Skip to content

metapic/pino-hook-structured-logger

Repository files navigation

pino-hook-structured-logger

Adds support for structured logs to pino via a logMethod hook.

Structured Logging

Captures structured data and adds it to the JSON payload. The unformatted message template is added to the payload too. Multiple ways of passing structured data are supported, including the default pino mergingObject.

TL;DR - Example Code

See all supported log formats in the spec.

// arguments after the message are captured as structured data
// in order of occurrence in the message template:
logger.info('User {user_id} logged in from {location}', 12345, 'Salzburg')

// alternatively, a dedicated object as first argument after the message can
// be used for structured data. this is more verbose, but *may* be more readable.
logger.info('User {user_id} logged in from {location}', {
  user_id: 12345,
  location: 'Salzburg',
})

// existing bindings (child logger) are respected.
// the dedicated structured data object takes precedence over the bindings,
// if both are present and there are any conflicting keys.
const loggerWithBindings = logger.child({ user_id: 12345 })
loggerWithBindings.info('User {user_id} logged in from {location}', 'Salzburg')

// pino's mergingObject is a valid source of structured data too.
// the dedicated structured data object takes precedence over the mergingObject,
// if both are present and there are any conflicting keys.
logger.info(
  { user_id: 12345, location: 'Salzburg' },
  'User {user_id} logged in from {location}',
)

// any combination of the above is supported too.
// though you should *probably* not do this...

logger.info({ user_id: 12345 }, 'User {user_id} logged in from {location}', {
  location: 'Salzburg',
})

logger.info(
  'User {user_id} logged in from {location}',
  { location: 'Salzburg' },
  12345,
)

All examples produce this output:

{
  "level": 30,
  "time": 1752419866224,
  "pid": 54900,
  "hostname": "953e747d91b4",
  "msg": "User 12345 logged in from Salzburg",
  "msg_tpl": "User {user_id} logged in from {location}",
  "data": {
    "user_id": 12345,
    "location": "Salzburg"
  }
}

Additional Arguments

Any additional, unmatched arguments are captured and added to the log output with the args key:

logger.info(
  'User {user_id} logged in from {location}',
  12345,
  'Salzburg',
  'arg1',
  { arg_num: 2 },
)
{
  //...
  "msg": "User 12345 logged in from Salzburg",
  "msg_tpl": "User {user_id} logged in from {location}",
  "data": {
    "user_id": 12345,
    "location": "Salzburg"
  },
  "args": [
    "arg1",
    {
      "arg_num": 2
    }
  ]
}

Usage

Simply configure your pino instance to use structuredLogger() as hooks. You can optionally customize the behaviour of the structured logger by passing a configuration object.

import pino from 'pino'
import { structuredLogger } from '@metapic/pino-hook-structured-logger'

const logger = pino({
  level: 'debug',
  hooks: structuredLogger({
    messageTemplateKey: 'msg_tpl',
    dataKey: 'data',
    argsKey: 'args',
    unwrapErrors: true,
  }),
})

Performance

The added functionality comes at a cost: the structured logger achieves roughly half the throughput of the standard pino logger. Pino is pretty fast, so in most real-world use cases this throughput is totally fine.

See benchmark tests in index.bench.ts.

$ npm run bench

> @metapic/pino-hook-structured-logger@0.0.0-development bench
> vitest bench

Benchmarking is an experimental feature.
Breaking changes might not follow SemVer, please pin Vitest's version when using it.

DEV  v3.2.4 /workspaces/pino-hook-structured-logger

  ✓ test/index.bench.ts > info with params 2811ms
      name                       hz     min      max    mean     p75     p99    p995    p999     rme  samples
    · standard logger    893,856.90  0.0006  33.1586  0.0011  0.0007  0.0013  0.0016  0.0038  ±9.51%   893857
    · structured logger  388,408.52  0.0018   7.2970  0.0026  0.0020  0.0034  0.0038  0.0085  ±4.67%   388409

  ✓ test/index.bench.ts > info with message only 3037ms
      name                         hz     min      max    mean     p75     p99    p995    p999      rme  samples
    · standard logger    1,073,454.90  0.0003  29.2559  0.0009  0.0004  0.0015  0.0017  0.0041  ±12.81%  1073455
    · structured logger    547,664.58  0.0011  15.1655  0.0018  0.0013  0.0025  0.0027  0.0062   ±6.78%   547665

BENCH  Summary

  standard logger - test/index.bench.ts > info with params
    2.30x faster than structured logger

  standard logger - test/index.bench.ts > info with message only
    1.96x faster than structured logger