Overview
This module provides a simple logging class, which appends to a text log file with bracket-delimited columns. You can define any number of columns you want, or use some of the built-in auto-populated columns. You can populate columns by name, or by using one of the shortcut methods described below.
Usage
Use npm to install the module:
npm install pixl-logger
Then use require()
to load it in your code:
var Logger = require('pixl-logger');
To use the module, instantiate an object, and start logging:
var columns = ['hires_epoch', 'date', 'hostname', 'component', 'category', 'code', 'msg', 'data'];
var logger = new Logger( 'logs/debug.log', columns );
logger.set('hostname', 'myserver.com');
logger.print({
category: 'debug',
component: 'main',
code: 1,
msg: "Hello log!"
});
This would append the following like to logs/debug.log
:
[1423462332.437][2015-02-08 22:12:12][myserver.com][main][debug][1][Hello log!][]
Some column names are special, and automatically populated (see below), but the rest are free-form. You can include any number of columns you like, and name them however you want.
All the log "columns" are basically just key/value pairs stored in an args
property in the class, which don't have to be specified on every call to print()
. Meaning, you can set some of them once, and only have to set them again when they change. Example:
logger.set('component', 'main');
logger.set('category', 'debug');
logger.print({
code: 1,
msg: "Hello log!"
});
Note that when you pass arguments to print()
they are not saved for subsequent calls. They are only used for that one log row. You need to call set()
to make them stick.
You can also fetch arg values using get()
. Pass in a key to fetch one arg, or omit to get the entire args
object back.
Data Cleansing
In order to protect the log format and syntax, all column values are "cleansed" before being written. Specifically, any newlines are converted to single spaces, and the character sequence ][
is stripped (as it would corrupt the log column layout). All other characters are allowed. Example:
logger.debug( 1, "This won't go well\n[Hi][There]\r\nGoodbye.\n" );
This would be logged as:
[1423466726.159][2015-02-08 23:25:26][myserver.com][main][debug][1][This won't go well [HiThere] Goodbye. ]
Shortcut Methods
The logger library provides the following three shortcut methods, which accept a list of common arguments instead of a hash:
debug
The debug()
method is designed to assist with writing to a debug log. It automatically sets the category
column to debug
. It requires two arguments, which are values for the code
(debug level) and msg
columns, with the 3rd argument being an optional data
object, if you want. Examples:
logger.debug( 1, "Logged at debug level 1" );
logger.debug( 2, "Logged at debug level 2", {some:"data"} );
An extra feature with the debug()
call is that you can set a debugLevel
arg on your class instance, and it'll only log entries if they have an equal or lower level (a.k.a. code). So imagine this setup:
logger.set( 'debugLevel', 2 );
logger.debug( 1, "Logged at debug level 1" );
logger.debug( 2, "Logged at debug level 2" );
logger.debug( 3, "This won't be logged at all!" );
In this case we set the debugLevel
arg to 2, so only the first two calls will be logged. The third call, which is logged at a higher (more verbose) level than the debugLevel
value, will be silently skipped.
shouldLog
If you need to see if a particular debug level would be logged, you can call the shouldLog()
method and pass in a level to query. If the current debugLevel
is equal to or higher than the level number passed in, it returns true
, otherwise it returns false
. Example:
if (logger.shouldLog(9)) {
logger.debug(9, "Complex data here: " + complex_data);
}
The idea is that entire code block wouldn't even be executed if the debugLevel
was less than 9, saving you the CPU time of serializing the complex data to be logged.
Keep in mind that you can simply pass an Object to debug()
as an optional 3rd argument, and it'll automatically be serialized to JSON and logged if the debug level is acceptable.
error
The error()
method is designed to assist with logging errors. It automatically sets the category
column to error
. It requires two arguments, which are values for the code
and msg
columns, with the 3rd argument being an optional data
object, if you want. Example:
logger.error( 1005, "An error of type 1005 occurred!" );
This would be equivalent to calling print()
with the following:
logger.print({
category: 'error',
code: 1005,
msg: "An error of type 1005 occurred!"
