options and dependencies need to be inside arrays:

namespace :thing do
  desc "it does a thing"
  task :work, [:option, :foo, :bar] do |task, args|
    puts "work", args
  end

  task :another, [:option, :foo, :bar] do |task, args|
    puts "another #{args}"
    Rake::Task["thing:work"].invoke(args[:option], args[:foo], args[:bar])
    # or splat the args
    # Rake::Task["thing:work"].invoke(*args)
  end

end

Then

rake thing:work[1,2,3]
=> work: {:option=>"1", :foo=>"2", :bar=>"3"}

rake thing:another[1,2,3]
=> another {:option=>"1", :foo=>"2", :bar=>"3"}
=> work: {:option=>"1", :foo=>"2", :bar=>"3"}

NOTE: variable task is the the task object, not very helpful unless you know/care about Rake internals.

RAILS NOTE:

If running the task from rails, its best to preload the environment by adding => [:environment] which is a way to setup dependent tasks.

  task :work, [:option, :foo, :bar] => [:environment] do |task, args|
    puts "work", args
  end

• This worked for m in Aug 2016...updated answer should be updated to this I think! - rmcsharry
• I updated this to be accepted as the answer by request. It however doesn't demonstrate how to pass argument to another rake task from a task itself. - Tilendor
• Also, make sure you don't use spaces between the arguments. E.g don't do this: rake thing:work[1, 2, 3] as it won't work and you'll get an error Don't know how to build task - rpbaltazar
• Also, make sure you enclose the argument in string. e.g from your command line run the rake task like so rake thing:work'[1,2,3]' - Damian Simon Peter
• Unfortuanely, zsh can not parse the call correctly, you need type the command on zsh like this: rake thing:work\[1,2,3\], or this rake 'thing:work[1,2,3]' - hutusi

You can specify formal arguments in rake by adding symbol arguments to the task call. For example:

require 'rake'

task :my_task, [:arg1, :arg2] do |t, args|
  puts "Args were: #{args}"
end

task :invoke_my_task do
  Rake.application.invoke_task("my_task[1, 2]")
end

# or if you prefer this syntax...
task :invoke_my_task_2 do
  Rake::Task[:my_task].invoke(3, 4)
end

# a task with prerequisites passes its 
# arguments to it prerequisites
task :with_prerequisite, [:arg1, :arg2] => :my_task #<- name of prerequisite task

# to specify default values, 
# we take advantage of args being a Rake::TaskArguments object
task :with_defaults, :arg1, :arg2 do |t, args|
  args.with_defaults(:arg1 => :default_1, :arg2 => :default_2)
  puts "Args with defaults were: #{args}"
end

Then, from the command line:

> rake my_task[1,2]
Args were: {:arg1=>"1", :arg2=>"2"}

> rake "my_task[1, 2]"
Args were: {:arg1=>"1", :arg2=>"2"}

> rake invoke_my_task
Args were: {:arg1=>"1", :arg2=>"2"}

> rake invoke_my_task_2
Args were: {:arg1=>3, :arg2=>4}

> rake with_prerequisite[5,6]
Args were: {:arg1=>"5", :arg2=>"6"}

> rake with_defaults
Args with defaults were: {:arg1=>:default_1, :arg2=>:default_2}

> rake with_defaults['x','y']
Args with defaults were: {:arg1=>"x", :arg2=>"y"}

As demonstrated in the second example, if you want to use spaces, the quotes around the target name are necessary to keep the shell from splitting up the arguments at the space.

Looking at the code in rake.rb, it appears that rake does not parse task strings to extract arguments for prerequisites, so you can't do task :t1 => "dep[1,2]". The only way to specify different arguments for a prerequisite would be to invoke it explicitly within the dependent task action, as in :invoke_my_task and :invoke_my_task_2.

