Permalink
Fetching contributors…
Cannot retrieve contributors at this time
433 lines (340 sloc) 14.9 KB

Changes in version 2.1.4

Remove expect-let and expect-let focused.

Changes in version 2.1.3

Minor update to move CustomPred checks in front of all other checking.

Changes in version 2.1.2

Minor update to allow redef-state to work with private vars.

Changes in version 2.1.1

ClojureScript support

Make Expectations work with recent ClojureScript versions (> 0.0-2985).

Since r2985 ClojureScript introduced macro symbols in its analysis map, and Expectations tried to use them to construct vars, hence the failure described in a comment to #51. This is now fixed.

Changes in version 2.1.0

ClojureScript support

This release adds support for ClojureScript. You can write your tests the same as in Clojure, except for the usual ClojureScript quirks.

Running your tests

Your tests written in ClojureScript should be compiled to JS and run by a JavaScript runtime (such as Node.js, Phantom.js, your browser or whatnot). lein-cljsbuild is the Leiningen plugin that does that for you.

Here's a sample project.clj excerpt:

  :profiles {:dev {:node-dependencies [[source-map-support "^0.2.9"]]
                   :plugins           [[lein-cljsbuild "1.0.5"]
                                       [lein-npm "0.5.0"]]}}
  :cljsbuild {:builds [{:id             "test"                                ;; your build config name
                        :source-paths   ["src/cljs" "test/cljs"]              ;; your source and test dirs
                        :notify-command ["node" "./out/tests/test.js"]        ;; `node ./out/tests/test.js` to be
                                                                              ;; run automatically after compile
                        :compiler       {:target         :nodejs              ;; use this if you want to run on Node.js
                                         :main           my-lib.test          ;; your tests main namespace
                                         :output-to      "out/tests/test.js"  ;; compiled JS main file
                                         :output-dir     "out/tests"          ;; compiled JS dir
                                         :optimizations  :none                ;; maybe just leave as it is ;)
                                         :cache-analysis true
                                         :source-map     true
                                         :pretty-print   true}}]}

To compile and run your tests just once, run

lein cljsbuild once test

To run automatic incremental compilation and testing your changes, run

lein cljsbuild auto test

Tests main namespace (entry point)

Your tests main namespace should require expectations.cljs to be able to run your tests:

(ns my-lib.test
  (:require-macros [expectations.cljs :as ecljs])
  (:require [expectations]
            [my-lib.expectations-options]
            [my-lib.core-test]
            [my-lib.util-test]))

(defn -main []
  (ecljs/run-all-tests))

(enable-console-print!)
(set! *main-cli-fn* -main)

You should use run-all-tests macro:

(expectations.cljs/run-all-tests)

or run-tests, like this:

(expectations.cljs/run-tests my-lib.core-test my-lib.util-test my-other.namespaces)

Requiring the expectations namespace

A usual tiny difference from Clojure.

Clojure:

(ns success.nested.success-examples
  (:require [expectations :refer :all]))

ClojureScript:

(ns success.nested.success-examples
  (:require-macros [expectations :refer [expect
                                         expect-focused
                                         ;; etc.
                                         ]]))

Implementation notes

pprint is not supported in ClojureScript yet, so datastructures will not get pretty printed in failed test reports.

freeze-time macro is not yet implemented for ClojureScript.

JavaScript error stack traces will look a bit verbose as we're not eliding system and expectations files just yet.

Maintainer notes

This version uses cljx to add support for ClojureScript.

Use lein test to run Expectations' internal tests. Currently only success-examples tests have been migrated.

Changes to expectations in Version 2.0

CONTENTS

1 new and improved features

1.1 side-effects

expectations 2.0 introduces the 'side-effects macro, which allows you to capture arguments to functions you specify. Previous to version 2.0 behavior was often tested using the 'interaction macro. expectations 2.0 removes the 'interaction macro and embraces the idea of verifying interactions at the data level, using the same type of comparison that is used for all other data.

