On this page
deno bench
, benchmarking tool
Command line usage
deno bench [OPTIONS] [files]... [-- [SCRIPT_ARG]...]
Run benchmarks using Deno's built-in bench tool.
Evaluate the given files, run all benches declared with 'Deno.bench()' and report results to standard output:
deno bench src/fetch_bench.ts src/signal_bench.ts
If you specify a directory instead of a file, the path is expanded to all contained files matching the glob {*_,*.,}bench.{js,mjs,ts,mts,jsx,tsx}
:
deno bench src/
Type checking options Jump to heading
--check
Jump to heading
Set type-checking behavior. This subcommand type-checks local modules by default, so adding --check
is redundant
If the value of "all" is supplied, remote modules will be included.
Alternatively, the 'deno check' subcommand can be used.
--no-check
Jump to heading
Skip type-checking. If the value of "remote" is supplied, diagnostic errors from remote modules will be ignored.
Dependency management options Jump to heading
--cached-only
Jump to heading
Require that remote dependencies are already cached.
--frozen
Jump to heading
Error out if lockfile is out of date.
--import-map
Jump to heading
Load import map file from local file or remote URL.
--lock
Jump to heading
Check the specified lock file. (If value is not provided, defaults to "./deno.lock").
--no-lock
Jump to heading
Disable auto discovery of the lock file.
--no-npm
Jump to heading
Do not resolve npm modules.
--no-remote
Jump to heading
Do not resolve remote modules.
--node-modules-dir
Jump to heading
Sets the node modules management mode for npm packages.
--reload
Jump to heading
Short flag: -r
Reload source code cache (recompile TypeScript) no value Reload everything jsr:@std/http/file-server,jsr:@std/assert/assert-equals Reloads specific modules npm: Reload all npm modules npm:chalk Reload specific npm module.
--vendor
Jump to heading
Toggles local vendor folder usage for remote modules and a node_modules folder for npm packages.
Options Jump to heading
--allow-scripts
Jump to heading
Allow running npm lifecycle scripts for the given packages
Note: Scripts will only be executed when using a node_modules directory (--node-modules-dir
).
--cert
Jump to heading
Load certificate authority from PEM encoded file.
--config
Jump to heading
Short flag: -c
Configure different aspects of deno including TypeScript, linting, and code formatting
Typically the configuration file will be called deno.json
or deno.jsonc
and
automatically detected; in that case this flag is not necessary.
--env-file
Jump to heading
Load environment variables from local file Only the first environment variable with a given key is used. Existing process environment variables are not overwritten.
--ext
Jump to heading
Set content type of the supplied file.
--filter
Jump to heading
Run benchmarks with this string or regexp pattern in the bench name.
--ignore
Jump to heading
Ignore files.
--json
Jump to heading
UNSTABLE: Output benchmark result in JSON format.
--location
Jump to heading
Value of globalThis.location used by some web APIs.
--no-config
Jump to heading
Disable automatic loading of the configuration file.
--no-run
Jump to heading
Cache bench modules, but don't run benchmarks.
--seed
Jump to heading
Set the random number generator seed.
--v8-flags
Jump to heading
To see a list of all available flags use --v8-flags=--help
Flags can also be set via the DENO_V8_FLAGS environment variable.
Any flags set with this flag are appended after the DENO_V8_FLAGS environment variable.
File watching options Jump to heading
--no-clear-screen
Jump to heading
Do not clear terminal screen when under watch mode.
--watch
Jump to heading
Watch for file changes and restart process automatically. Only local files from entry point module graph are watched.
--watch-exclude
Jump to heading
Exclude provided files/patterns from watch mode.
Quickstart Jump to heading
Firstly, let's create a file url_bench.ts
and register a bench using the
Deno.bench()
function.
