Cogger

When creating a new logger, you can configure behavior via the following attributes:

• id: The program/process ID which shows up in the logs as your id. Default: $PROGRAM_NAME. For example, if run within a demo.rb script, the id would be "demo",

• io: The input/output stream. This can be STDOUT/$stdout, a file/path, or nil. Default: $stdout.

• level: The severity level you want to log at. Can be :debug, :info, :warn, :error, :fatal, or :unknown. Default: :info.

• formatter: The formatter to use for formatting your log output. Default: Cogger::Formatter::Color. See the Formatters section for more info.

• tags: Global tagging for every log entry which must be an array of objects you wish to use for tagging purposes.

• mode: The binary mode which determines if your logs should be written in binary mode or not. Can be true or false and is identical to the binmode functionality found in the Logger class. Default: false.

• age: The rotation age of your log. This only applies when logging to a file. This is equivalent to the shift_age as found with the Logger class. Default: 0.

• size: The rotation size of your log. This only applies when logging to a file. This is equivalent to the shift_size as found with the Logger class. Default: 1,048,576 (i.e. 1 MB).

• suffix: The rotation suffix. This only applies when logging to a file. This is equivalent to the shift_period_suffix as found with the Logger class and is used when creating new rotation files. Default: %Y-%m-%d.

Given the above description, here’s how’d you create a new logger instance with all attributes:

Supported levels can be obtained via Cogger::LEVELS. Example:

You can use your environment to define the desired default log level. The default log level is: "info". Although, you can set the log level to any of the following:

While downcase is preferred for the log level, you can use upcased values as well. If the LOG_LEVEL environment variable is not set, Cogger will fall back to "info" unless overwritten during initialization. Example: Cogger.new level: :debug. Otherwise, an invalid log level will result in an ArgumentError.

Each instance can be mutated using the following messages:

Please see the Logger documentation for more information.

Templates are used by all formatters and adhere to the String Format Specification as used by Kernel#format. All specifiers, flags, width, and precision are supported except for the following restrictions:

• Use of reference by name is required which means %<demo>s is allowed but %{demo} is not. This is because reference by name is required for regular expressions and/or pattern matching.

• Use of the n$ flag is prohibited because this isn’t compatible with the above.

In addition to the above, the String Format Specification is further enhanced with the use of universal and individual directives which are primarily used by the color formatter but can prove useful for other formatters. Example:

Here’s a detailed breakdown of the above:

• Universal: Applies color universally to the entire template and requires you to:

  • Wrap your entire template in a and start (<example>) and end tag (</example>) which works much like an HTML tag in this context.

  • Your tag names must either be <dynamic></dynamic>, any default color (example: <green></green>), or alias (i.e. <your_alias></your_alias>) as supported by the Tone gem.

• Individual: Individual templates allow you to apply color to specific attributes and require you to:

  • Format your attributes as attribute:directive. The colon delimiter is required to separate your attribute for your color choice.

  • The color value (what follows after the colon) can be dynamic, any default color (example: green), or alias (i.e. your_alias) as supported by the Tone gem.

In addition to the general categorization of universal and individual tags, each support the following directives:

• Dynamic: A dynamic directive means that color will be determined by severity level only. This means if info level is used, the associated color (alias) for info will be applied. Same goes for warn, error, etc.

• Specific: A specific directive means the color you use will be applied without any further processing regardless of the severity level. This gives you the ability to customize your colors further in situations where dynamic coloring isn’t enough.

At this point, you might have gathered that there are specific keys you can use for the log event metadata in your template and everything else is up to you. This stems from the fact that Logger entries always have the following metadata:

• id: This is the program/process ID you created your logger with (i.e. Cogger.new id: :demo).

• severity: This is the severity at which you messaged your logger (i.e. logger.info).

• at: This is the date/time as which your log event was created.

This also means if you pass in these same keys as a log event (example: logger.info id: :bad, at: Time.now, severity: :bogus) they will be ignored.

The last key (or keys) is variable and customizable to your needs which is the log event message. The only special key is the tags key which is explained later. Here a couple of examples to illustrate:

💡 In situations where a message hash is logged but the keys of that hash don’t match the keys in the template, then an empty message will be logged. This applies to all formatters except the JSON formatter which will log any key/value that doesn’t have a nil value.

In addition to coloring to your log output, you can add emojis as well. Here are the defaults:

The :emoji formatter is the default formatter which provides dynamic rendering of emojis based on severity level. Example:

To add one or more emojis, you can chain messages together when registering them:

If you always want to use the same emoji, you could use the emoji formatter with a specific template:

As you can see, using a specific and non-dynamic emoji will always display regardless of the current severity level.

Aliases are specific to the Tone gem which allows you alias specific colors/styles via a new name. Here’s how you can use them:

The above would add a :haze alias which consists of bold white text on a purple background. Once added, you’d then be able to view a list of all default and custom aliases. You can also override an existing alias if you’d like something else.

