A more efficient splitter for bytes and strings, with a focus on zero allocation, for Go. Use this where you would might use bytes.Split
or strings.Split
.
See pkg.go.dev
go get github.com/clipperhouse/split
import "github.com/clipperhouse/split"
text := "Hello, 世界. Nice dog! 👍🐶"
sep := " "
split := split.String(text, sep)
for split.Next() {
fmt.Println(split.Value())
}
Some initial benchmarks:
split.String
(this package)
1185 ns/op 404.28 MB/s 0 B/op 0 allocs/op
strings.Split
(standard library)
1267 ns/op 378.07 MB/s 1280 B/op 1 allocs/op
Overall, this package is a little faster, but more importantly, notice the difference in allocations. If you're on a hot path, this might add up, and reducing GC might help your app to scale.
The standard library collects all the splits at once into an array, and allocates to do so (this is true in other languages as well).
This package lazily iterates over each split as needed, and avoids that allocation. Think of it as streaming instead of batching.
If you do not actually need the array, but only need to iterate over the splits, this package may be useful.
This packages handles string
and []byte
. If you have an io.Reader
, we suggest bufio.Scanner
from the standard library.
We work to ensure that split.Bytes
and split.String
offer an identical API and results as their standard library counterparts, bytes.Split
and strings.Split
. Have a look at the tests to verify that this is true.
We've published a v0.1.0. Try it and leave feedback.
PR's are welcome, perhaps you'd like to implement a range iterator.