FEATURE: add the bench command to the standard library (#8969)

Should address the first point of
- https://github.com/nushell/nushell/issues/8696.

# Description
i've seen a few appearances of this `benchmark` idea in recent works on
`nu-std`, so i thought it would be great to add it finally.

# User-Facing Changes
a new `std bench` command to measure the performance of `nushell`
closures and code blocks.

# Tests + Formatting
- 🟢 `toolkit fmt`
- 🟢 `toolkit clippy`
-  `toolkit test`
- 🟢 `toolkit test stdlib`

# After Submitting
```
$nothing
```
This commit is contained in:
Antoine Stevan 2023-04-28 09:07:23 +02:00 committed by GitHub
parent 44493dac51
commit 3268ecd116
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23

View File

@ -148,3 +148,73 @@ export def clip [
notify-send "std clip" "saved to clipboard" notify-send "std clip" "saved to clipboard"
} }
} }
# convert an integer amount of nanoseconds to a real duration
def "from ns" [] {
[$in "ns"] | str join | into duration
}
# run a piece of `nushell` code multiple times and measure the time of execution.
#
# this command returns a benchmark report of the following form:
# ```
# record<
# mean: duration
# std: duration
# times: list<duration>
# >
# ```
#
# > **Note**
# > `std bench --pretty` will return a `string`.
#
# # Examples
# measure the performance of simple addition
# > std bench { 1 + 2 } -n 10
# ╭───────┬────────────────────╮
# │ mean │ 4µs 956ns │
# │ std │ 4µs 831ns │
# │ │ ╭───┬────────────╮ │
# │ times │ │ 0 │ 19µs 402ns │ │
# │ │ │ 1 │ 4µs 322ns │ │
# │ │ │ 2 │ 3µs 352ns │ │
# │ │ │ 3 │ 2µs 966ns │ │
# │ │ │ 4 │ 3µs │ │
# │ │ │ 5 │ 3µs 86ns │ │
# │ │ │ 6 │ 3µs 84ns │ │
# │ │ │ 7 │ 3µs 604ns │ │
# │ │ │ 8 │ 3µs 98ns │ │
# │ │ │ 9 │ 3µs 653ns │ │
# │ │ ╰───┴────────────╯ │
# ╰───────┴────────────────────╯
#
# get a pretty benchmark report
# > std bench { 1 + 2 } --pretty
# 3µs 125ns +/- 2µs 408ns
export def bench [
code: closure # the piece of `nushell` code to measure the performance of
--rounds (-n): int = 50 # the number of benchmark rounds (hopefully the more rounds the less variance)
--verbose (-v): bool # be more verbose (namely prints the progress)
--pretty: bool # shows the results in human-readable format: "<mean> +/- <stddev>"
] {
let times = (
seq 1 $rounds | each {|i|
if $verbose { print -n $"($i) / ($rounds)\r" }
timeit { do $code } | into int | into decimal
}
)
if $verbose { print $"($rounds) / ($rounds)" }
let report = {
mean: ($times | math avg | from ns)
std: ($times | math stddev | from ns)
times: ($times | each { from ns })
}
if $pretty {
$"($report.mean) +/- ($report.std)"
} else {
$report
}
}