logger

A simple logger tool to send messages to the console at set levels

Usage

The logger function can be imported in to any and every file where logging is required just by using a simple import;

import { logger } from "@nexustech/logger";

It is then as easy as calling the various log levels in your code to output consistently labelled and timestamped log messages to the console. The timestamps are trimmed to only the microtime, and do not show the date, to prevent the timestamp from taking up too much screen space.

The arguments passed to the logger are infinite, and separated by commas (,).

const a = 1;
const b = "foo";

try {
  myFunction(a, b);
} catch (error) {
  logger.error("An error occured", error.message, "using", a, b);
}

// [ERROR] 23:19:51.450: An error occured myFunction is not defined using 1 foo

Variables can also be wrapped in to an object so that output will be labelled and coloured using the Node inspect utility

NB: inspect is not used when logger is running in a browser, and when in server mode color output is false

logger.error("An error occured", error.message, { a, b });

// [ERROR] 23:19:51.452: An error occured myFunction is not defined { a: 1, b: 'foo' }
// NB: the `1` would be yellow, and the `'foo'` would be green indicating a Number and a String

Logger Class

The Logger can be implemented as a class that has more options available, such as setting a correlation value to maintain a unique identity across microservices. The class can also have a custom expander injected (at construction), as well as logging in a standard object style using the serverMode flag or setting.

import { Logger, ServerMode } from "@nexustech/logger";

const logger = new Logger({ serverMode: ServerMode.STD });

logger.setCorrelation("123456")

{
  level: 1,
  severity: 'error',
  datetime: '2023-08-25T01:07:22.170Z',
  timestamp: 1692925642170,
  correlation: { id: '123456' },
  message: [
    'An error occured',
    'myFunction is not defined',
    { a: 1, b: 'foo' }
  ]
}

Class Constructor

The constructor paramaters are an initialization object that support the following options;

{
  correlation?: string;     // can be modified using .setCorrelation
  serverMode?: ServerMode;  // can be modified using .serverMode(ServerMode.AWS|GCP|STD|OFF)
  serverCall?: Function     // can be set at construction time only. Will receive the structured object as an argument.
  expandedMode?: boolean;   // can be modified using .expandedMode(true|false)
  logLimit?: number;        // can be modified using .setLimit(1-5, or LogLevel)
  expander?: LogExpander;   // can be set at construction time only.
}

Serialized Errors

If a call is made to the error factory with an Error type object passed as an argument, the serialize-error handler is invoked to return a JSON representation of the Error. For details on the handler, see the README at serialize-error on Github.

import { Logger } from "@nexustech/logger";

const logger = new Logger({ correlation: "123456", serverMode: "STD" });

{
  level: 1,
  severity: 'error',
  datetime: '2023-08-25T13:09:06.863Z',
  timestamp: 1692968946863,
  correlation: { id: '123456' },
  message: [
    {
      name: 'ReferenceError',
      message: 'myFunction is not defined',
      stack: 'ReferenceError: myFunction is not defined\n' +
        '    at file:///.../logger/.readme.mjs:33:3\n' +
        '    at ModuleJob.run (node:internal/modules/esm/module_job:194:25)'
    }
  ]
}

Log Expander

The log expander is the data serializer (not the log serializer) and can be any valid array function that follows the expected standard for Array.map function constraints, e.g.

const expander = (value, index?, array?) => {
  if (value && (typeof value == "object" || Array.isArray(value))) {
    return JSON.stringify(value, null, 2);
  }
  return value;
};

If the expander is replaced, the expanded and server flags may no longer be injected, thus expandedMode and serverMode are potentially moot on custom expanders, and therefore assumed to be always on (true).

the default expander will expand any JSON strings to JSON and should be adequate for most data serialization tasks. If you are logging a JSON string and specifically need to inspect the string, disable the expander using

logger.expandedMode(false);

Log Levels

Available log levels are

["log", "error", "warn", "info", "debug", "trace"];

The output level can be set using the setLevel function. This supports either a Number (1-5), or a LogLevel. It also has a second boolean argument to control output of the "success" message. Setting this false will turn the message off, otherwise it is true by default.

NB: Log level cannot be set lower than 1; All log and error messages will always be output by the logger.

// log all messages INFO and below (LOG, ERROR, and WARN)
logger.setLevel(LogLevel.INFO, false);

// no output

// log all messages
logger.setLevel(5);

// [  LOG] 01:26:50.301: logger: set to trace (5)

By default, log output is limited at INFO (3) unless the environment variable LOG_LEVEL is defined, or the setLevel function is called at run-time.

To set the log level for a single run, prefix the launch command with the desired log level

LOG_LEVEL=trace node my-script.js

Environment Variables

LOG_LEVEL can be set to any of the valid log levels. Case is not important.

LOG_EXPANDED can be set to true or false

LOG_MODE can be set to AWS, GCP, or STD to default the Logger class to server mode when constructed, using specific output formats that appear in cloud logging services, or to STDOUT (via console) when set to STD

LOG_DEPTH can be set to expand objects to a set depth. The default setting is 6 levels deep in expanded objects.

Replacement for console

All logging events are called with the same factory functions defined in console. If the terminal where the logger is sending output to does not support the relevant factory, the output is redirected to the log factory.

It should be possible to search and replace /console(\.[a-z]{3,5})/ with logger$1 across your project to implement logger everywhere there exists a console.<level> command, however this may need a global definition for logger, or an import in every file, depending on your project.

NB: Logger does not support the table and dir console functions

console.log("🚀 Launch message in the code");
console.debug("myFunction called");
console.trace({ a, b });

// Search & Replace...

logger.log("🚀 Launch message in the code");
logger.debug("myFunction called");
logger.trace({ a, b });

Installation

To use as a package in your project,

npm i @nexustech/logger

then in your project

import { logger } from "@nexustech/logger";

in the files that would output logs.

It is often expected that a Logger Service (class) is passed as a dependency to child classes from a parent, such that those classes inherit things like the correlation from the logger instance

Example: AWS Lambda function

handler

import { Logger } from "@nexustech/logger";

const logger = new Logger({ serverMode: "AWS" });

export default handler = async function (event: APIGatewayProxyEvent) {
  const correlation = getCorrelation(event); // possibly passed from previous microservice
  logger.setCorrelation(correlation);
  logger.debug("myHandler invoked", { event });
  // event checks...

  const payload = event.body;
  // payload checks

  const controller = new MyController(logger);
  return controller.myFunction(payload);
};

controller

export class MyController {
  constructor(private logger = console) {
    // ...
  }

  myFunction(payload: MyControllerPayload) {
    this.logger.debug("MyController:myFunction invoked", { payload });
    // ...
  }
}

In this configuration the Logger service can be passed down the chain and will maintain the correlation value, as well as allow for missing a missing dependency injection by replacing the logger with the standard console. (YMMV).

More advance dependency injection patterns can be achieved with other modules that are specifically designed for the job.

Modification

Clone the repo and use npm or yarn to pull dependencies.

npm install

Build

Packaging for end-use is normally handled with the project the logger is shipped with. Locally building for testing is done with tsc, with all files compiled to JS as separate files.

npm run build

Example Output

The output shown in this README can be generated with

node .readme.mjs

from the project root once installed and built.