Custom commands

Seyon Vasantharajan edited this page Feb 25, 2017 · 8 revisions

Custom commands

Quick Menu:

Back to Main Menu


Pry ships with a large number of builtin commands to cover common use-cases. Unfortunately we can't hope to include every command that might be useful to anyone, so we also make it easy to define new commands for yourself. This page explains how to do so.

For further explanation of the concepts behind commands, and how to use them, see the page on the command system.

### Block commands

Block commands are a quick way of definining simple commands, unlike Class commands they don't do any options parsing making them only suitable for quick hacks and personal utilities.

To create a new block command, you call Pry::CommandSet#block_command, for example:

Pry::Commands.block_command "hello", "Say hello to three people" do |x, y, z|
  output.puts "hello there #{x}, #{y}, and #{z}!"

This creates a command, hello which you can call from within Pry:

pry(main)> hello tom dick harry
hello there tom, dick and harry!

It also sets up the commands description, so that it appears in the list of help. For block commands the description is also used for the output of help <command>:

pry(main)> help
hello         Say hello to three people
pry(main)> help hello
Say hello to three people

As illustrated in the example, and explained further in Conventions, you should use output.puts rather than puts to make your command easily re-usable in situations where Pry is not using $stdout.

Just as with normal commands, block commands are executed with a number of Context methods available.

Block command options

The third argument to Pry::CommandSet#block_command are the command options, these define extra behaviours for commands, for example setting keep_retval allows you to return a value from your command:

Pry::Commands.block_command "add-numbers", "Add a list of numbers", :keep_retval => true do |*args|

This means that the output will be pretty-printed using the user's pretty printer and more importantly, it is made available in _.

pry(main)> add-numbers 1 2 3 4
=> 10
pry(main)> _ + 5
=> 15
### Class commands

Creating a simple command is very similar to creating a block command, though a little more verbose:

Pry::Commands.create_command "echo" do
  description "Echo the input: echo [ARGS]"

  def process
    output.puts args.join(" ")

Every time the command is invoked by the user, a new instance of the command class is created, it is setup as described in the Complete example, and then the process method is called.

Class commands were designed to help commands introspectable, they provide a declarative syntax for setting up the description, and in addition to setting up help like a Block command, class commands also respond to --help:

pry(main)> help
echo      Echo the input: echo [ARGS]
pry(main)> echo --help
Echo the input: echo [ARGS]


    -h, --help      Show this message.

NOTE: it is possible to pass the description and command_options as arguments to create_command, just like block_command.

Handling options

The main advantage of class commands is that you can easily support command-line style options or flags. This is done by overriding the options method:

Pry::Commands.create_command "echo" do
  description "Echo the input: echo [ARGS]"

  # opt is a Slop object, see
  def options(opt)
    opt.on :u, :upcase, "Upper-case the arguments before echoing"
    opt.on :d, :downcase, "Lower-case the arguments before echoing"

  def process
    string = args.join(" ")
    string.upcase! if opts.upcase?
    string.downcase! if opts.downcase?

    output.puts string

As you can see there are two methods defined on commands that make it easy to handle options:

  • opts returns the Slop object that parsed the flags.
  • args returns any positional arguments that are left over.

Pry currently uses the Slop parser library internally, so please look at their README for exactly what kind of options you can support.

### Complete example

This section gives a summary of all the things you can do when defining a new Class Command using the echo command from the previous section as an example.

Pry::Commands.create_command "echo" do
  # The description is a short summary (aim for a maximum of about 80 characters).
  # It is displayed after the command's name in the list of commands output by `help`.
  # If you don't set a description, the String "No description." is used.
  description "Echo the input."

  # The banner is a longer summary (aim for several lines).
  # It is displayed above the options when running `command --help` or `help command`
  # Normally the first line is a summary of the usage.
  # If you don't set a banner, the description is used.
  banner <<-BANNER
    Usage: echo [ -u | -d ] [ -j <joiner> ] <words>

    Prints the words to the output, upcasing or downcasing as appropriate, separated
    by the joiner string.

  # The command options are explained in the Command options section, they allow
  # tweaking of the default behaviour of commands. You can normally omit them as the
  # default options will usually work.
    :shellwords => false

  # The {setup} method is called right at the start of command invocation, before
  # {options} is called, and before {process} is called. You can use it to initialize
  # state that the rest of your command will need, please don't override {initialize}
  # itself. Every time setup is called will be on a fresh instance of the Command Class
  # so you don't need to clean up old state left by previous invocations.
  def setup
    @joiner = " "

  # The {options} method is called between {setup} and {process} in order to initialize
  # the {opts} and {args}.
  # It is also called (without {setup} or {process}) when displaying the help message, so
  # it should not have any side-effects.
  def options(opt)
    opt.on :u, :upcase, "Upper-case the arguments before echoing"
    opt.on :d, :downcase, "Lower-case the arguments before echoing"
    opt.on :j, :joiner, "Join arguments with a string (default: ' ')", true do |j|
      @joiner = j

  # The {process} method is where all the magic of your command should happen.
  # It isn't passed any options, but see the section on Context methods for an
  # overview of what it can access in addition to {opts} and {args}
  def process

    # CommandErrors can be raised when the command has gone wrong in an
    # expected way. They render nicely in the Pry REPL.
    raise Pry::CommandError, "-u doesn't make sense with -d" if opts.u? && opts.d?

    string = joined_args
    string.upcase! if opts.upcase?
    string.downcase! if opts.downcase?
    output.puts string

  # You can define your own private helper methods for this command. If you want
  # to share helpers, then define them using Pry::Commands.helpers{ def foo; end }
  # instead.

  def joined_args
### Regexp commands

Both Block Commands and Class Commands can be defined as a Regexp instead of a String. This is most useful if you want to provide fun syntax for commands. If you want to do this, then you should also set the :listing command option to a String so that the output of "help" looks correct.

