Run external commands with pretty output logging and capture stdout, stderr and exit status. Redirect stdin, stdout and stderr of each command to a file or a string.
TTY::Command provides independent command execution component for TTY toolkit.
Complex software projects aren't just a single app. These projects usually spawn dozens or hundreds of supplementary standalone scripts which are just as important as the app itself. Examples include - data validation, deployment, monitoring, database maintenance, backup & restore, configuration management, crawling, ETL, analytics, log file processing, custom reports, etc. One of the contributors to TTY::Command counted 222 scripts in the bin
directory for his startup.
Why should we be handcuffed to sh
or bash
for these scripts when we could be using Ruby? Ruby is easier to write and more fun, and we gain a lot by using a better language. It's nice for everyone to just use Ruby everywhere.
TTY::Command tries to add value in other ways. It'll halt automatically if a command fails. It's easy to get verbose or quiet output as appropriate, or even capture output and parse it with Ruby. Escaping arguments is a breeze. These are all areas where traditional shell scripts tend to fall flat.
Add this line to your application's Gemfile:
gem "tty-command"
And then execute:
$ bundle
Or install it yourself as:
$ gem install tty-command
- 1. Usage
- 2. Interface
- 3. Advanced Interface
- 4. Example
Create a command instance and then run some commands:
require "tty-command"
cmd = TTY::Command.new
cmd.run("ls -la")
cmd.run("echo Hello!")
Note that run
will throw an exception if the command fails. This is already an improvement over ordinary shell scripts, which just keep on going when things go bad. That usually makes things worse.
You can use the return value to capture stdout and stderr:
out, err = cmd.run("cat ~/.bashrc | grep alias")
Instead of using a plain old string, you can break up the arguments and they'll get escaped if necessary:
path = "hello world"
FileUtils.touch(path)
cmd.run("sum #{path}") # this will fail due to bad escaping
cmd.run("sum", path) # this gets escaped automatically
Run starts the specified command and waits for it to complete.
The argument signature of run
is as follows:
run([env], command, [argv1, ...], [options])
The env
, command
and options
arguments are described in the following sections.
For example, to display file contents:
cmd.run("cat file.txt")
If the command succeeds, a TTY::Command::Result
is returned that records stdout and stderr:
out, err = cmd.run("date")
puts "The date is #{out}"
# => "The date is Tue 10 May 2016 22:30:15 BST\n"
You can also pass a block that gets invoked anytime stdout and/or stderr receive output:
cmd.run("long running script") do |out, err|
output << out if out
errors << err if err
end
If the command fails (with a non-zero exit code), a TTY::Command::ExitError
is raised. The ExitError
message will include:
- the name of command executed
- the exit status
- stdout bytes
- stderr bytes
If the error output is very long, the stderr may contain only a prefix, number of omitted bytes and suffix.
If you expect a command to fail occasionally, use run!
instead. Then you can detect failures and respond appropriately. For example:
if cmd.run!("which xyzzy").failure?
cmd.run("brew install xyzzy")
end
By default, when a command is run, the command and the output are printed to stdout
using the :pretty
printer. If you wish to change printer you can do so by passing a :printer
option:
:null
- no output:pretty
- colorful output:progress
- minimal output with green dot for success and F for failure:quiet
- only output actual command stdout and stderr
like so:
cmd = TTY::Command.new(printer: :progress)
By default the printers log to stdout
but this can be changed by passing an object that responds to <<
message:
logger = Logger.new("dev.log")
cmd = TTY::Command.new(output: logger)
You can force the printer to always in print in color by passing the :color
option:
cmd = TTY::Command.new(color: true)
If the default printers don't meet your needs you can always create a custom printer
When using printers you can switch off coloring by using :color
option set to false
.
By default, when logging is enabled and pretty
printer is used, each log entry is prefixed by specific command run uuid number. This number can be switched off using the :uuid
option at initialization:
cmd = TTY::Command.new(uuid: false)
cmd.run("rm -R all_my_files")
# =>
# Running rm -r all_my_files
# ...
# Finished in 6 seconds with exit status 0 (successful)
or individually per command run:
cmd = TTY::Command.new
cmd.run("echo hello", uuid: false)
# =>
# Running echo hello
# hello
# Finished in 0.003 seconds with exit status 0 (successful)
When using a command that can fail, setting :only_output_on_error
option to true
hides the output if the command succeeds:
cmd = TTY::Command.new
cmd.run("non_failing_command", only_output_on_error: true)
This will only print the Running
and Finished
lines, while:
cmd.run("non_failing_command")
will also print any output that the non_failing_command
might generate.
