Clojure tools for using the Selenium-WebDriver library
Pull request Compare This branch is 876 commits behind semperos:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.

Selenium-WebDriver Support for Clojure

This is a Clojure library for driving a web browser using Selenium-WebDriver as the backend. Credits to mikitebeka/webdriver-clj for the initial code for this project and many of the low-level wrappers around the WebDriver API.


Clojure version 1.2.1 is required to use this library due to a breaking change between Clojure versions 1.2.0 and 1.2.1 with regards to namespaces with dashes.

Use/require the library in your code:

(use 'clj-webdriver.core)

Start up a browser:

(def b (start :firefox ""))

At the moment, the best documentation is the source code itself. While there are many functions in the core namespace, they're mostly short and straightforward wrappers around WebDriver API's. For the task of finding elements on the page, I've added some utility functions at the end of the core namespace.

Here's an example of logging into Github:

;; Start the browser and bind it to `b`
(def b (start :firefox ""))

;; Click "Login" link
(-> b
    (find-it {:text "Login"})

;; Input username/email into the "Login or Email" field
(-> b
    (find-it {:class "text", :name "login"}) ; use multiple attributes
    (input-text "username"))

;; Input password into the "Password" field
(-> b
    (find-it {:xpath "//input[@id='password']"}) ; :xpath and :css options
    (input-text "password"))

;; Click the "Log in" button"
(-> b
    (find-it :input {:value #"(?i)log"}) ; use of regular expressions

The key functions for finding an element on the page are find-it and find-them. The find-it function returns the first result that matches the criteria, while find-them returns a seq of all matches for the given criteria. Both support the same syntax and set of attributes.

Here is an overview of the arguments you can pass these functions:

  • HTML Tag as keyword: Pass in the name of an HTML tag as a keyword (:div, :a, :span, :img, etc.) (find-it :a) will find the first <a> tag on the page. There are also special keywords such as :* (match any tag), :text (match textfields and textareas), :window (to match an open browser window by title or url)
  • HTML Tag plus attributes: Pass in the name of an HTML tag as a keyword plus some attributes to describe it. (find-it :a {:class "external"}) will return the first <a> tag with a class of "external"
  • HTML attributes alone: You don't have to pass in a tag. (find-it {:class "external"}) will find the first element of any tag with class "external"
  • Multiple HTML attributes: You can pass in as many attribute-value pairs as you like. (find-it {:class "external", :text "Moustache"}) will find the first HTML element on the page with both a class of "external" and visible text of "Moustache"
  • Regular Expressions: Instead of looking for an exact match, you can use Java-style regular expressions to find elements. (find-it :a {:class #"exter"}) will find the first <a> tag with a class which matches the regular expression #"exter". You can also use regexes in the final position of an ancestry-based query (see below).
  • Ancestry-based queries: This library provides a pure-Clojure mechanism for finding an element based on parent elements. (find-it [:div {:id "content"}, :a {:class "external"}]) will find the first <a> tag with a class of "external" that is located within the <div> with id "content". This is equivalent to the XPath //div[@id='content']//a[@class='external']. You can also include regular expressions in the final attribute-value map which you supply. (Note: Due to issues of ambiguity and in order not to reinvent the wheel any further, applying regexes higher up the query is not supported and will cause an exception. In addition, none of the "semantic" tags such as `:button,:radio,:checkbox,:textfield`, etc. that do not map directly to HTML tags are not supported. If you need more advanced querying, use XPath or CSS selectors directly.)*
  • XPath and CSS Selectors: You can use the :xpath and :css attributes to use such queries in place of simple HTML attributes. If you use one of these attributes, you shouldn't use any others, as they will be ignored (e.g. {:xpath "//a", :class "external"} will only utilize the xpath //a). (find-it {:xpath "//a[@class='external']"}) will return the first <a> tag with a class of "external"

As mentioned above, the find-it and find-them functions share the same features and syntax; find-it returns a single element, find-them returns a seq of all matched elements.

To demonstrate how to use arguments in different ways, consider the following example. If I wanted to find <a href="/contact" id="contact-link" class="menu-item" name="contact">Contact Us</a> in a page and click on it I could perform any of the following:

(-> b
    (find-it :a)    ; assuming its the first <a> on the page

(-> b
    (find-it {:id "contact-link"})    ; :id is unique, so only one is needed

(-> b
    (find-it {:class "menu-item", :name "contact"})    ; use multiple attributes

(-> b
    (find-it :a {:class "menu-item", :name "contact"})    ; specify tag

(-> b
    (find-it :a {:text "Contact Us"})    ; special :text attribute, uses XPath's
    click)                               ; text() function to find the element

(-> b
    (find-it :a {:class #"(?i)menu-"})  ; use Java-style regular
    click)                               ; expressions

(-> b
    (find-it [:div {:id "content"}, :a {:id "contact-link"}]) ; hierarchical/ancestry-based query
    click)                                                    ; equivalent to
                                                              ; //div[@id='content']//a[@id='contact-link']

(-> b
    (find-it [:div {:id "content"}, :a {}]) ; ancestry-based query, tag with
    click)                                  ; no attributes (empty map required)

(-> b
    (find-it {:xpath "//a[@id='contact-link']"})    ; XPath query

(-> b
    (find-it {:css "a#contact-link"})    ; CSS selector

So, to describe the general pattern of interacting with the page:

(-> browser-instance
    (find-it options)

Firefox Functionality

Support for Firefox currently exceeds that for all other browsers, most notably via support for customizable Firefox profiles. I've included support for several of these advanced featues in the clj-webdriver.firefox namespace. Here are a few examples (borrowed from here:

(use 'clj-webdriver.core)
(require '[clj-webdriver.firefox :as ff])

(def b (new-driver :firefox
                   (doto (ff/new-profile)
                         ;; Enable Firebug
                         (ff/enable-extension "/path/to/extensions/firebug.xpi")))

                         ;; Auto-download certain file types to a specific folder
                         (ff/set-preferences { "C:/Users/semperos/Desktop",
                                              :browser.helperApps.neverAsk.saveToDisk "application/pdf"})))

Running Tests

The namespace clj-webdriver.test.example-app.core contains a Ring app (routing by Moustache) that acts as my "control application" for this project's test suite. Instead of running my tests against a remote server on the Internet (prone to change, not always available), I've packaged this small web application to be run locally for the purposes of testing.

Due to some Java server/socket issues, you cannot start both this Ring app and the WebDriver browser instance in the test itself (in this situation, the Ring app starts and WebDriver opens the browser, but then a host of errors follow).

Here's how I run these tests:

  • Open a terminal and run lein repl or lein swank at the root of this project
  • Evaluate (use 'clj-webdriver.test.example-app.core 'ring.adapter.jetty)
  • Evaluate (defonce my-server (run-jetty #'routes {:port 8080, :join? false}))
  • Open a new terminal tab/window and run lein test at the root of this project

The last test in the suite closes the WebDriver browser instance; you can stop the Jetty server by evaluating (.stop my-server) or just killing the REPL with Ctrl+C or C-c C-c.

If anyone can figure out how to solve this issue (i.e. be able to run just lein test and start both the Ring app and WebDriver browser instance in one go), I'd be most appreciative. Until then, testing multiple server-based apps in separate "sandboxes" is acceptable to me.


Distributed under the Eclipse Public License, the same as Clojure.