When we want to do logging from a library, we don't want to force the choice of
logging framework on the application developer. Instead, we want to use whatever
logging framework the application developer selected. anylogger let's you do
just that.
Library | App with debug | App with loglevel | App with ulog | App with log4js |
|---|---|---|---|---|
Installnpm i -P anylogger Useimport anylogger from 'anylogger'
const log = anylogger('my-library')
log('Anylogger is easy!')
Install your preferred logger and it's adapter as dev dependencies. | Installnpm i -P anylogger debug anylogger-debug Add to entry pointindex.jsimport "anylogger-debug" Useimport anylogger from 'anylogger'
const log = anylogger('my-app')
log('Anylogger is easy!')
| Installnpm i -P anylogger loglevel anylogger-loglevel Add to entry pointindex.jsimport "anylogger-loglevel" Useimport anylogger from 'anylogger'
const log = anylogger('my-app')
log('Anylogger is easy!')
| Installnpm i -P anylogger ulog Add to entry pointindex.jsimport "ulog" Useimport anylogger from 'anylogger'
const log = anylogger('my-app')
log('Anylogger is easy!')
| Installnpm i -P anylogger log4js anylogger-log4js Add to entry pointindex.jsimport "anylogger-log4js" Useimport anylogger from 'anylogger'
const log = anylogger('my-app')
log('Anylogger is easy!')
|
We, the Javascript community, really need a logging facade. There are dozens of logging libraries around and we library authors face a challenge. Which logger do we pick? Should we make this configurable? Should we just not log? Use the console directly? How do we deal with this complexity?
In software architecture, a facade hides a complex system behind a simple interface. In the context of our logging problem, we can have our library log to a simple facade object. In our application we back the facade object by the actual logging framework with an adapter.
So what we need is a simple and small logging facade and a bunch of adapters for popular loggers.
anyloggerA tiny ~330 bytes logging facade that you can include in your library to support logging, while at the same time allowing application developers to plug in any logging framework they choose.
Instead of building in your own library specific configuration mechanism,
or forcing the choice for a certain logging framework on your users,
or just abandoning logging altogether, choose anylogger and for just
~330 bytes shared between all libraries doing this, we can
plug in any framework of our choice and all libraries will automatically
start to use that framework. Wouldn't it be much better and easier?
index.html
<script src="https://unpkg.com/anylogger@1.0.11"></script>
<script>(function(){ // IIFE
var log = anylogger('index.html')
log.info('Logging is simple!')
})()</script>
Note that anylogger by default does nothing. You need an adapter to see output.
For example, to send all logging to the console, you can use
anylogger-console.
Depending on your project type, install anylogger, your logging framework of choice and an anylogger adapter if needed.
If you are building a library, install anylogger as a dependency:
npm install --save anylogger
This will add anylogger as a dependency to your package.json.
If you are building a library, you can use the logging framework you prefer without tightly coupling your library to it by installing that library and the adapter for it if needed as development dependencies:
For anylogger-console:
npm install --save-dev anylogger-console
For debug:
npm install --save-dev debug anylogger-debug
See anylogger-debug
For loglevel:
npm install --save-dev loglevel anylogger-loglevel
For ulog:
npm install --save-dev ulog
No adapter is needed for
ulog
For log4js:
npm install --save-dev log4js anylogger-log4js
See anylogger-log4js
If you are building an application project and have selected a logging framework, install anylogger, the selected logging framework and the adapter for that logging framework if needed.
For anylogger-console:
npm install --save anylogger anylogger-console
For debug:
npm install --save anylogger debug anylogger-debug
See anylogger-debug
For loglevel:
npm install --save anylogger loglevel anylogger-loglevel
For ulog
npm install --save anylogger ulog
No adapter is needed for
ulog
For log4js:
npm install --save anylogger log4js anylogger-log4js
See anylogger-log4js
Check out all available adapters.
Depending on the type of project, either just use anylogger, or also include the adapter.