Note that some shells (like zsh) require you to escape the brackets: rake my_task\['arg1'\]

• To invoke a task within a namespace simpy do: Rake::Task['namespace:task'].invoke - gaqzi
• That's a separate question, Igoru, but the reason your call to invoke only runs once is that rake is dependency-oriented, so it will only execute a task if it is needed. For generic tasks that means if it hasn't already run. To explicitly execute a task regardless of its dependencies or if it is needed, call execute instead of invoke. - Nick Desjardins
• Note: According to rake, this syntax for accepting variables in tasks is deprecated: WARNING: 'task :t, arg, :needs => [deps]' is deprecated. Please use 'task :t, [args] => [deps]' instead. - Ajedi32
• Note that zsh fails to parse the command line arguments correctly (zsh: no matches found: ...), so you need to escape the brackets: rake my_task\['arg1'\]. From robots.thoughtbot.com/post/18129303042/… - Seth Bro
• @SethBro YES. If only your comment hadn't been hidden behind the "See more comments" link I wouldn't have wasted 10 minutes unable to make this work. - GMA

In addition to answer by kch (I didn't find how to leave a comment to that, sorry):

You don't have to specify variables as ENV variables before the rake command. You can just set them as usual command line parameters like that:

rake mytask var=foo

and access those from your rake file as ENV variables like such:

p ENV['var'] # => "foo"

• It's really convenient. Thanks! - Hoang Le
• Ah, so this is how Rails does it. Makes sense... - Ajedi32

If you want to pass named arguments (e.g. with standard OptionParser) you could use something like this:

$ rake user:create -- --user test@example.com --pass 123

note the --, that's necessary for bypassing standard Rake arguments. Should work with Rake 0.9.x, <= 10.3.x.

Newer Rake has changed its parsing of --, and now you have to make sure it's not passed to the OptionParser#parse method, for example with parser.parse!(ARGV[2..-1])

require 'rake'
require 'optparse'
# Rake task for creating an account

namespace :user do |args|
  desc 'Creates user account with given credentials: rake user:create'
  # environment is required to have access to Rails models
  task :create do
    options = {}
    OptionParser.new(args) do |opts|
      opts.banner = "Usage: rake user:create [options]"
      opts.on("-u", "--user {username}","User's email address", String) do |user|
        options[:user] = user
      end
      opts.on("-p", "--pass {password}","User's password", String) do |pass|
        options[:pass] = pass
      end
    end.parse!

    puts "creating user account..."
    u = Hash.new
    u[:email] = options[:user]
    u[:password] = options[:pass]
    # with some DB layer like ActiveRecord:
    # user = User.new(u); user.save!
    puts "user: " + u.to_s
    puts "account created."
    exit 0
  end
end

exit at the end will make sure that the extra arguments won't be interpreted as Rake task.

Also the shortcut for arguments should work:

 rake user:create -- -u test@example.com -p 123

When rake scripts look like this, maybe it's time to look for another tool that would allow this just out of box.

• From my perspective this really is the best answer. Bypass environment variable kludges, strange syntax with task arguments, the additional benefit for standard --option-names. My only suggestion would be to use exit rather than abort as abort will leave you with a return code of 1 to the shell. If the rake task is a part of a higher-level script it's more common to assume a non-zero exit is some type of error. - Joe
• I agree with Joe, this is the best answer. The natural thing is to use the same interface for passing options to rake as you would when passing options to a script. - Rik Smith-Unna
• I agree this is the best answer. Ain't there a way to bypass the ugly --? Like passing rake arguments to the actual task or something? Like task :my_task, :*args do |t, args| or something? - Augustin Riedinger
• Besides, I don't understand what the {username} is here for. Where is it used? Why isn't it there in -u {username}? Cheers - Augustin Riedinger
• The way how Rake parses ARGV was changed in 10.4.1 and reverted in 10.4.2. github.com/ruby/rake/commit/… - Tombart

I've found the answer from these two websites: Net Maniac and Aimred.

You need to have version > 0.8 of rake to use this technique

The normal rake task description is this:

desc 'Task Description'
task :task_name => [:depends_on_taskA, :depends_on_taskB] do
  #interesting things
end

To pass arguments, do three things:

1. Add the argument names after the task name, separated by commas.
2. Put the dependencies at the end using :needs => [...]
3. Place |t, args| after the do. (t is the object for this task)

To access the arguments in the script, use args.arg_name

desc 'Takes arguments task'
task :task_name, :display_value, :display_times, :needs => [:depends_on_taskA, :depends_on_taskB] do |t, args|
  args.display_times.to_i.times do
    puts args.display_value
  end
end

