benchee";
// the functions to benchmark
const add = (a, b) => a + b;
const subtract = (a, b) => a - b;
// create an individual benchmark
benchmark("add", () => add(1, 2)).then(results => console.log(results));
/*
{
"stats": {
"elapsed": 677,
"endTime": 1540050973491,
"startTime": 1540050972814,
"iterations": 82165907,
"ops": 121367661,
"tpe": 0.00000823942708987561
},
"name": "add"
}
*/
// or create a suite of benchmarks
createSuite()
.add("add", () => add(1, 2))
.add("subtract", () => subtract(1, 2))
.run()
.then(results => console.log(results));
/*
{
"ungrouped": [
{
"stats": {
"elapsed": 802,
"endTime": 1540050973617,
"startTime": 1540050972815,
"iterations": 28777534,
"ops": 35882211,
"tpe": 0.000027868961947886152
},
"name": "add"
},
{
"stats": {
"elapsed": 750,
"endTime": 1540050974473,
"startTime": 1540050973723,
"iterations": 106326932,
"ops": 141769242,
"tpe": 0.000007053716174186235
},
"name": "subtract"
}
]
}
*/
The results contract is Promise
-based, however you can also access the results in a callback format if preferred.
In addition to running standard benchmarks, you can group benchmarks together within the same suite. The results of each group can be accessed through the onGroupComplete
callback, and will namespaced under the group name in the final results.
To apply a group, simply add a group name as the second parameter to your test.
createSuite()
.add("add", "math", () => add(1, 2))
.add("trim", "string", () => " trimmed ".trim())
.run()
.then(results => console.log(results));
/*
{
"math": [
{
"stats": {
"elapsed": 888,
"endTime": 1540051045206,
"startTime": 1540051044318,
"iterations": 38415463,
"ops": 43260656,
"tpe": 0.000023115691720284616
},
"name": "add"
}
],
"string": [
{
"stats": {
"elapsed": 568,
"endTime": 1540051045880,
"startTime": 1540051045312,
"iterations": 15363261,
"ops": 27047994,
"tpe": 0.00003697131748266205
},
"name": "trim"
}
]
}
*/
Statistics for each benchmark have the following shape:
{
// time of total benchmark run
elapsed: number;
// timestamp of benchmark complete
endTime: number;
// number of operations executed in benchmark
iterations: number;
// operations per second calculation
ops: number;
// time per execution calculation
tpe: number;
// timestamp of benchmark start
startTime: number;
}
The time wait between execution of benchmark groups (defaults to 100ms)
The minimum number of iterations that need to occur before the benchmark is considered complete (defaults to 10)
The minimum amount of time that needs to elapse before the benchmark is considered complete (defaults to 500ms)
NOTE: This is ignored when type
is set to fixed
.
Function called when suite has finished running. This is the callback method to receive results, if preferred over the standard promised-based method.
onComplete: (results: Benchee.Results) => void;
Function called when a given group has completed all of its benchmarks.
onGroupComplete: (results: Benchee.ResultsGroup) => void;
Function called when a given group has started running its benchmarks.
onGroupStart: (group: string) => void;
Function called when a specific benchmark has finished running.
onResult: (result: Benchee.Result) => void;
The type of benchmark to perform. (defaults to adaptive
)
Valid values:
adaptive
=> number of iterations performed is based on an exponential algorithm driven by theminTime
fixed
=> number of iterations performed is based directly onminIterations
- Chrome (33+)
- Edge (all)
- Firefox (29+)
- Opera (20+)
- Safari (7.1+)
NOTE: If a Promise
polyfill is provided, then older versions / unlisted browsers should be supported as well (notably IE11).
- 6+
Standard stuff, clone the repo and npm install
dependencies. The npm scripts available:
build
=> run rollup to build the distributed files indist
clean
=> runrimraf
on thedist
folderdev
=> run webpack dev server to run example app (playground!)dist
=> runsclean
,build
, andbuild:types
lint
=> runs TSLint against all files in thesrc
folderlint:fix
=> runslint
, fixing any errors if possibleprepublish
=> runsprepublish:compile
prepublish:compile
=> runlint
,test:coverage
, anddist
test
=> run AVA test functions withNODE_ENV=test
test:coverage
=> runtest
but withnyc
for coverage checkertest:watch
=> runtest
, but with persistent watcher