In your library code, only use anylogger and restrict yourself to the Anylogger API to stay framework-independent:
my-library.js
var log = require('anylogger')('my-library')
my-library.js
import anylogger from 'anylogger'
const log = anylogger('my-library')
This way, your library does not get tightly coupled to any specific logger
In the tests for your library code, you can include an adapter and make your library use the logging framework of your choice without having to add it as a dependency for your library. You can add them as development dependencies.
my-library.test.js
// e.g. for ulog
require('ulog')
// all anylogger loggers will use ulog
var log = require('anylogger')('my-lbrary:test')
my-library.test.js
// e.g. for ulog
import 'ulog'
// all anylogger loggers will use ulog
import anylogger from 'anylogger'
const log = anylogger('my-lbrary:test')
In your main entry point, include your adapter or library with native support so it extends anylogger:
main.js
// e.g. for debug
require('anylogger-debug')
// all anylogger loggers will use debug
main.js
// e.g. for debug
import 'anylogger-debug'
// all anylogger loggers will use debug
In your other modules, use only anylogger and restrict yourself to the Anylogger API to stay framework-independent:
my-module.js
var log = require('anylogger')('my-module')
my-module.js
import anylogger from 'anylogger'
const log = anylogger('my-module')
Anylogger is very natural to use:
var log = require('anylogger')('my-module')
log('A log message')
log('debug', 'A debug message')
log('warn', 'A warning message')
log.info(log.name + ' starting...')
log.error('Something went wrong', new Error('Oh no!'))
if (log.enabledFor('warn')) {
log.warn(expensiveArguments())
}
If you are able to restrict yourself to the Anylogger API, your code will be framework independent and will work with any supported logging library.
log.info('Logging is easy!')
So what does this API look like?
function anylogger(name, options) => logger
The main function to call to get a logger. Accepts two arguments.
The name of the logger. String. Optional. Defaults to undefined.
The recommended format is <package-name>[:<sub>[:<sub> [..]]],
as this is the convention
used by the highly popular debug module. But you are free to pick any name
you want. You only get a logger if you supply a name. If the name is
not given anylogger() will return an object containing all loggers,
keyed by name.
An optional options object. Object. Optional. Defaults to undefined.
The use of such options objects varies wildly amongst implementations so
it is recommended to avoid using it where possible. However in case of
implementations that require it, anylogger passes any options object it
is given on to anylogger.new to allow it to be used
where needed.
When no arguments are given anylogger returns an object containing all loggers created so far, keyed by name.
When a name is given anylogger returns the existing logger with that
name, or creates a new one by calling anylogger.new.
The returned logger adheres to the Logging API described below.
The logger returned by anylogger is a function that can
do logging on it's own:
log('message') // logs a message at `log` level
log('info', 'message') // logs a message at `info` level
In addition, the logger looks like a simple console object:
log.debug('message')
log.info('message')
Because of this, the logger created by anylogger is compatible with most logging frameworks out there, which mostly use one or both of these approaches.
The main API looks like this (in pseudo code):
log: function([level='log'], ...args)
log.error: function(...args)
log.warn: function(...args)
log.info: function(...args)
log.log: function(...args)
log.debug: function(...args)
log.trace: function(...args)
log.enabledFor: function(level) => truthy or falsey
And that's about it. However this covers the basic logging needs.
Note that all logging methods here are part of the upcoming Console standard, but not all platforms and frameworks support all of them. In particular the
debugmethod is not available everywhere. Anylogger will make sure that thedebugfunction is polyfilled if needed.
Is your logging framework not supported? No fear, just...
To write an anylogger adapter, you need to make a project that includes both anylogger and the logging framework the adapter is for as peer dependencies.
You then need to modify one or more of the anylogger extension points so the created loggers will be compliant with both the anylogger Logging API as well as with the logging framework's own API.
It is recommended you call your library anylogger-[adapter], where
[adapter] should be replaced with the name of the logging framework
the adapter is for. For example, the adapter for debug is called
anylogger-debug.
In addition, it is recommended you add the keyword "anylogger" to the
package.json file of your adapter project, so it will show up in the list of
available adapters.
The process of logger creation and invocation is split up in such a way as to optimize possible extension points allowing extensions to re-use anylogger functionality and avoid having to duplicate code. The extension points are:
anylogger.levels = {error:1, warn:2, info:3, log:4, debug:5, trace:6}
An object containing a mapping of level names to level values.