To call this task from the command line, pass it the arguments in []s

rake task_name['Hello',4]

will output

Hello
Hello
Hello
Hello

and if you want to call this task from another task, and pass it arguments, use invoke

task :caller do
  puts 'In Caller'
  Rake::Task[:task_name].invoke('hi',2)
end

then the command

rake caller

will output

In Caller
hi
hi

I haven't found a way to pass arguments as part of a dependency, as the following code breaks:

task :caller => :task_name['hi',2]' do
   puts 'In Caller'
end

• The format for this functionality has changed as this warning states: 'task :t, arg, :needs => [deps]' is deprecated. Please use 'task :t, [args] => [deps]' instead. - madh

Another commonly used option is to pass environment variables. In your code you read them via ENV['VAR'], and can pass them right before the rake command, like

$ VAR=foo rake mytask

• Frankly I was hoping for rake task -- these --go --to -a program and my task could get them from ARGV. Unfortunately I'm not sure if that's possible however I am currently using your solution: rake var1=val1 var2=val2 - JasonSmith
• @jhs: rake blah -- --these --go --to --a-program (note the -- to tell rake that its switches have ended), see stackoverflow.com/questions/5086224/… - mu is too short

Actually @Nick Desjardins answered perfect. But just for education: you can use dirty approach: using ENV argument

task :my_task do
  myvar = ENV['myvar']
  puts "myvar: #{myvar}"
end 

rake my_task myvar=10
#=> myvar: 10

I couldn't figure out how to pass args and also the :environment until I worked this out:

namespace :db do
  desc 'Export product data'
  task :export, [:file_token, :file_path] => :environment do |t, args|
    args.with_defaults(:file_token => "products", :file_path => "./lib/data/")

       #do stuff [...]

  end
end

And then I call like this:

rake db:export['foo, /tmp/']

desc 'an updated version'
task :task_name, [:arg1, :arg2] => [:dependency1, :dependency2] do |t, args|
    puts args[:arg1]
end

• To call this, go: rake task_name[hello, world] - Dex
• from rake.rubyforge.org/files/doc/rakefile_rdoc.html "Just a few words of caution. The rake task name and its arguments need to be a single command line argument to rake. This generally means no spaces. If spaces are needed, then the entire rake + argument string should be quoted. Something like this: rake "name[billy bob, smith]" " - Gayle

I just wanted to be able to run:

$ rake some:task arg1 arg2

Simple, right? (Nope!)

Rake interprets arg1 and arg2 as tasks, and tries to run them. So we just abort before it does.

namespace :some do
  task task: :environment do
    arg1, arg2 = ARGV

    # your task...

    exit
  end
end

Take that, brackets!

Disclaimer: I wanted to be able to do this in a pretty small pet project. Not intended for "real world" usage since you lose the ability to chain rake tasks (i.e. rake task1 task2 task3). IMO not worth it. Just use the ugly rake task[arg1,arg2].

• Needed to make this _, arg1, arg2 = ARGV as the first arg was seen to be the name of the rake task. But that exit is a neat trick. - fatty
• rake task[arg1,arg2] && rake task2 && rake task3 Not sure if that's less ugly than rake task[arg1,arg2] task2 task3. Probably less efficient though. - Nuclearman
• _, *args = ARGV is perfect for capturing all subsequent arguments! Thanks heaps! - XtraSimplicity

I use a regular ruby argument in the rake file:

DB = ARGV[1]

then I stub out the rake tasks at the bottom of the file (since rake will look for a task based on that argument name).

task :database_name1
task :database_name2

command line:

rake mytask db_name

this feels cleaner to me than the var=foo ENV var and the task args[blah, blah2] solutions.
the stub is a little jenky, but not too bad if you just have a few environments that are a one-time setup

• To prevent frozen strings issues, use dup at the end: db = ARGV[1].dup - Juanda
• Event better db = ARGV[1].dup unless ARGV[1].nil? to prevent exception of duping a nil. - Andre Figueiredo

The ways to pass argument are correct in above answer. However to run rake task with arguments, there is a small technicality involved in newer version of rails

It will work with rake "namespace:taskname['argument1']"

Note the Inverted quotes in running the task from command line.

I like the "querystring" syntax for argument passing, especially when there are a lot of arguments to be passed.