Aliases are a powerful way to customize your colors and use short syntax in your templates. Building upon the alias, added above, you’d be able to use it in your templates as follows:

💡 These aliases are used by the color and emoji formatters but check out the Tone documentation and Formatters section below for further examples.

Multiple formatters are provided for you which can be further customized as needed. Here’s what is provided by default:

You can add a formatter by providing a key, class, and optional template. If a template isn’t supplied, then the formatter’s default template will be used instead (more on that shortly). Example:

Symbols or strings can be used interchangeably when adding/getting formatters. As mentioned above, a template doesn’t have to be supplied if you want to use the formatter’s default template which can be inspected as follows:

💡 When you find yourself customizing any of the default formatters, you can reduce typing by adding your custom configuration to the registry and then referring to it via it’s associated key when initializing a new logger.

Tags allow you to tag your messages at both a global and local (i.e. per message) level. For example, here’s what tagging looks like when used globally:

Each tag is wrapped in brackets (i.e. []) and you can use multiple tags:

You are not limited to string-based tags. Any object will work:

With the above, we have string, symbol, integer, hash, and proc tags. With hashes, you’ll always get a the key/value pair formatted as: key=value. Procs/lambdas allow you to lazy evaluate your tag at time of logging which provides a powerful way to acquire the current process ID, thread ID, and so forth.

In addition to global tags, you can use local tags per log message. Example:

You can also combine global and local tags:

As you can see, tags are highly versatile. That said, the following guidelines are worth consideration when using them:

• Prefer uppercase tag names to make them visually stand out.

• Prefer short names, ideally 1-4 characters since long tags defeat the purpose of brevity.

• Prefer consistent tag names by using tags that are not synonymous or ambiguous.

• Prefer using tags by feature rather than things like environments. Examples: API, DB, MAILER.

• Prefer the JSON formatter for structured metadata instead of tags. Logging JSON formatted messages with tags will work but sticking with a traditional hash, instead of tags, will probably serve you better.

Filters allow you to mask sensitive information you don’t want showing up in your logs. The default is an empty set:

To add filters, use:

Symbols and strings can be used interchangeably but are stored as symbols since symbols are used when filtering log entries. Once your filters are in place, you can immediately see their effects:

You can add multiple log streams (outputs) by using:

The above would log the "Demo." message to $stdout — the default stream — to the tmp/demo.log file, and to /dev/null. All attributes used to construct your default logger apply to all additional streams unless customized further. This means any custom template/formatter can be applied to your streams. Example:

In this situation, you’d get colorized output to $stdout and JSON output to the tmp/demo.log file.

There is a lot you can do with streams. For example, if you wanted to experiment with the same message formatted by multiple formatters, you could add a stream per format. Example:

Aborting a program is mostly syntax sugar for Command Line Interfaces (CLIs) which aids in situations where you need to log an error message and exit the program at the same time with an exit code of 1 (similar to how Kernel#abort behaves). This allows your CLI to log an error and ensure the exit status is correct when displaying status, piping commands together, etc. All of the arguments, when messaging #error directly, are the same. Here’s how it works:

Rack is implicitly supported which means your middleware must be Rack-based and must require the Rack gem since Cogger::Rack::Logger doesn’t explicitly require Rack by default. If these requirements are met then, to add HTTP request logging, you only need to use it. Example:

Like any other Rack middleware, Rails::Rack::Logger is initialized with your current application along with any custom options. Example:

The following defaults are supported:

The defaults can be customized. Example:

In the above example, we see Cogger.new overrides the default Cogger.new(formatter: :json). In practice, you’ll want to customize the logger and key map. Here’s how each default is configured to be used:

• logger: Defaults to JSON formatted logging but you’ll want to pass in the same logger as globally configured for your application in order to reduce duplication and save on memory.

• timer: The timer calculates the total duration of the request and defaults to nanosecond precision but you can swap this out with your own timer if desired. When providing your own timer, the only requirement is that the timer respond to the #call message with a block.

• key_map: The key map is used to map the HTTP Headers to keys (i.e. tags) used in the log output. You can use the existing key map, provide your own, or use a hybrid.

Once this middleware is configured and used within your application, you’ll start seeing the following kinds of log entries (depending on your specific settings and tags used):

Rails

To build upon the above — and if using the Rails framework — you could configure your application as follows:

The above defines Cogger as the default logger for the entire application, ensures Cogger::Rack::Logger is configured to use it and swaps itself with the default Rails::Rack::Logger so you don’t have two pieces of middleware logging the same HTTP requests.

Should you ever need quick access to the defaults, you can use:

This is primarily meant for display/inspection purposes, though.

Each instance can be inspected via the #inspect message:

You can also look at individual attributes:

When testing, you might find it convenient to rewind and read from the stream you are writing too (i.e. IO, StringIO, File). For instance, here is an example where I inject the default logger into my Demo class and then, for testing purposes, create a new logger to be injected which only logs to StringIO so I can buffer and read for test verification:

The ability to #reread is only available for the default (first) stream and doesn’t work with any additional streams that you add to your logger. That said, this does make it easy to test the Demo implementation while also keeping your test suite output clean at the same time. 🎉