(deftest inc-test
(is (= 1 (inc 0))))
in eg becomes:
(eg inc 0 1)
eg supports Clojure, Babashka, ClojureScript JVM, and ClojureScript JS, the same as clojure.test.
Vanilla clojure.test
and eg test results are included in the same report.
Also, no custom test runner is required – run your clojure.test tests as usual.
Check the ideas driving eg.
Leiningen/Boot
[eg "0.5.6-alpha"]
Clojure CLI/deps.edn
eg {:mvn/version "0.5.6-alpha"}
Let's try eg! Start by creating a REPL session and then requiring eg
and ge
:
(require '[eg :refer [eg ge]])
eg
stands for e.g. (short for example), and ge
is just eg
reversed.
Each eg test tests one function using examples. You could think of it as a function's test definition:
(eg not ; testing clojure.core/not
false ; with input parameter `false`
true) ; returning expected value `true`
a clojure.test
test named not-test
was generated.
If a function under test takes multiple parameters, we need to wrap these with []
:
(eg * [3 2] 6)
When parameter is a vector, it needs to be wrapped with []
as well:
(eg set
[[1 2]] #{1 2}) ; cannot do: [1 2] #{1 2}
Each eg test can contain an arbitrary number of examples:
(eg *
[3] 3
[3 2] 6)
There are times when we prefer to have expected values
on the left, and input parameters on the right.
For that we use ge
, a useful mnemonic for the inverted flow of the test example:
(ge + 10 [3 7])
Expected values can be checked against a predicate or a spec:
(require '[clojure.spec.alpha :as spec])
(spec/def ::int integer?)
(eg dec
[4] integer?
[5] ::int)
=>
or <=
operators between input parameters and expected value can be used to improve readability, or
override the default example direction of eg
or ge
.
(eg hash-map
[:d 1] {:d 1}
[:a 1 :b 2 :c 3 :d 4] => {:a 1 :b 2 :c 3 :d 4})
=>
or <=
could be read as is. {:a 1 :b 2 :c 3 :d 4}
is a map?
:
(eg hash-map
map? <= {:a 1}
{:a 1} coll?) ; when no operator is used, 'is' operator semantics are assumed
ex
makes it possible to test the result from evaluating an arbitrary s-expression. Typical scenarios include testing the result of calling a macro (eg
, and ge
only support function testing), or decomposing the assertion of different properties or values from calling a form:
; let must be used outside of 'ex' when surrounding examples
(let [test-eg-ret (ex (inc 0) => 1)
f-len (count "eg-test-")]
; using arrows/equal is compulsory
(let [test-eg-ret (ex (inc 0) 1)
f-len (count "eg-test-")]
(ex var? <= test-eg-ret)
(ex (-> test-eg-ret meta :test) => boolean)
(ex (-> test-eg-ret meta :test) => fn?)
(ex (-> test-eg-ret meta :name name (subs f-len)) => not-empty))
;=> eg-test-<rand-id>
(ex (true? false) => false) ;=> eg-test-<rand-id>
; when '=>' is omitted, <expected | checker> and <actual> placeholders are
; reversed. This syntax is ideal for multiline tests
(ex "Long string to examplify 'ex' syntax that is ideal for multiline tests"
(identity "Long string to examplify 'ex' syntax that is ideal for multiline tests"))
;=> eg-test-<rand-id>
(ex string?
"Long string to examplify 'ex' syntax that is ideal for multiline tests")
;=> eg-test-<rand-id>
; tests truthy expression
(ex "bar")
;=> eg-test-<rand-id>
; checks expression against a spec
; recommended to do (eg ::fixtures/string "foo") instead,
; but implemented for consistency
(ex "foo" => ::fixtures/string)
;=> eg-test-<rand-id>
If we want to check if the expected value is a function or a spec, the operator =
is used:
(defn foo [] inc)
(eg foo [] = inc)
; or
(ge foo [] = 2)
(ex (foo 2) = inc)
(ex inc = (foo))
(ex (foo) = inc)
(defn bar [] ::inc)
(eg bar [] ::inc)
There are times when we just want to test a certain input parameter value, but fill the
remainder input parameters nevertheless. eg provides a don't care placeholder _
,
for these cases:
(eg vector
[1 2 3 4] [1 2 3 4]
[5 6 _ 8] vector?
