A validation DSL for Clojure & Clojurescript applications
- Annotated Source
- Motivation
- Setup
- Usage
- Validator sets
- Customization support
- Custom validators using arbitrary functions
- Writing validators
- [Internationalization and customised error messages] (#internationalization-and-customised-error-messages)
- Built-in validators
- Contributing
- TODO
- CHANGELOG
- CONTRIBUTORS
- License
Check this blog post where I explain in detail the motivation behind this library
Stable release:
Development release:
[bouncer "0.3.2-SNAPSHOT"]
Check out the CHANGELOG to see what's new.
Then, require the library:
(ns some.ns
(:require [bouncer.core :as b]
[bouncer.validators :as v]))
bouncer provides two main validation functions, validate
and valid?
valid?
is a convenience function built on top of validate
:
(b/valid? {:name nil}
:name v/required)
;; false
validate
takes a map and one or more validation forms and returns a vector.
The first element in this vector contains a map of the error messages, whereas the second element contains the original map, augmented with the error messages.
Let's look at a few examples:
Below is an example where we're validating that a given map has a value for both the keys :name
and :age
.
(def person {:name "Leo"})
(b/validate person
:name v/required
:age v/required)
;; [{:age ("age must be present")}
;; {:name "Leo", :bouncer.core/errors {:age ("age must be present")}}]
As you can see, since age is missing, it's listed in the errors map with the appropriate error messages.
Error messages can be customized by providing a :message
option - e.g: in case you need them internationalized:
(b/validate person
:age [[v/required :message "Idade é um atributo obrigatório"]])
;; [{:age ("Idade é um atributo obrigatório")}
;; {:name "Leo", :bouncer.core/errors {:age ("Idade é um atributo obrigatório")}}]
Note the double vector:
- the inner one wraps a single validation where the first element is the validating function and the rest are options for that validation.
- the outer vector simply denotes a list of validations to be applied
Below is an example of date-time validation. A clj-time formatter may optionally be supplied.
(ns some.ns
(:require [bouncer.core :as b]
[bouncer.validators :as v]
[clj-time.format :as f])) ;; cljs-time for clojurescript users
(def person {:name "Jeb" :last-login "2014-10-21 18:00:00"})
(b/validate person
:name v/required
:last-login v/datetime (:mysql f/formatters))
;; [nil {:name "Jeb" :last-login "2014-10-21 18:00:00"}]
Nested maps can easily be validated as well, using the built-in validators:
(def person-1
{:address
{:street nil
:country "Brazil"
:postcode "invalid"
:phone "foobar"}})
(b/validate person-1
[:address :street] v/required
[:address :postcode] v/number
[:address :phone] [[v/matches #"^\d+$"]])
;;[{:address
;; {:phone ("phone must match the given regex pattern"),
;; :postcode ("postcode must be a number"),
;; :street ("street must be present")}}
;; {:bouncer.core/errors {:address {
;; :phone ("phone must match the given regex pattern"),
;; :postcode ("postcode must be a number"),
;; :street ("street must be present")}},
;; :address {:country "Brazil", :postcode "invalid", :street nil,
;; :phone "foobar"}}]
In the example above, the vector of keys is assumed to be the path in an associative structure.
bouncer
features a short circuit mechanism for multiple validations within a single field.
For instance, say you're validating a map representing a person and you expect the key :age
to be required
, a number
and also be positive
:
(b/validate {:age nil}
:age [v/required v/number v/positive])
;; [{:age ("age must be present")} {:bouncer.core/errors {:age ("age must be present")}, :age nil}]
As you can see, only the required
validator was executed. That's what I meant by the short circuit mechanism. As soon as a validation fails, it exits and returns that error, skipping further validators.
However, note this is true within a single map entry. Multiple map entries will have all its messages returned as expected:
(b/validate person-1
[:address :street] v/required
[:address :postcode] [v/number v/positive])
;; [{:address {:postcode ("postcode must be a number"), :street ("street must be present")}} {:bouncer.core/errors {:address {:postcode ("postcode must be a number"), :street ("street must be present")}}, :address {:country "Brazil", :postcode "invalid", :street nil, :phone "foobar"}}]
Also note that if we need multiple validations against any keyword or path, we need only provide them inside a vector, like [v/number v/positive]
above.
Sometimes it's useful to perform simple, ad-hoc checks in collections contained within a map. For that purpose, bouncer provides every
.
Its usage is similar to the validators seen so far. This time however, the value in the given key/path must be a collection (vector, list etc...)
