logfmt

key=value logger and parser

node-logfmt

"logfmt" is the name for a key value logging convention we've adopted at Heroku.

This library is for both converting lines in logfmt format to objects and for logging objects to a stream in logfmt format.

It provides a logfmt parser, logfmt stringifier, a logging facility, and both streaming and non-streaming body parsers for express and restify.

You should use this library if you're trying to write structured logs or if you're consuming them (especially if you're writing a logplex drain).

npm install logfmt

use

The logfmt module is a singleton that works directly from require.

var logfmt = require('logfmt');
 
logfmt.stringify({foo: 'bar'});
// 'foo=bar' 
 
logfmt.parse('foo=bar');
// {foo: 'bar'} 

It is also a constructor function, so you can use new logfmt to create a new logfmt that you can configure differently.

var logfmt2 = new logfmt;
 
// replace our stringify with JSON's 
logfmt2.stringify = JSON.stringify
 
// now we log JSON! 
logfmt2.log({foo: 'bar'})
// {"foo":"bar"} 
 
// and the original logfmt is untouched 
logfmt.log({foo: 'bar'})
// foo=bar 

accepts lines on STDIN and converts them to json

> echo "foo=bar a=14 baz=\"hello kitty\" cool%story=bro f %^asdf" | logfmt
{ "foo": "bar", "a": 14, "baz": "hello kitty", "cool%story": "bro", "f": true, "%^asdf": true }

accepts JSON on STDIN and converts them to logfmt

> echo '{ "foo": "bar", "a": 14, "baz": "hello kitty", \
"cool%story": "bro", "f": true, "%^asdf": true }' | logfmt -r
foo=bar a=14 baz="hello kitty" cool%story=bro f=true %^asdf=true

round trips for free!

> echo "foo=bar a=14 baz=\"hello kitty\" cool%story=bro f %^asdf" | logfmt | logfmt -r | logfmt
{ "foo": "bar", "a": 14, "baz": "hello kitty", "cool%story": "bro", "f": true, "%^asdf": true }

API

Serialize an object to logfmt format

Serializes a single object.

logfmt.stringify({foo: "bar", a: 14, baz: 'hello kitty'})
//> 'foo=bar a=14 baz="hello kitty"' 

Parse a line in logfmt format

logfmt.parse("foo=bar a=14 baz=\"hello kitty\" cool%story=bro f %^asdf code=H12")
//> { "foo": "bar", "a": '14', "baz": "hello kitty", "cool%story": "bro", "f": true, "%^asdf": true, "code" : "H12" } 

The only conversions are from the strings true and false to their proper boolean counterparts.

We cannot arbitrarily convert numbers because that will drop precision for numbers that require more than 32 bits to represent them.

Put this in your pipe and smoke it.

Creates a streaming parser that will automatically split and parse incoming lines and emit javascript objects.

Stream in from STDIN

process.stdin.pipe(logfmt.streamParser())

Or pipe from an HTTP request

req.pipe(logfmt.streamParser())

Pipe objects into the stream and it will write logfmt. You can customize the delimiter via the options object, which defaults to \n (newlines).

  var parseJSON = function(line) {
    if(!line) return;
    this.queue(JSON.parse(line.trim()))
  }
 
  process.stdin
    .pipe(split())
    .pipe(through(parseJSON))
    .pipe(logfmt.streamStringify())
    .pipe(process.stdout)

Example command line of parsing logfmt and echoing objects to STDOUT:

var logfmt  = require('logfmt');
var through = require('through');
 
process.stdin
  .pipe(logfmt.streamParser())
  .pipe(through(function(object){
    console.log(object);
  }))

Example HTTP request parsing logfmt and echoing objects to STDOUT:

var http    = require('http');
var logfmt  = require('logfmt');
var through = require('through');
 
http.createServer(function (reqres) {
  req.pipe(logfmt.streamParser())
     .pipe(through(function(object){
       console.log(object);
     }))
 
  res.writeHead(200, {'Content-Type': 'text/plain'});
  res.end('OK');
}).listen(3000);
  // streaming 
  app.use(logfmt.bodyParserStream());
  // buffering 
  app.use(logfmt.bodyParser());

Valid Options:

  • contentType: defaults to application/logplex-1

If you use the logfmt.bodyParserStream() for a body parser, you will have a req.body that is a readable stream.

