The Amazing Mustermann

Make sure you view the correct docs: latest release, master.

Welcome to Mustermann. Mustermann is your personal string matching expert. As an expert in the field of strings and patterns, Mustermann keeps its runtime dependencies to a minimum and is fully covered with specs and documentation.

Given a string pattern, Mustermann will turn it into an object that behaves like a regular expression and has comparable performance characteristics.

if '/foo/bar' =~ Mustermann.new('/foo/*')
  puts 'it works!'
end

case 'something.png'
when Mustermann.new('foo/*') then puts "prefixed with foo"
when Mustermann.new('*.pdf') then puts "it's a PDF"
when Mustermann.new('*.png') then puts "it's an image"
end

pattern = Mustermann.new('/:prefix/*.*')
pattern.params('/a/b.c') # => { "prefix" => "a", splat => ["b", "c"] }

Overview

Features

  • Pattern Types: Mustermann supports a wide variety of different pattern types, making it compatible with a large variety of existing software.
  • Fine Grained Control: You can easily adjust matching behavior and add constraints to the placeholders and capture groups.
  • Binary Operators: Patterns can be combined into composite patterns using binary operators.
  • Regexp Look Alike: Mustermann patterns can be used as a replacement for regular expressions.
  • Parameter Parsing: Mustermann can parse matched parameters into a Sinatra-style "params" hash, including type casting.
  • Peeking: Lets you check if the beginning of a string matches a pattern.
  • Expanding: Besides parsing a parameters from an input string, a pattern object can also be used to generate a string from a set of parameters.
  • Generating Templates: This comes in handy when wanting to hand on patterns rather than fully expanded strings as part of an external API.
  • Proc Look Alike: Pass on a pattern instead of a block.
  • Duck Typing: You can create your own pattern-like objects by implementing to_pattern.
  • Performance: Patterns are implemented with both performance and a low memory footprint in mind.

Additional Tooling

These features are included in the library, but not loaded by default

  • Mapper: A simple tool for mapping one string to another based on patterns.
  • Routers: Model execution flow based on pattern matching. Comes with a simple Rack router.
  • Sinatra Integration: Mustermann can be used as a Sinatra extension. Sinatra 2.0 and beyond will use Mustermann by default.

Pattern Types

Mustermann support multiple pattern types. A pattern type defines the syntax, matching semantics and whether certain features, like expanding and generating templates, are available.

You can create a pattern of a certain type by passing type option to Mustermann.new:

require 'mustermann'
pattern = Mustermann.new('/*/**', type: :shell)

Note that this will use the type as suggestion: When passing in a string argument, it will create a pattern of the given type, but it might choose a different type for other objects (a regular expression argument will always result in a regexp pattern, a symbol always in a sinatra pattern, etc).

Alternatively, you can also load and instantiate the pattern type directly:

require 'mustermann/shell'
pattern = Mustermann::Shell.new('/*/**')

Mustermann itself includes the sinatra, identity and regexp pattern types. Other pattern types are available as separate gems.

Binary Operators

Patterns can be combined via binary operators. These are:

  • | (or): Resulting pattern matches if at least one of the input pattern matches.
  • & (and): Resulting pattern matches if all input patterns match.
  • ^ (xor): Resulting pattern matches if exactly one of the input pattern matches.
require 'mustermann'

first  = Mustermann.new('/foo/:input')
second = Mustermann.new('/:input/bar')

first | second === "/foo/foo" # => true
first | second === "/foo/bar" # => true

first & second === "/foo/foo" # => false
first & second === "/foo/bar" # => true

first ^ second === "/foo/foo" # => true
first ^ second === "/foo/bar" # => false

These resulting objects are fully functional pattern objects, allowing you to call methods like params or to_proc on them. Moreover, or patterns created solely from expandable patterns will also be expandable. The same logic also applies to generating templates from or patterns.

Regexp Look Alike

Pattern objects mimic Ruby's Regexp class by implementing match, =~, ===, names and named_captures.

require 'mustermann'

pattern = Mustermann.new('/:page')
pattern.match('/')     # => nil
pattern.match('/home') # => #<MatchData "/home" page:"home">
pattern =~ '/home'     # => 0
pattern === '/home'    # => true (this allows using it in case statements)
pattern.names          # => ['page']
pattern.names          # => {"page"=>[1]}

pattern = Mustermann.new('/home', type: :identity)
pattern.match('/')     # => nil
pattern.match('/home') # => #<Mustermann::SimpleMatch "/home">
pattern =~ '/home'     # => 0
pattern === '/home'    # => true (this allows using it in case statements)
pattern.names          # => []
pattern.names          # => {}

Moreover, patterns based on regular expressions (all but identity and shell) automatically convert to regular expressions when needed:

require 'mustermann'

pattern = Mustermann.new('/:page')
union   = Regexp.union(pattern, /^$/)

union =~ "/foo" # => 0
union =~ ""     # => 0

Regexp.try_convert(pattern) # => /.../

This way, unless some code explicitly checks the class for a regular expression, you should be able to pass in a pattern object instead even if the code in question was not written with Mustermann in mind.