Let's see it in action:
(def person-with-pets {:name "Leo"
:pets [{:name nil}
{:name "Gandalf"}]})
(b/validate person-with-pets
:pets [[v/every #(not (nil? (:name %)))]])
;;[{:pets ("All items in pets must satisfy the predicate")}
;; {:name "Leo", :pets [{:name nil} {:name "Gandalf"}],
;; :bouncer.core/errors {:pets ("All items in pets must satisfy the predicate")}}]
All we need to do is provide a predicate function to every
. It will be invoked for every item in the collection, making sure they all pass.
Note that if a map is threaded through multiple validators, bouncer will leave it's errors map untouched and simply add new validation errors to it:
(-> {:age "NaN"}
(b/validate :name v/required)
second
(b/validate :age v/number)
second
::b/errors)
;; {:age ("age must be a number"), :name ("name must be present")}
Validators can take a pre-condition option :pre
that causes it to be executed only if the given pre-condition - a truthy function - is met.
Consider the following:
(b/valid? {:a -1 :b "X"}
:b [[v/member #{"Y" "Z"} :pre (comp pos? :a)]])
;; true
As you can see the value of b
is clearly not in the set #{"Y" "Z"}
, however the whole validation passes because the v/member
check states is should only be run if :a
is positive.
Let's now make it fail:
(b/valid? {:a 1 :b "X"}
:b [[v/member #{"Y" "Z"} :pre (comp pos? :a)]])
;; false
If you find yourself repeating a set of validators over and over, chances are you will want to group and compose them somehow. Validator sets are simply plain Clojure maps:
;; first we define the set of validators we want to use
(def address-validations
{:postcode [v/required v/number]
:street v/required
:country v/required})
;;just something to validate
(def person {:address {
:postcode ""
:country "Brazil"}})
;;now we compose the validators
(b/validate person
:name v/required
:address address-validations)
;;[{:address
;; {:postcode ("postcode must be a number" "postcode must be present"),
;; :street ("street must be present")},
;; :name ("name must be present")}
;;
;; {:bouncer.core/errors {:address {:postcode ("postcode must be a number" "postcode must be present"),
;; :street ("street must be present")}, :name ("name must be present")},
;; :address {:country "Brazil", :postcode ""}}]
You can also compose validator sets together and use them as top level validations:
(def address-validator
{:postcode v/required})
(def person-validator
{:name v/required
:age [v/required v/number]
:address address-validator})
(b/validate {}
person-validator)
;;[{:address {:postcode ("postcode must be present")}, :age ("age must be present"), :name ("name must be present")} {:bouncer.core/errors {:address {:postcode ("postcode must be present")}, :age ("age must be present"), :name ("name must be present")}}]
Using your own functions as validators is simple:
(defn young? [age]
(< age 25))
(b/validate {:age 29}
:age [[young? :message "Too old!"]])
;; [{:age ("Too old!")}
;; {:bouncer.core/errors {:age ("Too old!")}, :age 29}]
As shown above, validators as just functions. The downside is that by using a function bouncer will default to a validation message that might not make sense in a given scenario:
(b/validate {:age 29}
:age young?)
;; [{:age ("Custom validation failed for age")}
;; {:bouncer.core/errors {:age ("Custom validation failed for age")}, :age 29}]
You could of course use the message keyword as in previous examples but if you reuse the validation in several places, you'd need a lot of copying and pasting.
Another way - and the preferred one - to provide custom validations is to use the macro defvalidator
in the bouncer.validators
namespace.
The advantage of this approach is that it attaches the needed metadata for bouncer to know which message to use.
As an example, here's a simplified version of the bouncer.validators/number
validator:
(use '[bouncer.validators :only [defvalidator]])
(defvalidator my-number-validator
{:default-message-format "%s must be a number"}
[maybe-a-number]
(number? maybe-a-number))
defvalidator
takes your validator name, an optional map of options and the body of your predicate function.
Options is a map of key/value pairs where:
:default-message-format
- to be used when clients of this validator don't provide one:optional
- a boolean indicating if this validator should only trigger for keys that have a value different thannil
. Defaults to false.
That's all syntactic sugar for:
(def my-number-validator
(with-meta (fn my-number-validator
([maybe-a-number]
(number? maybe-a-number)))
{:default-message-format "%s must be a number", :optional false}))
Using it is then straightforward:
(b/validate {:postcode "NaN"}
:postcode my-number-validator)
;; [{:postcode ("postcode must be a number")}
;; {:bouncer.core/errors {:postcode ("postcode must be a number")}, :postcode "NaN"}]
As you'd expect, the message can be customized as well:
(b/validate {:postcode "NaN"}
:postcode [[my-number-validator :message "must be a number"]])
Your validators aren't limited to a single argument though.