Pipes FTW:

var app    = require('express')();
var http   = require('http');
var through = require('through');
var logfmt  = require('logfmt');
 
app.use(logfmt.bodyParserStream());
 
app.post('/logs', function(reqres){
  if(!req.body) return res.send('OK');
 
  req.body.pipe(through(function(line){
    console.dir(line);
  }))
 
  res.send('OK');
})
 
http.createServer(app).listen(3000);

Or you can just use the readable event:

var app    = require('express')();
var http   = require('http');
var logfmt  = require('logfmt');
 
app.use(logfmt.bodyParserStream());
 
// req.body is now a Readable Stream 
app.post('/logs', function(reqres){
  req.body.on('readable', function(){
    var parsedLine = req.body.read();
    if(parsedLine) console.log(parsedLine);
    else res.send('OK');
  })
})
 
http.createServer(app).listen(3000);

Valid Options:

  • contentType: defaults to application/logplex-1

If you use the logfmt.bodyParser() for a body parser, you will have a req.body that is an array of objects.

var logfmt   = require('logfmt');
 
app.use(logfmt.bodyParser());
 
// req.body is now an array of objects 
app.post('/logs', function(reqres){
  console.log('BODY: ' + JSON.stringify(req.body));
  req.body.forEach(function(data){
    console.log(data);
  });
  res.send('OK');
})
 
http.createServer(app).listen(3000);

test it:

curl -X POST --header 'Content-Type: application/logplex-1' -d "foo=bar a=14 baz=\"hello kitty\" cool%story=bro f %^asdf" http://localhost:3000/logs

Log an object to logfmt.stream (defaults to STDOUT)

Uses the logfmt.stringify function to write the result to logfmt.stream

logfmt.log({foo: "bar", a: 14, baz: 'hello kitty'})
//=> foo=bar a=14 baz="hello kitty" 
//> undefined 

Defaults to logging to process.stdout

logfmt.log({ "foo": "bar", "a": 14, baz: 'hello kitty'})
//=> foo=bar a=14 baz="hello kitty" 

logfmt.log() Accepts as 2nd argument anything that responds to write(string)

var logfmt = require('logfmt');
logfmt.log({ "foo": "bar", "a": 14, baz: 'hello kitty'}, process.stderr)
//=> foo=bar a=14 baz="hello kitty" 

Overwrite the default global location by setting logfmt.stream

var logfmt = require('logfmt');
logfmt.stream = process.stderr
 
logfmt.log({ "foo": "bar", "a": 14, baz: 'hello kitty'})
//=> foo=bar a=14 baz="hello kitty" 

You can have multiple, isolated logfmts by using new.

var logfmt = require('logfmt');
var errorLogger = new logfmt;
errorLogger.stream = process.stderr
 
logfmt.log({hello: 'stdout'});
//=> hello=stdout 
 
errorLogger.log({hello: 'stderr'});
//=> hello=stderr 

Returns a new logfmt with object's data included in every log call.

var logfmt = require('logfmt').namespace({app: 'logfmt'});
 
logfmt.log({ "foo": "bar", "a": 14, baz: 'hello kitty'})
//=> app=logfmt foo=bar a=14 baz="hello kitty" 
 
logfmt.log({})
//=> app=logfmt 
 
logfmt.log({hello: 'world'})
//=> app=logfmt hello=world 

Log how long something takes. Returns a new logfmt with elapsed milliseconds included in every log call.

  • label: optional name for the milliseconds key. defaults to: elapsed=<milliseconds>ms
var timer = logfmt.time();
timer.log();
//=> elapsed=1ms 

String label changes the key to <string>=<milliseconds>ms

var timer = logfmt.time('time');
timer.log();
//=> time=1ms 
timer.log();
//=> time=2ms 

If you'd like to include data, just chain a call to namespace.

var timer = logfmt.time('time').namespace({foo: 'bar'});
timer.log();
//=> time=1ms foo=bar 
timer.log();
//=> time=2ms foo=bar 

Accepts a Javascript Error object and converts it to logfmt format.

It will print up to logfmt.maxErrorLines lines.

var logfmt = require('logfmt');
logfmt.error(new Error('test error'));
//=> at=error id=12345 message="test error" 
//=> at=error id=12345 line=0 trace="Error: test error" 
//=> ... 
app.use(logfmt.requestLogger());
//=> ip=127.0.0.1 time=2013-08-05T20:50:19.216Z method=POST path=/logs status=200 content_length=337 content_type=application/logplex-1 elapsed=4ms 

If no formatter is supplied it will default to logfmt.requestLogger.commonFormatter which is based on having similiar fields to the Apache Common Log format.

Valid Options:

  • immediate: log before call to next() (ie: before the request finishes)
  • elapsed: renames the elapsed key to a key of your choice when in non-immediate mode

Defaults to immediate: true and elapsed: 'elapsed'

app.use(logfmt.requestLogger({immediate: true}, function(reqres){
  return {
    method: req.method
  }
}));
//=> method=POST 
app.use(logfmt.requestLogger({elapsed: 'request.time'}, function(reqres){
  return {
    "request.method": req.method
  }
}));
//=> request.method=POST request.time=12ms 

A formatter takes the request and response and returns a JSON object for logfmt.log

app.use(logfmt.requestLogger(function(reqres){
  return {
    method: req.method
  }
}));
//=> method=POST elapsed=4ms 

It's always possible to piggyback on top of the commonFormatter

app.use(logfmt.requestLogger(function(reqres){
  var data = logfmt.requestLogger.commonFormatter(req, res)
  return {
    ip: data.ip,
    time: data.time,
    foo: 'bar'
  };
}));
//=> ip=127.0.0.1 time=2013-08-05T20:50:19.216Z foo=bar elapsed=4ms 

Development

Pull Requests welcome.

> npm test

License

MIT