Examples:

    (expect [["/tmp/hello-world" "some data" :append true]
             ["/tmp/hello-world" "some data" :append true]]
      (side-effects [spit]
                (spit "/tmp/hello-world" "some data" :append true)
                (spit "/tmp/hello-world" "some data" :append true)))

In the above example, you specify that spit is a side effect fn, and the 'side-effects macro will return a list of all calls made, with the arguments used in the call. The above example uses simple equality for verification.

    (expect empty?
      (side-effects [spit] "spit never called"))

The above example demonstrates how you can use non-equality to to verify the data returned.

    (expect ["/tmp/hello-world" "some data" :append true]
      (in (side-effects [spit]
                    (spit "some other stuff" "xy")
                    (spit "/tmp/hello-world" "some data" :append true))))

Immediately above is an example of combining 'side-effects with 'in for a more concise test. Here we're testing that the expected data will exist somewhere within the list returned by 'side-effects

1.2 more, more->, & more-of

expectations has always given you the ability to test against an arbitrary fn, similar to the example below.

    (expect nil? nil)

The ability to specify any fn is powerful, but it doesn't always give you the most descriptive failure messages. In expectations 2.0 we introduce the 'more, 'more->, & 'more-of macros, which are designed to allow you to expect more of your actual values.

Below is a simple example of using the more macro.

    (expect (more vector? not-empty) [1 2 3])

As you can see from the above example, we're simply expecting that the actual value '[1 2 3] is both a 'vector? and 'not-empty. The 'more macro is great when you want to test a few 1 arg fns; however, I expect you'll more often find yourself reaching for 'more-> and 'more-of.

The 'more-> macro is used for threading the actual value and comparing the result to an expected value. Below is a simple example of using 'more to pull values out of a vector and test their equality.

    (expect (more-> 1 first
                    3 last)
      [1 2 3])

The 'more-> macro threads using -> (thread-first), so you're able to put any form you'd like in the actual transformation.

    (expect (more-> 2 (-> first (+ 1))
                    3 last)
      [1 2 3])

Finally, 'more-> can be very helpful for testing various kv pairs within a map, or various Java fields.

    (expect (more-> 0 .size
                    true .isEmpty)
       (java.util.ArrayList.))

    (expect (more-> 2 :a
                    4 :b)
       {:a 2 :b 4})

Threading is great work, if you can get it. For the times when you need to name your actual value, 'more-of should do the trick. The following example demonstrates how to name your actual value and then specify a few expectations.

    (expect (more-of x
                     vector? x
                     1 (first x))
      [1 2 3])

If you've ever found yourself wishing you had destructuring in clojure.test/are or expectations/given, you're not alone. The good news is, 'more-of supports any destructuring you want to give it.

    (expect (more-of [x :as all]
                     vector? all
                     1 x)
      [1 2 3])

1.3 combining side-effects and more-of

It's fairly common to expect some behavior where you know the exact values for some of the args, and you have something more general in mind for the additional args. By combining 'side-effects and 'more-of you can easily destructure a call into it's args and verify as many as you care to verify.

    (expect (more-of [path data action {:keys [a c]}]
                     String path
                     #"some da" data
                     keyword? action
                     :b a
                     :d c)
      (in (side-effects [spit]
            (spit "/tmp/hello-world" "some data" :append {:a :b :c :d :e :f}))))

The above test is a bit much to swallow at first glance; however, it's actually very straightforward once you've gotten used to the 'more-of syntax. In the above example the 'spit fn is called with the args "/tmp/hello-world", "some data" :append {:a :b :c :d :e :f}. Using 'more-of, we destructure those args, and expect them individually. The path arg is expected to be of type String. The data arg is expected to be a string that matches the regex "some da". The action is expected to be a 'keyword?. Finally, the options map is destructured to it's :a and :c values, and equality expected.

1.4 from-each

It's common to expect something from a list of actual values. Traditionally 'given was used to generate many tests from one form. Unfortunately 'given suffered from many issues: no ability to destructure values, failure line numbers were almost completely useless, and little visibility into what the problem was when a failure did occur.

In expectations 2.0 'from-each was introduced to provide a more powerful syntax as well as more helpful failure messages.

Below you can see a very simple expectation that verifies each of the elements of a vector is a String.

