Blazing fast, structured, leveled logging in Go.
go get -u github.com/uber-go/zap
Zap takes an opinionated stance on logging and doesn't provide any
printf-style helpers. Rather than logger.Printf("Failed to fetch URL %s (attempt %v), sleeping %s before retry.", url, tryNum, sleepFor), zap
encourages the more structured
logger.Info("Failed to fetch URL.",
zap.String("url", url),
zap.Int("attempt", tryNum),
zap.Duration("backoff", sleepFor),
)
This a bit more verbose, but it enables powerful ad-hoc analysis, flexible dashboarding, and accurate message bucketing. In short, it helps you get the most out of tools like ELK, Splunk, and Sentry. All log messages are JSON-serialized, though PRs to support other formats are welcome.
For compatibility with the standard library and bark, zap provides the
zwrap.Standardize and zbark.Barkify wrappers. Both are slower than the core
zap logger, but faster than the libraries they replace.
For applications that log in the hot path, reflection-based serialization and
string formatting are prohibitively expensive — they're CPU-intensive and
make many small allocations. Put differently, using encoding/json and
fmt.Fprintf to log tons of interface{}s makes your application slow.
Zap takes a different approach. It includes a reflection-free, zero-allocation JSON encoder, and it offers a variety of type-safe ways to add structured context to your log messages. It strives to avoid serialization overhead and allocations wherever possible, so collecting rich debug logs doesn't impact normal operations.
As measured by its own benchmarking suite, not only is zap more performant than comparable structured logging libraries — it's also faster than the standard library. Like all benchmarks, take these with a grain of salt.1
Log a message and 10 fields:
| Library | Time | Bytes Allocated | Objects Allocated |
|---|---|---|---|
| ⚡ zap | 1279 ns/op | 705 B/op | 2 allocs/op |
| logrus | 10369 ns/op | 5275 B/op | 78 allocs/op |
| go-kit | 6969 ns/op | 3204 B/op | 70 allocs/op |
| log15 | 22246 ns/op | 4783 B/op | 91 allocs/op |
| apex/log | 16379 ns/op | 3608 B/op | 63 allocs/op |
Log a message using a logger that already has 10 fields of context:
| Library | Time | Bytes Allocated | Objects Allocated |
|---|---|---|---|
| ⚡ zap | 231 ns/op | 0 B/op | 0 allocs/op |
| logrus | 8532 ns/op | 3438 B/op | 61 allocs/op |
| go-kit | 6874 ns/op | 2486 B/op | 48 allocs/op |
| log15 | 20462 ns/op | 4118 B/op | 70 allocs/op |
| apex/log | 13886 ns/op | 2384 B/op | 48 allocs/op |
Log a static string, without any context or printf-style formatting:
| Library | Time | Bytes Allocated | Objects Allocated |
|---|---|---|---|
| ⚡ zap | 222 ns/op | 0 B/op | 0 allocs/op |
| standard library | 565 ns/op | 32 B/op | 2 allocs/op |
| logrus | 3085 ns/op | 1336 B/op | 26 allocs/op |
| go-kit | 1061 ns/op | 624 B/op | 13 allocs/op |
| log15 | 5462 ns/op | 1351 B/op | 23 allocs/op |
| apex/log | 3009 ns/op | 584 B/op | 11 allocs/op |
Ready for adventurous users, but we're planning several breaking changes before releasing version 1.0. This milestone tracks our progress toward a stable release.
Released under the [MIT License](LICENSE.txt).
1 In particular, keep in mind that we may be benchmarking against slightly older versions of other libraries. Versions are pinned in zap's glide.lock file. ↩