// url_bench.ts
Deno.bench("URL parsing", () => {
new URL("https://deno.land");
});
Secondly, run the benchmark using the deno bench
subcommand.
deno bench url_bench.ts
cpu: Apple M1 Max
runtime: deno 1.21.0 (aarch64-apple-darwin)
file:///dev/deno/url_bench.ts
benchmark time (avg) (min … max) p75 p99 p995
--------------------------------------------------- -----------------------------
URL parsing 17.29 µs/iter (16.67 µs … 153.62 µs) 17.25 µs 18.92 µs 22.25 µs
Writing benchmarks Jump to heading
To define a benchmark you need to register it with a call to the Deno.bench
API. There are multiple overloads of this API to allow for the greatest
flexibility and easy switching between the forms (eg. when you need to quickly
focus a single bench for debugging, using the only: true
option):
// Compact form: name and function
Deno.bench("hello world #1", () => {
new URL("https://deno.land");
});
// Compact form: named function.
Deno.bench(function helloWorld3() {
new URL("https://deno.land");
});
// Longer form: bench definition.
Deno.bench({
name: "hello world #2",
fn: () => {
new URL("https://deno.land");
},
});
// Similar to compact form, with additional configuration as a second argument.
Deno.bench("hello world #4", { permissions: { read: true } }, () => {
new URL("https://deno.land");
});
// Similar to longer form, with bench function as a second argument.
Deno.bench(
{ name: "hello world #5", permissions: { read: true } },
() => {
new URL("https://deno.land");
},
);
// Similar to longer form, with a named bench function as a second argument.
Deno.bench({ permissions: { read: true } }, function helloWorld6() {
new URL("https://deno.land");
});
Async functions Jump to heading
You can also bench asynchronous code by passing a bench function that returns a
promise. For this you can use the async
keyword when defining a function:
Deno.bench("async hello world", async () => {
await 1;
});
Critical sections Jump to heading
Sometimes the benchmark case needs to include setup and teardown code that would taint the benchmark results. For example, if you want to measure how long it takes to read a small file, you need to open the file, read it, and then close it. If the file is small enough the time it takes to open and close the file might outweigh the time it takes to read the file itself.
To help with such situations you can Deno.BenchContext.start
and
Deno.BenchContext.end
to tell the benchmarking tool about the critical section
you want to measure. Everything outside of the section between these two calls
will be excluded from the measurement.
Deno.bench("foo", async (b) => {
// Open a file that we will act upon.
const file = await Deno.open("a_big_data_file.txt");
// Tell the benchmarking tool that this is the only section you want
// to measure.
b.start();
// Now let's measure how long it takes to read all of the data from the file.
await new Response(file.readable).arrayBuffer();
// End measurement here.
b.end();
// Now we can perform some potentially time-consuming teardown that will not
// taint out benchmark results.
file.close();
});
Grouping and baselines Jump to heading
When registering a bench case, it can be assigned to a group, using
Deno.BenchDefinition.group
option:
// url_bench.ts
Deno.bench("url parse", { group: "url" }, () => {
new URL("https://deno.land");
});
It is useful to assign several cases to a single group and compare how they perform against a "baseline" case.
In this example we'll check how performant is Date.now()
compared to
performance.now()
, to do that we'll mark the first case as a "baseline" using
Deno.BenchDefinition.baseline
option:
// time_bench.ts
Deno.bench("Date.now()", { group: "timing", baseline: true }, () => {
Date.now();
});
Deno.bench("performance.now()", { group: "timing" }, () => {
performance.now();
});
$ deno bench time_bench.ts
cpu: Apple M1 Max
runtime: deno 1.21.0 (aarch64-apple-darwin)
file:///dev/deno/time_bench.ts
benchmark time (avg) (min … max) p75 p99 p995
--------------------------------------------------------- -----------------------------
Date.now() 125.24 ns/iter (118.98 ns … 559.95 ns) 123.62 ns 150.69 ns 156.63 ns
performance.now() 2.67 µs/iter (2.64 µs … 2.82 µs) 2.67 µs 2.82 µs 2.82 µs
summary
Date.now()
21.29x times faster than performance.now()
You can specify multiple groups in the same file.