To be compliant with the anylogger API, loggers should support at least the log methods corresponding to the default levels, but they may define additional levels and they may choose to use different numeric values for all the levels.
The guarantees the Anylogger API makes are:
anylogger.levelserror, warn, info, log, debug and trace are always thereNote that the Anylogger API explicitly does not guarantee that all levels have distinct values or that the numeric values will follow any pattern or have any specific order. For this reason it is best to think of levels as separate log channels, possibly going to different output locations.
You can replace or change this object to include levels corresponding with
those available in the framework you are writing an adapter for. Please
make sure to always include the default levels as well so all code can
rely on the 6 console methods error, warn, info, log, debug and
trace to always be there.
anylogger.new(name, options) => logger
Creates a new logger function that calls anylogger.log when invoked.
Uses new Function(..) to create a named function so that function.name
corresponds to the module name given. Polyfills function.name on platforms
where it is not natively available.
The name of the new logger. String. Required.
An optional options object. Object. Optional.
If the logging framework you are writing an adapter for uses an options
object, you should override anylogger.new and do something useful with
the options object here (set it as a property on the logger for example),
because the default implementation just ignores it.
Instead of completely trying to replace the original method, I recommend you chain it to include your one-time customizations like this:
import anylogger from 'anylogger'
// save the original function
const make = anylogger.new
// override anylogger.new
anylogger.new = (name, options) => {
// call the original function to chain it
var logger = make(name, options)
// do something useful with the options object
logger.options = options
// return the customized logger
return logger
}
All anylogger methods are independent of
thisso they can all be easily chained
If you need to re-apply customizations any time relevant config changes (such
as active log level changing), override anylogger.ext.
anylogger.ext(logger) => logger
Called when a logger needs to be extended, either because it was newly created, or because it's configuration or settings changed in some way.
This method must ensure that a log method is available on the logger for
each level in anylogger.levels.
When overriding anylogger.ext, please ensure the function can safely
be called multiple times on the same object
The logger that should be (re-)extended. Function. Required.
The default implementation loops over the
anylogger.levels and creates noop methods for each level.
Additionally it creates a noop enabledFor that always returns undefined.
You can override or chain this method to change the way the log methods are
(re-)created. In a library that supports log levels, all methods corresponding
to log levels equal to or higher than the currently active level might be
replaced with console methods instead. Or maybe the destination of the log
messages might change dynamically based on configuration. Apply such changes
in anylogger.ext as it will be called again whenever relevant config changes.
This allows adapters to (re-)extend the logger so that the new configuration
takes effect.
You may need to ensure in your adapter that
anylogger.extis called whenever relevant config changes. By hooking into setters for example.
anylogger.log([level='log'], ...args)
The log function returned by anylogger calls anylogger.log, which determines
the log level and invokes the appropriate log method.
Please have a look at the source it should make it more clear how to write an adapter. Also consider studying the available adapters and learn by example.
If you wrote an anylogger adapter, make sure to share it back with the
community. Publish it to NPM for all to use!
Credits go to these people, who helped with this project:
Add an issue in this project's issue tracker to let me know of any problems you find, or questions you may have.
© 2020 by Stijn de Witt. Some rights reserved. Contributions by Jakub Jirutka.
Licensed under the MIT Open Source license.
The GZIP algorithm is available in different flavours and with different possible compression settings. The sizes quoted in this README have been measured using gzip-size by Sindre Sorhus, your mileage may vary.
FAQs
Get a logger. Any logger.
We found that anylogger demonstrated a not healthy version release cadence and project activity because the last version was released a year ago. It has 1 open source maintainer collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Security News
Research
A malicious npm package disguised as a WhatsApp client is exploiting authentication flows with a remote kill switch to exfiltrate data and destroy files.
Security News
PyPI now supports digital attestations, enhancing security and trust by allowing package maintainers to verify the authenticity of Python packages.
Security News
GitHub removed 27 malicious pull requests attempting to inject harmful code across multiple open source repositories, in another round of low-effort attacks.