});
transaction
The transaction()
method is designed to assist with logging transactions. A "transaction" is whatever action you define in your app as something you want logged, for audit or replay purposes. It automatically sets the category
column to transaction
. It requires two arguments, which are values for the code
and msg
columns, with the 3rd argument being an optional data
object, if you want. Example:
logger.transaction( "user_update", "User jhuckaby was updated", {username:"jhuckaby"} );
This would be equivalent to calling print()
with the following:
logger.print({
category: 'transaction',
code: "user_update",
msg: "User jhuckaby was updated",
data: {username:"jhuckaby"}
});
Echo to Console
If you set the echo
arg to any true value, the logger will echo all log entries to process.stdout, in addition to the log file. This is useful for running CLI scripts or your app in debug mode. Example:
logger.set( 'echo', true );
logger.debug( 1, "This will be logged and echoed to the console!" );
Colored Logs
If you set the color
arg to any true value, the logger will echo all log entries in color (assuming you have a terminal that supports color), using the chalk module. The color sequence is gray
, red
, green
, yellow
, blue
, magenta
and cyan
. If your log has more than 7 columns, the colors repeat. The bracket dividers are printed in dim
. Here is a screenshot:
Example:
logger.set( 'echo', true );
logger.set( 'color', true );
logger.debug( 1, "This will be colored in the console!" );
Note that the color only affects the local echo in your terminal. The log file itself is still written in plain text.
Last Line Logged
To grab the last line logged by the logger, pull the lastRow
property off the class instance. It is the fully formatted line, including an EOL. Example:
var line = logger.lastRow;
Special Column Names
Several column names are special, in that they are automatically populated, or have special behavior. Here they are:
epoch
Any column named epoch
will automatically be populated with the current local server time, represented in integer Epoch seconds. Example:
[1423433821]
hires_epoch
Any column named hires_epoch
will automatically be populated with the current local server time, represented in high precision floating point Epoch seconds, with up to 3 digits after the decimal. Example:
[1423433807.277]
date
Any column named date
will automatically be populated with a human-friendly version of the current local server time, in YYYY-MM-DD HH:MI:SS
format. Example:
[2015-02-08 14:16:58]
hostname
Any column named hostname
will automatically be populated with the server's hostname, obtained by calling os.hostname() once at construction. Example:
[host01.mydomain.com]
pid
Any column named pid
will automatically be populated with the current Process ID (PID), obtained by calling process.pid once at construction. Example:
[13702]
data
The data
column is special, in that if it contains an object, it will be serialized to JSON. Example:
[{"code":0,"description":"Success"}]
category
The category
column is only special in that the shortcut methods debug()
, error()
and transaction()
automatically populate it to match their names.
Special Hooks
There are a number of optional hooks available for you to customize your logs even further. These can be specified in the args
object passed to the constructor, by calling set()
, or simply by setting them on an instance object.
pather
The pather
hook allows you to intercept and alter the log path on disk per each log row, and generate your own path based on the current args
passed to the logger. Using this you can change the log file per each log row. Your hook function is passed the current path, and the current args
object containing all the column keys/values. Example:
logger.pather = function(path, args) {
return '/var/log/MyApp-' + args.category + '.log';
};
In this example the category
column is used to dynamically construct the log filename. So if the category
column was set to debug
, the log path would be /var/log/MyApp-debug.log
. This is a great way to generate multiple logs based on any criteria you want.
filter
The filter
hook allows you to intercept each log column value as it is being added to the columns array for serialization. This is typically where column values are stripped of any illegal characters, and whitespace compressed. If you specify this hook, the default cleansing operation does not take place, and your function is expected to do all the filtering necessary. Your hook function is called for each column in each row, so keep that in mind for performance purposes, and passed the current column value, and its array index. Example:
logger.filter = function(value, idx) {
if (typeof(value) == 'object') value = JSON.stringify(value);
return value.toString().replace(/[\r\n]/g, ' ').replace(/\]\[/g, '');
};
Note that it is up to your filter function to convert objects into strings here, if you are trying to log any raw objects.