Running either:
cmd.run("failing_command", only_output_on_error: true)
either:
cmd.run("failing_command")
will also print the output.
Setting this option will cause the output to show at once, at the end of the command.
By default commands will produce warnings when, for example pty
option is not supported on a given platform. You can switch off such warnings with :verbose
option set to false
.
cmd.run("echo '\e[32mColors!\e[0m'", pty: true, verbose: false)
Sometimes it can be useful to put your script into a "dry run" mode that prints commands without actually running them. To simulate execution of the command use the :dry_run
option:
cmd = TTY::Command.new(dry_run: true)
cmd.run(:rm, "all_my_files")
# => [123abc] (dry run) rm all_my_files
To check what mode the command is in use the dry_run?
query helper:
cmd.dry_run? # => true
If you need to wait for a long running script and stop it when a given pattern has been matched use wait
like so:
cmd.wait "tail -f /var/log/production.log", /something happened/
To simulate classic bash test command you case use test
method with expression to check as a first argument:
if cmd.test "-e /etc/passwd"
puts "Sweet..."
else
puts "Ohh no! Where is it?"
exit 1
end
In order to run a command with Ruby interpreter do:
cmd.ruby %q{-e "puts 'Hello world'"}
The environment variables need to be provided as hash entries, that can be set directly as a first argument:
cmd.run({"RAILS_ENV" => "PRODUCTION"}, :rails, "server")
or as an option with :env
key:
cmd.run(:rails, "server", env: {rails_env: :production})
When a value in env is nil, the variable is unset in the child process:
cmd.run(:echo, "hello", env: {foo: "bar", baz: nil})
When a hash is given in the last argument (options), it allows to specify a current directory, umask, user, group and zero or more fd redirects for the child process.
There are few ways you can redirect commands output.
You can directly use shell redirection like so:
out, err = cmd.run("ls 1&>2")
puts err
# =>
# CHANGELOG.md
# CODE_OF_CONDUCT.md
# Gemfile
# ...
You can provide redirection as additional hash options where the key is one of :in
, :out
, :err
, an integer (a file descriptor for the child process), an IO or array. For example, stderr
can be merged into stdout as follows:
cmd.run(:ls, :err => :out)
cmd.run(:ls, :stderr => :stdout)
cmd.run(:ls, 2 => 1)
cmd.run(:ls, STDERR => :out)
cmd.run(:ls, STDERR => STDOUT)
The hash key and value specify a file descriptor in the child process (stderr & stdout in the examples).
You can also redirect to a file:
cmd.run(:cat, :in => "file")
cmd.run(:cat, :in => open("/etc/passwd"))
cmd.run(:ls, :out => "log")
cmd.run(:ls, :out => "/dev/null")
cmd.run(:ls, :out => "out.log", :err => "err.log")
cmd.run(:ls, [:out, :err] => "log")
cmd.run("ls 1>&2", :err => "log")
It is possible to specify flags and permissions of file creation explicitly by passing an array value:
cmd.run(:ls, :out => ["log", "w"]) # 0664 assumed
cmd.run(:ls, :out => ["log", "w", 0600])
cmd.run(:ls, :out => ["log", File::WRONLY|File::EXCL|File::CREAT, 0600])
You can, for example, read data from one source and output to another:
cmd.run("cat", :in => "Gemfile", :out => "gemfile.log")
You can provide input to stdin stream using the :input
key. For instance, given the following executable called cli
that expects name from stdin
:
name = $stdin.gets
puts "Your name: #{name}"
In order to execute cli
with name input do:
cmd.run("cli", input: "Piotr\n")
# => Your name: Piotr
Alternatively, you can pass input via the :in option, by passing a StringIO
Object. This object might have more than one line, if the executed command reads more than once from STDIN.
Assume you have run a program, that first asks for your email address and then for a password:
in_stream = StringIO.new
in_stream.puts "[email protected]"
in_stream.puts "password"
in_stream.rewind
cmd.run("my_cli_program", "login", in: in_stream).out
You can timeout command execution by providing the :timeout
option in seconds:
cmd.run("while test 1; sleep 1; done", timeout: 5)
And to set it for all commands do:
cmd = TTY::Command.new(timeout: 5)
Please run examples/timeout.rb
to see timeout in action.
By default the standard input, output and error are non-binary. However, you can change to read and write in binary mode by using the :binmode
option like so:
cmd.run("echo 'hello'", binmode: true)
To set all commands to be run in binary mode do:
cmd = TTY::Command.new(binmode: true)
You can specify process termination signal other than the default SIGTERM
:
cmd.run("whilte test1; sleep1; done", timeout: 5, signal: :KILL)
The :pty
configuration option causes the command to be executed in subprocess where each stream is a pseudo terminal
. By default this options is set to false
.