Parameter Parsing

Besides being a Regexp look-alike, Mustermann also adds a params method, that will give you a Sinatra-style hash:

require 'mustermann'

pattern = Mustermann.new('/:prefix/*.*')
pattern.params('/a/b.c') # => { "prefix" => "a", splat => ["b", "c"] }

For patterns with typed captures, it will also automatically convert them:

require 'mustermann'

pattern = Mustermann.new('/<prefix>/<int:id>', type: :flask)
pattern.params('/page/10') # => { "prefix" => "page", "id" => 10 }

Peeking

Peeking gives the option to match a pattern against the beginning of a string rather the full string. Patterns come with four methods for peeking:

  • peek returns the matching substring.
  • peek_size returns the number of characters matching.
  • peek_match will return a MatchData or Mustermann::SimpleMatch (just like match does for the full string)
  • peek_params will return the params hash parsed from the substring and the number of characters.

All of the above will turn nil if there was no match.

require 'mustermann'

pattern = Mustermann.new('/:prefix')
pattern.peek('/foo/bar')      # => '/foo'
pattern.peek_size('/foo/bar') # => 4

path_info    = '/foo/bar'
params, size = patter.peek_params(path_info)  # params == { "prefix" => "foo" }
rest         = path_info[size..-1]            # => "/bar"

Expanding

Similarly to parsing, it is also possible to generate a string from a pattern by expanding it with a hash. For simple expansions, you can use Pattern#expand.

pattern = Mustermann.new('/:file(.:ext)?')
pattern.expand(file: 'pony')             # => "/pony"
pattern.expand(file: 'pony', ext: 'jpg') # => "/pony.jpg"
pattern.expand(ext: 'jpg')               # raises Mustermann::ExpandError

Expanding can be useful for instance when implementing link helpers.

Expander Objects

To get fine-grained control over expansion, you can use Mustermann::Expander directly.

You can create an expander object directly from a string:

require 'mustermann/expander'
expander = Mustermann::Expander("/:file.jpg")
expander.expand(file: 'pony') # => "/pony.jpg"

expander = Mustermann::Expander(":file(.:ext)", type: :rails)
expander.expand(file: 'pony', ext: 'jpg') # => "/pony.jpg"

Or you can pass it a pattern instance:

require 'mustermann'
pattern = Mustermann.new("/:file")

require 'mustermann/expander'
expander = Mustermann::Expander.new(pattern)

Expanding Multiple Patterns

You can add patterns to an expander object via <<:

require 'mustermann'

expander = Mustermann::Expander.new
expander << "/users/:user_id"
expander << "/pages/:page_id"

expander.expand(user_id: 15) # => "/users/15"
expander.expand(page_id: 58) # => "/pages/58"

You can set pattern options when creating the expander:

require 'mustermann'

expander = Mustermann::Expander.new(type: :template)
expander << "/users/{user_id}"
expander << "/pages/{page_id}"

Additionally, it is possible to combine patterns of different types:

require 'mustermann'

expander = Mustermann::Expander.new
expander << Mustermann.new("/users/{user_id}", type: :template)
expander << Mustermann.new("/pages/:page_id",  type: :rails)

Handling Additional Values

The handling of additional values passed in to expand can be changed by setting the additional_values option:

require 'mustermann'

expander = Mustermann::Expander.new("/:slug", additional_values: :raise)
expander.expand(slug: "foo", value: "bar") # raises Mustermann::ExpandError

expander = Mustermann::Expander.new("/:slug", additional_values: :ignore)
expander.expand(slug: "foo", value: "bar") # => "/foo"

expander = Mustermann::Expander.new("/:slug", additional_values: :append)
expander.expand(slug: "foo", value: "bar") # => "/foo?value=bar"

It is also possible to pass this directly to the expand call:

require 'mustermann'

pattern = Mustermann.new('/:slug')
pattern.expand(:append, slug: "foo", value: "bar") # => "/foo?value=bar"

Generating Templates

... TODO ...

