Cri is a library for building easy-to-use commandline tools with support for nested commands.

The central concept in Cri is the command, which has option definitions as well as code for actually executing itself. In Cri, the commandline tool itself is a command as well.

Here’s a sample command definition:

To run this command, invoke the #run method with the raw arguments. For example, for a root command (the commandline tool itself), the command could be called like this:

Each command has automatically generated help. This help can be printed using Cri::Command#help; something like this will be shown:

Let’s disect the command definition and start with the first five lines:

These lines of the command definition specify the name of the command (or the commandline tool, if the command is the root command), the usage, a list of aliases that can be used to call this command, a one-line summary and a (long) description. The usage should not include a “usage:” prefix nor the name of the supercommand, because the latter will be automatically prepended.

Aliases don’t make sense for root commands, but for subcommands they do.

The next few lines contain the command’s option definitions:

Options can be defined using the following methods:

• Cri::CommandDSL#option or Cri::CommandDSL#opt

• Cri::CommandDSL#flag (implies no arguments passed to option)

• Cri::CommandDSL#required (implies required argument)

• Cri::CommandDSL#optional (implies optional argument)

All these methods take the short option form as their first argument, and a long option form as their second argument. Either the short or the long form can be nil, but not both (because that would not make any sense). In the example above, the --more option has no short form.

Each of the above methods also take a block, which will be executed when the option is found. The argument to the block are the option value (true in case the option does not have an argument) and the command.

The last part of the command defines the execution itself:

The Cri::CommandDSL#run method takes a block with the actual code to execute. This block takes three arguments: the options, any arguments passed to the command, and the command itself.

Instead of defining a run block, it is possible to declare a class, the command runner class (Cri::CommandRunner) that will perform the actual execution of the command. This makes it easier to break up large run blocks into manageable pieces.

Commands can have subcommands. For example, the git commandline tool would be represented by a command that has subcommands named commit, add, and so on. Commands with subcommands do not use a run block; execution will always be dispatched to a subcommand (or none, if no subcommand is found).

To add a command as a subcommand to another command, use the Cri::Command#add_command method, like this:

• Toon Willems

• Ken Coar

Thanks for Lee “injekt” Jarvis for Slop, which has inspired the design of Cri 2.0.