UUID

A high-performance, zero-allocation platform for UUID versions 1-8.

Go

go get github.com/mikhaildadaev/uuid

Info

The latest stable version of uuid is v1.26.11.

Run Test

go test ./...
go test -bench=. ./...
go test -cover ./...
go test -race ./...

Key Features

• UUIDv1..v8 & Null — Generate all UUID versions and SQL-compatible null UUID;
• Parse & Validate — Parse and validate standard and null UUID strings;
• Metadata Extraction — Extract timestamp, sequence, node, POSIX, variant, and version;
• Serialization — Full Binary, JSON, and Text marshal/unmarshal support;
• SQL Integration — Native database/sql driver with Scan/Value interfaces;
• Zero-Allocation — Fast, allocator-free core routines on all hot paths.

Limits

• V2: posix should be 0..255;
• V3/V5: name should be 1..512 symbols (byte length);
• V8: node should be 0..16383;

Note

Strings name that are 0 bytes or larger than 512 bytes return a null UUID to prevent allocation.

Benchmarks

Info

The best way to compare libraries is to run benchmarks in your own environment with your own workload. Each project has unique requirements — latency, throughput, memory usage, and integration complexity — and no single test can cover them all.

I recommend that you test uuid alongside other libraries and choose the tool that best suits your needs.

Core Performance

Pure generation and serialization overhead. Benchmarks write to io.Discard — no I/O involved.

MultiThread

Version	Operations	Time (ns/op)	Memory (B/op)	Allocs (allocs/op)
V1	34.1M	29.29	0	0
V2	59.6M	16.79	0	0
V3	84.7M	11.81	0	0
V4	18.9M	52.88	0	0
V5	51.2M	19.51	0	0
V6	32.5M	30.75	0	0
V7	22.5M	44.45	0	0
V8	32.0M	31.21	0	0

SingleThread

Version	Operations	Time (ns/op)	Memory (B/op)	Allocs (allocs/op)
V1	13.5M	85.46	0	0
V2	32.1M	36.61	0	0
V3	9.9M	117.30	7	0
V4	26.7M	44.28	0	0
V5	7.7M	152.60	7	0
V6	13.7M	85.78	0	0
V7	10.7M	109.50	0	0
V8	10.2M	109.10	0	0

Note

All benchmarks measure pure generation and parsing overhead. Benchmarks marked as zero-allocation (0 B/op, 0 allocs/op) perform no heap allocations in hot-path operations — minor artifacts in single-threaded V3/V5 are from the Go testing framework itself, not from UUID generation. Real-world performance depends on CPU, memory latency, and concurrency.

Benchmarked on Intel Core i9-9880H (2.30 GHz)

Usage

import (
    "fmt"
    "log"
    "github.com/mikhaildadaev/uuid"
)
func main() {
    name := "github.com/mikhaildadaev/uuid"
    node := 1995
    posix := 0
    uid := "01968727-8c7e-8000-87cb-bdba4f634d9f"
    fmt.Println(uuid.NewV1())
    fmt.Println(uuid.NewV2(posix))
    fmt.Println(uuid.NewV3(uuid.NameSpaceURL,name))
    fmt.Println(uuid.NewV4())
    fmt.Println(uuid.NewV5(uuid.NameSpaceURL,name))
    fmt.Println(uuid.NewV6())
    fmt.Println(uuid.NewV7())
    fmt.Println(uuid.NewV8(node))
    parsed, err := uuid.Parse(uid)
    if err != nil {
        log.Fatal(err)
    }
    fmt.Println("Parsed:", parsed)
    fmt.Println("Is zero:", parsed.IsZero())
    var nu uuid.NullUUID
    if err := nu.Scan(nil); err != nil {
        log.Fatal(err)
    }
    fmt.Println("NULL valid:", nu.Valid)
    if err := nu.Scan(uid); err != nil {
        log.Fatal(err)
    }
    fmt.Println("UUID valid:", nu.Valid)
}