Since v0.2.2, defvalidator
takes an arbitrary number of arguments. The only thing you need to be aware of is that the value being validated will always be the first argument you list - this applies if you're using plain functions too. Let's see an example with the member
validator:
(defvalidator member
[value coll]
(some #{value} coll))
Yup, it's that simple. Let's use it:
(def kid {:age 10})
(b/validate kid
:age [[member (range 5)]])
In the example above, the validator will be called with 10
- that's the value the key :age
holds - and (0 1 2 3 4)
- which is the result of (range 5)
and will be fed as the second argument to the validator.
In some cases the default behaviour might not be enough.
Perhaps you'd like to customise your error messages with the value that has been provided.
Or maybe you need access to the options that were given to a specific validator.
And what if you need to know which validator generated a specific error message?
Since 0.3.1-beta1, this is possible. validate
takes as an optional first argument a function - called a message-fn - that is applied to a map containing validation metadata that allows you to customise error messages in any way you like - or even return other data structures instead.
This map has the following keys:
- :path - where in the map has the error ocurred?
- :value - what was the value at the time of the validation?
- :args - which arguments - if any - were passed to the validator?
- :metadata - what is the metadata associated with this validation?
- :message - what - if any - is the message passed to this validator instance?
Let's see how this works in practice. For the first example, we'll simply use identity
as our message-fn:
(def person {:name "Leo" :age "NaN"})
(b/validate identity
person
:name v/required
:age v/number)
As you can see we simply supply identity
as the first argument. This is what gets returned:
[{:age
({:path [:age],
:value "NaN",
:args nil,
:metadata
{:optional true,
:default-message-format "%s must be a number",
:validator :bouncer.validators/number},
:message nil})}
{:age "NaN",
:name "Leo",
:bouncer.core/errors
{:age
({:path [:age],
:value "NaN",
:args nil,
:metadata
{:optional true,
:default-message-format "%s must be a number",
:validator :bouncer.validators/number},
:message nil})}}]
Contrast this with what the function would have returned by default:
[{:age ("age must be a number")}
{:age "NaN",
:name "Leo",
:bouncer.core/errors {:age ("age must be a number")}}]
This opens a number of possibilities around customising errors messages. Let's create a simple message-fn to illustrate:
(defn custom-message-fn [{:keys [path value metadata]}]
(format "'%s' in field %s should be a %s" value path (:validator metadata)))
(b/validate custom-message-fn
person
:name v/required
:age v/number)
This time we get a much more informative message:
[{:age ("'NaN' in field [:age] should be a :bouncer.validators/number")}
{:age "NaN",
:name "Leo",
:bouncer.core/errors
{:age
("NaN in field [:age] should be a :bouncer.validators/number")}}]
Hats off to @dm3 for this pull request.
I didn't spend a whole lot of time on bouncer so it only ships with the validations I've needed myself. At the moment they live in the validators namespace:
-
bouncer.validators/required
-
bouncer.validators/number
-
bouncer.validators/integer
-
bouncer.validators/string
-
bouncer.validators/boolean
-
bouncer.validators/email
-
bouncer.validators/positive
-
bouncer.validators/in-range
-
bouncer.validators/member
-
bouncer.validators/max-count
-
bouncer.validators/min-count
-
bouncer.validators/matches
(for matching regular expressions) -
bouncer.validators/every
(for ad-hoc validation of collections. All items must match the provided predicate) -
bouncer.validators/datetime
(uses clj-time formatters)
Pull requests of bug fixes and new validators are most welcome.
Note that if you wish your validator to be merged and considered built-in you must implement it using the macro defvalidator
shown above.
Feedback to both this library and this guide is welcome.
Bouncer is assumed to work with Clojure 1.4 and up, as well as ClojureScript.
There is a leiningen alias that makes it easy to run the tests against multiple Clojure versions:
λ lein all-tests
It'll run all tests against Clojure 1.4, 1.5 and 1.6, as well as Clojurescript - make sure all tests pass before submitting a pull request.
- Add more validators (help is appreciated here)
- Docs are getting a bit messy. Fix that.
Copyright © 2012-2014 Leonardo Borges
Distributed under the MIT License.