Proc Look Alike

Patterns implement to_proc:

require 'mustermann'
pattern  = Mustermann.new('/foo')
callback = pattern.to_proc # => #<Proc>

callback.call('/foo') # => true
callback.call('/bar') # => false

They can therefore be easily passed to methods expecting a block:

require 'mustermann'

list    = ["foo", "[email protected]", "bar"]
pattern = Mustermann.new(":name@:domain.:tld")
email   = list.detect(&pattern) # => "[email protected]"

Mapper

You can use a mapper to transform strings according to two or more mappings:

require 'mustermann/mapper'

mapper = Mustermann::Mapper.new("/:page(.:format)?" => ["/:page/view.:format", "/:page/view.html"])
mapper['/foo']     # => "/foo/view.html"
mapper['/foo.xml'] # => "/foo/view.xml"
mapper['/foo/bar'] # => "/foo/bar"

Routers

Mustermann comes with basic router implementations that will call certain callbacks depending on the input.

Simple Router

The simple router chooses callbacks based on an input string.

require 'mustermann/router/simple'

router = Mustermann::Router::Simple.new(default: 42)
router.on(':name', capture: :digit) { |string| string.to_i }
router.call("23")      # => 23
router.call("example") # => 42

Rack Router

This is not a full replacement for Rails, Sinatra, Cuba, etc, as it only cares about path based routing.

require 'mustermann/router/rack'

router = Mustermann::Router::Rack.new do
  on '/' do |env|
    [200, {'Content-Type' => 'text/plain'}, ['Hello World!']]
  end

  on '/:name' do |env|
    name = env['mustermann.params']['name']
    [200, {'Content-Type' => 'text/plain'}, ["Hello #{name}!"]]
  end

  on '/something/*', call: SomeApp
end

# in a config.ru
run router

Sinatra Integration

All patterns implement match, which means they can be dropped into Sinatra and other Rack routers:

require 'sinatra'
require 'mustermann'

get Mustermann.new('/:foo') do
  params[:foo]
end

In fact, since using this with Sinatra is the main use case, it comes with a build-in extension for Sinatra 1.x.

require 'sinatra'
require 'mustermann'

register Mustermann

# this will use Mustermann rather than the built-in pattern matching
get '/:slug(.ext)?' do
  params[:slug]
end

Configuration

You can change what pattern type you want to use for your app via the pattern option:

require 'sinatra/base'
require 'mustermann'

class MyApp < Sinatra::Base
  register Mustermann
  set :pattern, type: :shell

  get '/images/*.png' do
    send_file request.path_info
  end

  get '/index{.htm,.html,}' do
    erb :index
  end
end

You can use the same setting for options:

require 'sinatra'
require 'mustermann'

register Mustermann
set :pattern, capture: { ext: %w[png jpg html txt] }

get '/:slug(.:ext)?' do
  # slug will be 'foo' for '/foo.png'
  # slug will be 'foo.bar' for '/foo.bar'
  # slug will be 'foo.bar' for '/foo.bar.html'
  params[:slug]
end

It is also possible to pass in options to a specific route:

require 'sinatra'
require 'mustermann'

register Mustermann

get '/:slug(.:ext)?', pattern: { greedy: false } do
  # slug will be 'foo' for '/foo.png'
  # slug will be 'foo' for '/foo.bar'
  # slug will be 'foo' for '/foo.bar.html'
  params[:slug]
end

Of course, all of the above can be combined. Moreover, the capture and the except option can be passed to route directly. And yes, this also works with before and after filters.

require 'sinatra/base'
require 'sinatra/respond_with'
require 'mustermann'

class MyApp < Sinatra::Base
  register Mustermann, Sinatra::RespondWith
  set :pattern, capture: { id: /\d+/ } # id will only match digits

  # only capture extensions known to Rack
  before '*:ext', capture: Rack::Mime::MIME_TYPES.keys do
    content_type params[:ext]                 # set Content-Type
    request.path_info = params[:splat].first  # drop the extension
  end

  get '/:id' do
    not_found unless page = Page.find params[:id]
    respond_with(page)
  end
end

Why would I want this?

  • It gives you fine grained control over the pattern matching
  • Allows you to use different pattern styles in your app
  • The default is more robust and powerful than the built-in patterns
  • Sinatra 2.0 will use Mustermann internally
  • Better exceptions for broken route syntax

Why not include this in Sinatra 1.x?