    (expect String
      (from-each [letter ["a" "b" "c"]]
        letter))

Hopefully the syntax of 'from-each feels very familiar, it's been written to handle the same options as 'for and 'doseq - :let and :when.

    (expect odd? (from-each [num [1 2 3]
                             :when (not= num 2)]
                   num))

    (expect odd? (from-each [num [1 2 3]
                             :when (not= num 2)
                             :let [numinc1 (inc num)]]
                   (inc numinc1)))

While 'from-each is helpful in creating concise tests, I actually find it's most value when a test fails. If you take the above test and remove the :when, you would have the test below.

    (expect odd? (from-each [num [1 2 3]
                             :let [numinc1 (inc num)]]
                   (inc numinc1)))

The above test would definitely fail, but it's not immediately obvious what the issue is. However, the failure message should quickly lead you to the underlying issue.

    failure in (success_examples.clj:206) : success.success-examples
    (expect
     odd?
     (from-each [num [1 2 3] :let [numinc1 (inc num)]] (inc numinc1)))

               the list: (3 4 5)

    (expect odd? (inc numinc1))

                 locals num: 2
                        numinc1: 3
               4 is not odd?

As you can see above, when 'from-each fails it will give you values of every var defined within the 'from-each bindings. As a result, it's fairly easy to find the combination of vars that led to a failing test.

2 changed

2.1 custom diff syntax removed

expectations was originally written before clojure.data/diff existed, and defined it's own syntax for printing diffed data between actual and expected. Now that there's a standard, it no longer makes sense to define a custom syntax. All error messages that previously used expectations custom diff syntax have been converted to simply use clojure.data/diff result maps.

Below is an example failure message that utilizes clojure.data/diff for reporting the inconsistencies.

    failure in (failure_examples.clj:23) : failure.failure-examples
    (expect
     {:foo 1, :bar 3, :dog 3, :car 4}
     (assoc {} :foo 1 :bar "3" :cat 4))

               expected: {:foo 1, :bar 3, :dog 3, :car 4}
                    was: {:cat 4, :bar "3", :foo 1}

               in expected, not actual: {:car 4, :dog 3, :bar 3}
               in actual, not expected: {:cat 4, :bar "3"}

3 removed

3.1 given

As I said above, 'given suffered from many issues: no ability to destructure values, failure line numbers were almost completely useless, and little visibility into what the problem was when a failure did occur. I personally found given expectations to often be the hardest to maintain, and often I felt they were completely unmaintainable.

As a result, 'given has been replaced with 'from-each, 'more, 'more->, & 'more-of. I've converted a few codebases over, and I've found the 'more-* or 'from-each post conversion test to be far more maintainable.

3.2 function interaction tests

expectations 2.0 abandons behavior based testing and the interaction syntax. Existing tests relying on interaction should be easy to convert to side-effects.

3.3 java mock interaction tests

java mock interaction tests never really fit well within expectations, and the removal of function interaction tests caused us to remove the java mock support as well.

3.4 Double/NaN support

The original project using expectations made heavy usage of Double/NaN, and the original version of expectations did it's best to hide the fact that Double/NaN isn't = to Double/NaN. Unfortunately, over time this hiding became confusing and complicated to maintain. In expectations 2.0 we gave up on this fight and removed Double/NaN support.

If you find yourself fighting the NaN battle, I'd suggest writing a helper method and simply using that in your tests. Here are a few examples

Changes in version 2.0.10

metadata added to 'work argument used in in-context

Since version 1.4.36 expectations has allowed you to alter the context in which your tests run by creating a function that takes the "run the tests" function as an arg, and do you as wish. more info

In version 2.0.10 of expecatations, the "run the tests" function has metadata that allows you to see the var for the test.

The following code would allow you to see the metadata of each test being run.

(defn in-context
  {:expectations-options :in-context}
  [work]
    (println (meta (:the-var (meta work))))
    (work))

Changes in version 2.0.11 and 2.0.12

None. Both versions were released as the result of automating deployment.

Changes in version 2.0.13

Tests within a namespace are run in order.

Changes from version 2.0.14 - 2.0.16

Small change allowing filenames that start with integers. The number of releases was due to tweaking automated deployment (again).