serializer
The serializer
hook allows you to intercept the log columns just as they are being serialized into a string for appending to your log file. Using this you can completely change how your log file is formatted. The default format is [bracket][delimited][rows]
but you can serialize columns into any format of string you want. Your function is passed the current columns array (filtered) and the args
object. For example, here is how to produce a CSV (comma-separated) log:
logger.serializer = function(cols, args) {
return cols.join(',') + "\n";
};
Your serializer function is expected to return a full formatted log row (line) ending in a EOL character.
echoer
The echoer
hook allows you to control exactly how your log is echoed to the console, if Echo to Console mode is enabled. The default behavior is to simply reprint the formatted log line to STDOUT, in addition to appending it to your log file. However, if you specify the echoer
hook, this no longer happens, and instead your function is expected to do the echoing. The function is passed the formatted log line string (already appended to the log file), the array of columns that were serialized (these are filtered), and the current args
object. Example:
logger.echoer = function(line, cols, args) {
console.log( args.category + ": " + args.msg );
if (args.data) console.dir(args.data);
};
In this example we are only echoing certain columns to the console (just category
and msg
) using the args
object, and allowing Node.js to serialize the data column via console.dir(). This way if you are using a debugger, it may be navigable.
If you specify a filesystem path as the echoer
instead of a function, it will be appended to. Using this technique you can have the logger append to two different log files at once (the primary log and the echoer log). This file append honors the sync
flag as well.
Note that your echoer
hook is only ever called if Echo to Console mode is enabled.
Rotating Logs
To rotate a log file, call the rotate()
method, passing in a destination filesystem path and a callback. This will atomically move the file to the destination directory/filename, attempting a rename, and falling back to a "copy to temp file + rename" strategy. Example:
logger.rotate( '/logs/pickup/myapp.log', function(err) {
if (err) throw err;
} );
If you omit a filename on the destination path and leave a trailing slash, the source log filename will be appended to it.
You can actually rotate any log file you want by specifying three arguments, with the custom source log file path as the first argument. Example:
logger.rotate( '/path/to/logfile.log', '/logs/pickup/otherapp.log', function(err) {
if (err) throw err;
} );
Archiving Logs
You can also "archive" logs using the archive()
method. Archiving differs from rotation in that the log file is atomically copied to a custom location which may contain date/time directories (all auto-created as needed), and then the file is compressed using gzip. You can archive any number of logs at once by using filesystem glob syntax. Example:
var src_spec = '/logs/myapp/*.log';
var dest_path = '/archives/myapp/[yyyy]/[mm]/[dd]/[filename]-[hh].log.gz';
var epoch = ((new Date()).getTime() / 1000) - 1800;
logger.archive( src_spec, dest_path, epoch, function(err) {
if (err) throw err;
} );
This example would find all the log files found in the /logs/myapp/
directory that end in .log
, and archive them to destination directory /archives/myapp/[yyyy]/[mm]/[dd]/
, with a destination filename pattern of [filename]-[hh].log.gz
. All the bracket-delimited placeholders are expanded using the timestamp provided in the epoch
variable. The special [filename]
placeholder expands to the source log filename, sans extension. All directories are created as needed.
Sync or Async
By default, the logger will append to your log files asynchronously. This has the benefit of not blocking your main thread, and can help if your log drive is suffering lag or high I/O wait. But it may cause issues with log entries appearing out of order for extremely high traffic apps, and also some final log entries may be lost if process.exit()
is called immediately after.
To get around these potential issues, you can write log entries synchronously. Just set the sync
arg to true:
logger.set( 'sync', true );
logger.debug( 1, "This will be logged synchronously, even if we exit right NOW!" );
process.exit(0);
License
The MIT License
Copyright (c) 2015 - 2018 Joseph Huckaby
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.