  • It would introduce breaking changes, even though these would be minor
  • Like Sinatra 2.0, Mustermann requires Ruby 2.0 or newer

Duck Typing

to_pattern

All methods converting string input to pattern objects will also accept any arbitrary object that implements to_pattern:

require 'mustermann'

class MyObject
  def to_pattern(**options)
    Mustermann.new("/foo", **options)
  end
end

object = MyObject.new
Mustermann.new(object, type: :rails) # => #<Mustermann::Rails:"/foo">

It might also be that you want to call to_pattern yourself instead of Mustermann.new. You can load mustermann/to_pattern to implement this method for strings, regular expressions and pattern objects:

require 'mustermann/to_pattern'

"/foo".to_pattern               # => #<Mustermann::Sinatra:"/foo">
"/foo".to_pattern(type: :rails) # => #<Mustermann::Rails:"/foo">
%r{/foo}.to_pattern             # => #<Mustermann::Regular:"\\/foo">
"/foo".to_pattern.to_pattern    # => #<Mustermann::Sinatra:"/foo">

You can also use the Mustermann::ToPattern mixin to easily add to_pattern to your own objects:

require 'mustermann/to_pattern'

class MyObject
  include Mustermann::ToPattern

  def to_s
    "/foo"
  end
end

MyObject.new.to_pattern # => #<Mustermann::Sinatra:"/foo">

respond_to?

You can and should use respond_to? to check if a pattern supports certain features.

require 'mustermann'
pattern = Mustermann.new("/")

puts "supports expanding"             if pattern.respond_to? :expand
puts "supports generating templates"  if pattern.respond_to? :to_templates

Alternatively, you can handle a NotImplementedError raised from such a method.

require 'mustermann'
pattern = Mustermann.new("/")

begin
  p pattern.to_templates
rescue NotImplementedError
  puts "does not support generating templates"
end

This behavior corresponds to what Ruby does, for instance for fork.

Available Options

capture

Supported by: All types except identity, shell and simple patterns.

Most pattern types support changing the strings named captures will match via the capture options.

Possible values for a capture:

# String: Matches the given string (or any URI encoded version of it)
Mustermann.new('/index.:ext', capture: 'png')

# Regexp: Matches the Regular expression
Mustermann.new('/:id', capture: /\d+/)

# Symbol: Matches POSIX character class
Mustermann.new('/:id', capture: :digit)

# Array of the above: Matches anything in the array
Mustermann.new('/:id_or_slug', capture: [/\d+/, :word])

# Hash of the above: Looks up the hash entry by capture name and uses value for matching
Mustermann.new('/:id.:ext', capture: { id: /\d+/, ext: ['png', 'jpg'] })

Available POSIX character classes are: :alnum, :alpha, :blank, :cntrl, :digit, :graph, :lower, :print, :punct, :space, :upper, :xdigit, :word and :ascii.

except

Supported by: All types except identity, shell and simple patterns.

Given you supply a second pattern via the except option. Any string that would match the primary pattern but also matches the except pattern will not result in a successful match. Feel free to read that again. Or just take a look at this example:

pattern = Mustermann.new('/auth/*', except: '/auth/login')
pattern === '/auth/dunno' # => true
pattern === '/auth/login' # => false

Now, as said above, except treats the value as a pattern:

pattern = Mustermann.new('/*anything', type: :rails, except: '/*anything.png')
pattern === '/foo.jpg' # => true
pattern === '/foo.png' # => false

greedy

Supported by: All types except identity and shell patterns. Default value: true

Simple patterns are greedy, meaning that for the pattern :foo:bar?, everything will be captured as foo, bar will always be nil. By setting greedy to false, foo will capture as little as possible (which in this case would only be the first letter), leaving the rest to bar.

All other supported patterns are semi-greedy. This means :foo(.:bar)? (:foo(.:bar) for Rails patterns) will capture everything before the last dot as foo. For these two pattern types, you can switch into non-greedy mode by setting the greedy option to false. In that case foo will only capture the part before the first dot.