If you require to interface with interactive subprocess then setting this option to true
will enable a pty
terminal device. For example, a command may emit colored output only if it is running via terminal device. You may also wish to run a program that waits for user input, and simulates typing in commands and reading responses.
This option will only work on systems that support BSD pty devices such as Linux or OS X, and it will gracefully fallback to non-pty device on all the other.
In order to run command in pseudo terminal
, either set the flag globally for all commands:
cmd = TTY::Command.new(pty: true)
or individually for each executed command:
cmd.run("echo 'hello'", pty: true)
Please note that setting :pty
to true
may change how the command behaves. It's important to understand the difference between interactive
and non-interactive
modes. For example, executing git log
to view the commit history in default non-interactive
mode:
cmd.run("git log") # => finishes and produces full output
However, in interactive
mode with pty
flag on:
cmd.run("git log", pty: true) # => uses pager and waits for user input (never returns)
In addition, when pty device is used, any input to command may be echoed to the standard output, as well as some redirects may not work.
To change directory in which the command is run pass the :chdir
option:
cmd.run(:echo, "hello", chdir: "/var/tmp")
To run command as a given user do:
cmd.run(:echo, "hello", user: "piotr")
To run command as part of group do:
cmd.run(:echo, "hello", group: "devs")
To run command with umask do:
cmd.run(:echo, "hello", umask: "007")
Each time you run command the stdout and stderr are captured and return as result. The result can be examined directly by casting it to tuple:
out, err = cmd.run(:echo, "Hello")
However, if you want to you can defer reading:
result = cmd.run(:echo, "Hello")
result.out
result.err
To check if command exited successfully use success?
:
result = cmd.run(:echo, "Hello")
result.success? # => true
To check if command exited unsuccessfully use failure?
or failed?
:
result = cmd.run(:echo, "Hello")
result.failure? # => false
result.failed? # => false
To check if command ran to completion use exited?
or complete?
:
result = cmd.run(:echo, "Hello")
result.exited? # => true
result.complete? # => true
The result itself is an enumerable and allows you to iterate over the stdout output:
result = cmd.run(:ls, "-1")
result.each { |line| puts line }
# =>
# CHANGELOG.md
# CODE_OF_CONDUCT.md
# Gemfile
# Gemfile.lock
# ...
# lib
# pkg
# spec
# tasks
By default the linefeed character \n
is used as a delimiter but this can be changed either globally by calling record_separator
:
TTY::Command.record_separator = "\n\r"
or configured per each
call by passing delimiter as an argument:
cmd.run(:ls, "-1").each("\t") { ... }
If the built-in printers do not meet your requirements you can create your own. A printer is a regular Ruby class that can be registered through :printer
option to receive notifications about received command data.
As the command runs the custom printer will be notified when the command starts, when data is printed to stdout, when data is printed to stderr and when the command exits.
Please see lib/tty/command/printers/abstract.rb for a full set of methods that you can override.
At the very minimum you need to specify the write
method that will be called during the lifecycle of command execution. The write
accepts two arguments, first the currently run command instance and second the message to be printed:
CustomPrinter < TTY::Command::Printers::Abstract
def write(cmd, message)
puts cmd.to_command + message
end
end
cmd = TTY::Command.new(printer: CustomPrinter)
Here's a slightly more elaborate example to illustrate how tty-command can improve on plain old shell scripts. This example installs a new version of Ruby on an Ubuntu machine.
cmd = TTY::Command.new
# dependencies
cmd.run "apt-get -y install build-essential checkinstall"
# fetch ruby if necessary
if !File.exists?("ruby-2.3.0.tar.gz")
puts "Downloading..."
cmd.run "wget http://ftp.ruby-lang.org/pub/ruby/2.3/ruby-2.3.0.tar.gz"
cmd.run "tar xvzf ruby-2.3.0.tar.gz"
end
# now install
Dir.chdir("ruby-2.3.0") do
puts "Building..."
cmd.run "./configure --prefix=/usr/local"
cmd.run "make"
end
After checking out the repo, run bin/setup
to install dependencies. Then, run rake spec
to run the tests. You can also run bin/console
for an interactive prompt that will allow you to experiment.
Bug reports and pull requests are welcome on GitHub at https://github.com/piotrmurach/tty-command. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the Contributor Covenant code of conduct.
The gem is available as open source under the terms of the MIT License.
Copyright (c) 2016 Piotr Murach. See LICENSE for further details.