The capture groups of the regular expression are made available in two places, firstly by calling the captures method, and secondly by being added to the front of the args array.

Pry::Commands.create_command /wtf([?!]*)/ do
  description "Show the exception backtrace, add more ?!s to see more lines."

  command_options :listing => "wtf?!"

  def process
    raise Pry::CommandError, "No most-recent exception" unless _pry_.last_exception
    number_of_lines = [captures[0].size, 0.5].max * 10
    output.puts _pry_.last_exception
    output.puts _pry_.last_exception.backtrace.first(number_of_lines)

This can then be used:

pry(main)> 1 / 0
ZeroDivisionError: divided by 0
pry(main)> wtf?!?
divided by 0
(pry):1:in `/'
### Command options

Command options are parameters that tweak the behaviour of commands, they are provided by calling the command_options method on Class Commands, or as the third parameter when defining Block Commands.

  • :keep_retval (Boolean) — Whether or not to use return value of the block for return of command or just to return void (the default). If your command returns void even if it has :keep_retval set to true, the command will have no return value and will show no eval output (e.g => result)
  • :requires_gem (Array) — Whether this command is an optional command contingent upon any gem dependencies. If it is and dependencies not met then command is disabled and a stub proc giving instructions to install command is provided. Note only use this option if you want your command to be optional; use of this option should be rare.
  • :interpolate (Boolean) — Whether string #{} based interpolation is applied to the command arguments before executing the command. Defaults to true.
  • :use_prefix (Boolean) — Whether or not the command responds to the command prefix config option. Defaults to true.
  • :use_shellwords (Boolean) — Whether the commands arguments are tokenized using Shellwords. Defaults to true. If false, arguments are simply split on spaces instead. Setting it to false is most useful if you want to accept arbitrary ruby as an argument, as you can re-combine the arguments before eval'ing them.
  • :listing (String) — The listing name of the command. That is the name by which the command is looked up by help and by show-source. Necessary for regex based commands.
  • :takes_block (Boolean) - Whether the command can take a block. A block is passed to a command by interposing a | character before providing the Ruby block syntax. e.g my-command arg1 | { puts 'a block!' }. The block is then made available inside the command via the command_block method. This method will return nil if no block was passed or :takes_block was false.
### Context methods

All commands can access a few special methods, in addition to those added to the helper module of their command sets.

Argument parsing

  • opts — The options as parsed by Slop. (Class Commands only)
  • args — The arguments that remained after Slop parsed the opts. (Class Commands only)
  • arg_string — The string that appeared after the command name, or after the end of the regexp match for regexp commands.
  • captures — The captures in the regexp (Regexp Commands only)
  • command_block - The passed block (only available if :takes_block => true and a block was in fact passed).

Pry context

  • _pry_ — The current instance of Pry that's running.
  • target — The binding that is currently being Pry'd. Most useful for target.eval, which lets you eval code as though you were the user.
  • target_self — A short-cut for target.eval('self')
  • eval_string — The cumulative lines of input for multi-line input. Modifying this modifies the input buffer (used by commands such as amend-line)
  • output — The current output object.

Command introspection

  • command_options — The current command's options.
  • command_name — The current command's name.
  • description — The current command's description.
  • block — The block in which the current command was defined.

Interacting with other commands

  • command_set — The command set through which the command was invoked.
  • run — Run another command in the current command set.
  • void — A value indicating that a command returned no value.
### Conventions

When you define your own commands and especially if you distribute them as plugins it is important you follow a few conventions.

Ensure all output goes to the output object

Ensuring that you use output.puts and not just the top-level puts allows all Pry output to be redirected by the user.


Pry::Commands.block_command "greet" do |name|
  output.puts "hello #{name}"

Provide non-color variation

If your command uses colored output please ensure that you provide a non-color variation. Use the Pry.config.color option to determine whether color should be used.

Pry::Commands.block_command "greet" do |text|
  output.puts Pry.config.color ? "\e[1m#{text}\e[0m" : text

Use hyphens not underscores for long command names

Commands are not methods, so to further distinguish them from methods (and avoid conflict) it is recommended to segment long names with hyphens rather than underscores.

Pry::Commands.block_command "long-command-name" do
  output.puts "a long name"

Provide a listing for Regexp commands

Regexp commands need a 'listing' name if they are to be looked up by commands such as help and show-command.


Pry::Commands.block_command /\.(.*)/, "Invoke a shell command", :listing => "shell-command" do |cmd|