Semi-greedy behavior is not specific to dots, it works with all characters or strings. For instance, :a(foo:b) will capture everything before the last foo as a, and :foo(bar)? will not capture a bar at the end.

pattern = Mustermann.new(':a.:b', greedy: true)
pattern.match('a.b.c.d') # => #<MatchData a:"a.b.c" b:"d">

pattern = Mustermann.new(':a.:b', greedy: false)
pattern.match('a.b.c.d') # => #<MatchData a:"a" b:"b.c.d">

space_matches_plus

Supported by: All types except identity, regexp and shell patterns. Default value: true

Most pattern types will by default also match a plus sign for a space in the pattern:

Mustermann.new('a b') === 'a+b' # => true

You can disable this behavior via space_matches_plus:

Mustermann.new('a b', space_matches_plus: false) === 'a+b' # => false

Important: This setting has no effect on captures, captures will always keep plus signs as plus sings and spaces as spaces:

pattern = Mustermann.new(':x')
pattern.match('a b')[:x] # => 'a b'
pattern.match('a+b')[:x] # => 'a+b'

uri_decode

Supported by all pattern types. Default value: true

Usually, characters in the pattern will also match the URI encoded version of these characters:

Mustermann.new('a b') === 'a b'   # => true
Mustermann.new('a b') === 'a%20b' # => true

You can avoid this by setting uri_decode to false:

Mustermann.new('a b', uri_decode: false) === 'a b'   # => true
Mustermann.new('a b', uri_decode: false) === 'a%20b' # => false

ignore_unknown_options

Supported by all patterns. Default value: false

If you pass an option in that is not supported by the specific pattern type, Mustermann will raise an ArgumentError. By setting ignore_unknown_options to true, it will happily ignore the option.

Performance

It's generally a good idea to reuse pattern objects, since as much computation as possible is happening during object creation, so that the actual matching or expanding is quite fast.

Pattern objects should be treated as immutable. Their internals have been designed for both performance and low memory usage. To reduce pattern compilation, Mustermann.new and Mustermann::Pattern.new might return the same instance when given the same arguments, if that instance has not yet been garbage collected. However, this is not guaranteed, so do not rely on object identity.

String Matching

When using a pattern instead of a regular expression for string matching, performance will usually be comparable.

In certain cases, Mustermann might outperform naive, equivalent regular expressions. It achieves this by using look-ahead and atomic groups in ways that work well with a backtracking, NFA-based regular expression engine (such as the Oniguruma/Onigmo engine used by Ruby). It can be difficult and error prone to construct complex regular expressions using these techniques by hand. This only applies to patterns generating an AST internally (all but identity, shell, simple and regexp patterns).

When using a Mustermann pattern as a direct Regexp replacement (ie, via methods like =~, match or ===), the overhead will be a single method dispatch, which some Ruby implementations might even eliminate with method inlining. This only applies to patterns using a regular expression internally (all but identity and shell patterns).

Expanding

Pattern expansion significantly outperforms other, widely used Ruby tools for generating URLs from URL patterns in most use cases.

This comes with a few trade-offs:

  • As with pattern compilation, as much computation as possible has been shifted to compiling expansion rules. This will add compilation overhead, which is why patterns only generate these rules on the first invocation to Mustermann::Pattern#expand. Create a Mustermann::Expander instance yourself to get better control over the point in time this computation should happen.
  • Memory is sacrificed in favor of performance: The size of the expander object will grow linear with the number of possible combination for expansion keys ("/:foo/:bar" has one such combination, but "/(:foo/)?:bar?" has four)
  • Parsing a params hash from a string generated from another params hash might not result in two identical hashes, and vice versa. Specifically, expanding ignores capture constraints, type casting and greediness.
  • Partial expansion is (currently) not supported.

Details on Pattern Types

identity

Supported options: uri_decode, ignore_unknown_options.

Syntax Element Description
any character Matches exactly that character or a URI escaped version of it.

regexp

Supported options: uri_decode, ignore_unknown_options.

Syntax Element Description
any string Interpreted as regular expression.

sinatra

Supported options: capture, except, greedy, space_matches_plus, uri_decode, ignore_unknown_options.

Syntax Element Description
:name or {name} Captures anything but a forward slash in a semi-greedy fashion. Capture is named name. Capture behavior can be modified with capture and greedy option.
*name or {+name} Captures anything in a non-greedy fashion. Capture is named name.
* or {+splat} Captures anything in a non-greedy fashion. Capture is named splat. It is always an array of captures, as you can use it more than once in a pattern.
(expression) Enclosed expression is a group. Useful when combined with ? to make it optional, or to separate two elements that would otherwise be parsed as one.
(expression|expression|...) Will match anything matching the nested expressions. May contain any other syntax element, including captures.
x? Makes x optional. For instance, (foo)? matches foo or an empty string.
/ Matches forward slash. Does not match URI encoded version of forward slash.
\x Matches x or URI encoded version of x. For instance \* matches *.
any other character Matches exactly that character or a URI encoded version of it.