Running benchmarks Jump to heading
To run a benchmark, call deno bench
with the file that contains your bench
function. You can also omit the file name, in which case all benchmarks in the
current directory (recursively) that match the glob
{*_,*.,}bench.{ts, tsx, mts, js, mjs, jsx}
will be run. If you pass a
directory, all files in the directory that match this glob will be run.
The glob expands to:
- files named
bench.{ts, tsx, mts, js, mjs, jsx}
, - or files ending with
.bench.{ts, tsx, mts, js, mjs, jsx}
, - or files ending with
_bench.{ts, tsx, mts, js, mjs, jsx}
# Run all benches in the current directory and all sub-directories
deno bench
# Run all benches in the util directory
deno bench util/
# Run just my_bench.ts
deno bench my_bench.ts
⚠️ If you want to pass additional CLI arguments to the bench files use
--
to inform Deno that remaining arguments are scripts arguments.
# Pass additional arguments to the bench file
deno bench my_bench.ts -- -e --foo --bar
deno bench
uses the same permission model as deno run
and therefore will
require, for example, --allow-write
to write to the file system during
benching.
To see all runtime options with deno bench
, you can reference the command line
help:
deno help bench
Filtering Jump to heading
There are a number of options to filter the benches you are running.
Command line filtering Jump to heading
Benches can be run individually or in groups using the command line --filter
option.
The filter flags accept a string or a pattern as value.
Assuming the following benches:
Deno.bench({
name: "my-bench",
fn: () => {/* bench function zero */},
});
Deno.bench({
name: "bench-1",
fn: () => {/* bench function one */},
});
Deno.bench({
name: "bench2",
fn: () => {/* bench function two */},
});
This command will run all of these benches because they all contain the word "bench".
deno bench --filter "bench" benchmarks/
On the flip side, the following command uses a pattern and will run the second and third benchmarks.
deno bench --filter "/bench-*\d/" benchmarks/
To let Deno know that you want to use a pattern, wrap your filter with forward-slashes like the JavaScript syntactic sugar for a regex.
Bench definition filtering Jump to heading
Within the benches themselves, you have two options for filtering.
Filtering out (ignoring these benches) Jump to heading
Sometimes you want to ignore benches based on some sort of condition (for
example you only want a benchmark to run on Windows). For this you can use the
ignore
boolean in the bench definition. If it is set to true the bench will be
skipped.
Deno.bench({
name: "bench windows feature",
ignore: Deno.build.os !== "windows",
fn() {
// do windows feature
},
});
Filtering in (only run these benches) Jump to heading
Sometimes you may be in the middle of a performance problem within a large bench
class and you would like to focus on just that single bench and ignore the rest
for now. For this you can use the only
option to tell the benchmark harness to
only run benches with this set to true. Multiple benches can set this option.
While the benchmark run will report on the success or failure of each bench, the
overall benchmark run will always fail if any bench is flagged with only
, as
this is a temporary measure only which disables nearly all of your benchmarks.
Deno.bench({
name: "Focus on this bench only",
only: true,
fn() {
// bench complicated stuff
},
});
JSON output Jump to heading
To retrieve the output as JSON, use the --json
flag:
$ deno bench --json bench_me.js
{
"runtime": "Deno/1.31.0 x86_64-apple-darwin",
"cpu": "Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz",
"benches": [
"origin": "file:///dev/bench_me.js",
"group": null,
"name": "Deno.UnsafePointerView#getUint32",
"baseline": false,
"result": {
"ok": {
"n": 49,
"min": 1251.9348,
"max": 1441.2696,
"avg": 1308.7523755102038,
"p75": 1324.1055,
"p99": 1441.2696,
"p995": 1441.2696,
"p999": 1441.2696
}
}
]
}
Help us make these docs great!
Make a contribution
Deno's docs are open source. See something that's wrong or unclear? Submit a pull request:
Edit this page