Example:

rake "mytask[width=10&height=20]"

The "querystring" being:

width=10&height=20

Warning: note that the syntax is rake "mytask[foo=bar]" and NOT rake mytask["foo=bar"]

When parsed inside the rake task using Rack::Utils.parse_nested_query , we get a Hash:

=> {"width"=>"10", "height"=>"20"}

(The cool thing is that you can pass hashes and arrays, more below)

This is how to achieve this:

require 'rack/utils'

task :mytask, :args_expr do |t,args|
  args.with_defaults(:args_expr => "width=10&height=10")
  options = Rack::Utils.parse_nested_query(args[:args_expr])
end

Here's a more extended example that I'm using with Rails in my delayed_job_active_record_threaded gem:

bundle exec rake "dj:start[ebooks[workers_number]=16&ebooks[worker_timeout]=60&albums[workers_number]=32&albums[worker_timeout]=120]"

Parsed the same way as above, with an environment dependency (in order load the Rails environment)

namespace :dj do
  task :start, [ :args_expr ] => :environment do |t, args|
    # defaults here...
    options = Rack::Utils.parse_nested_query(args[:args_expr])  
  end
end

Gives the following in options

=> {"ebooks"=>{"workers_number"=>"16", "worker_timeout"=>"60"}, "albums"=>{"workers_number"=>"32", "worker_timeout"=>"120"}}

• Downvoter, care to elaborate? - Abdo

To pass arguments to the default task, you can do something like this. For example, say "version" is your argument:

task :default, [:version] => [:build]

task :build, :version do |t,args|
  version = args[:version]
  puts version ? "version is #{version}" : "no version passed"
end

Then you can call it like so:

$ rake
no version passed

or

$ rake default[3.2.1]
version is 3.2.1

or

$ rake build[3.2.1]
version is 3.2.1

However, I have not found a way to avoid specifying the task name (default or build) while passing in arguments. Would love to hear if anyone knows of a way.

Most of the methods described above did not work for me, maybe they are deprecated in the newer versions. The up-to-date guide can be found here: http://guides.rubyonrails.org/command_line.html#custom-rake-tasks

a copy-and-paste ans from the guide is here:

task :task_name, [:arg_1] => [:pre_1, :pre_2] do |t, args|
  # You can use args from here
end

Invoke it like this

bin/rake "task_name[value 1]" # entire argument string should be quoted

If you can't be bothered to remember what argument position is for what and you want do something like a ruby argument hash. You can use one argument to pass in a string and then regex that string into an options hash.

namespace :dummy_data do
  desc "Tests options hash like arguments"
  task :test, [:options] => :environment do |t, args|
    arg_options = args[:options] || '' # nil catch incase no options are provided
    two_d_array = arg_options.scan(/\W*(\w*): (\w*)\W*/)
    puts two_d_array.to_s + ' # options are regexed into a 2d array'
    string_key_hash = two_d_array.to_h
    puts string_key_hash.to_s + ' # options are in a hash with keys as strings'
    options = two_d_array.map {|p| [p[0].to_sym, p[1]]}.to_h
    puts options.to_s + ' # options are in a hash with symbols'
    default_options = {users: '50', friends: '25', colour: 'red', name: 'tom'}
    options = default_options.merge(options)
    puts options.to_s + ' # default option values are merged into options'
  end
end

And on the command line you get.

$ rake dummy_data:test["users: 100 friends: 50 colour: red"]
[["users", "100"], ["friends", "50"], ["colour", "red"]] # options are regexed into a 2d array
{"users"=>"100", "friends"=>"50", "colour"=>"red"} # options are in a hash with keys as strings
{:users=>"100", :friends=>"50", :colour=>"red"} # options are in a hash with symbols
{:users=>"100", :friends=>"50", :colour=>"red", :name=>"tom"} # default option values are merged into options

• Your code needs a few well-placed empty lines. I don't know how you read that wall of text. - Joshua Pinter

While passing parameters, it is better option is an input file, can this be a excel a json or whatever you need and from there read the data structure and variables you need from that including the variable name as is the need. To read a file can have the following structure.

  namespace :name_sapace_task do
    desc "Description task...."
      task :name_task  => :environment do
        data =  ActiveSupport::JSON.decode(File.read(Rails.root+"public/file.json")) if defined?(data)
    # and work whit yoour data, example is data["user_id"]

    end
  end

Example json

{
  "name_task": "I'm a task",
  "user_id": 389,
  "users_assigned": [389,672,524],
  "task_id": 3
}

Execution

rake :name_task 

• If you need a JSON instructions file for your Rake task, you're probably doing too many things in your Rake task. - ZiggyTheHamster
• This is way over-complicating something that's incredibly simple. - jeffdill2