[4 _ 5] vector?)
We can arbitrarily name a don't care parameter by prefixing its name with $
. A named don't care can also be bound with parts on the expected result. Don't care parameters are generated from parameters on the same argument position:
(eg assoc-in
[{} [:a :b] {:eggs "boiled"}] => {:a {:b {:eggs "boiled"}}}
[_ $spam _] => map?
[_ _ $eggs] => {:a {:b $eggs}})
When writing examples, don't cares enable us to spend less time writting fillers, and the reader is able to better understand the focus of the example.
Expecting arbitrarily nested JavaScript objects or arrays works out of the box:
(eg identity
#js {:a [1]} => #js {:a [1]}
#js {:a [1]} => (clj->js {:a [1]}))
(ex (identity #js {:a [1]}) => #js {:a [1]}))
Check if your specs are on the right track using examples.
eg
validates examples against a spec defined with a qualified keyword:
(require '[clojure.spec.alpha :as spec])
(spec/def ::string (spec/nilable string?))
(spec/def ::pos-int pos-int?)
(eg ::string nil "foo") ; `^:focus` cannot be used here at the moment
;=> <current-ns>-:string-test
(ge ::pos-int
1
; test invalid examples, i.e., near a spec's boundary, using `!`
! 0
(! -1 -2 -3)) ; equivalent to: ! -1 ! -2 ! -3
Quite often, writing tests becomes an afterthought, because creating test boilerblate like a new test namespace, requiring test forms and functions under test is too much of a hassle, while being immersed in writing code. It would make sense to have test forms globally available that we use almost as often as defn
. Introducing set-eg!
– call it at the development entrypoint of your program:
(require '[eg :refer [set-eg!]])
To use eg
, ge
, and ex
in any namespace without additional requires, run:
(set-eg!)
;=> :reloading ()
To use eg
, and ge
in any namespace without additional requires, run:
(set-eg! 'eg 'ge)
PS - This functionality is only supported in Clojure.
It's possible to run only selected tests by using metadata ^:focus
on eg
or ge
:
(eg ^:focus false? [false] true)
There are some caveats to consider when using ^:focus
with ClojureScript:
- The tests report counts towards non focused tests, although examples under such tests are not executed.
- Assertions for tests defined directly with
clojure.test/deftest
will still be executed, despite the presence of focusedeg
, orge
tests.
Skip running certain tests or examples using vanilla Clojure(Script) code:
#_(eg count '(9 8 7)
3)
(eg count
; '(9 8 7) 3
'(9 8 7 6) 4)
Between eg
, and ge
, choose the form that is most convenient for your combination of function examples and use it only once for testing a function. For example, don't do this:
(ge inc [1] 2)
(ge inc [0] 1)
or this:
(eg inc [1] 2)
(ge inc [0] 1)
When building with ClojureScript JVM, remove WARNING: eg is a single segment namespace
warning with the compiler option single-segment-namespace
:
{:cljsbuild
{:builds
[{:compiler
{:warnings
{:single-segment-namespace false}}}]}}
Finally, run your tests as you normally would with clojure.test
.
(clojure.test/run-all-tests)
; or
(clojure.test/run-tests some.ns)
> lein test
Also see below Enable line information for ClojureScript (JVM) tests on Node.js.
(cljs.test/run-all-tests)
; or
(cljs.test/run-tests some.ns)
Get accurate test line information on reports.
First install source maps support node package:
npm install source-map-support
Enable source maps on your project.clj
test build config:
{:cljsbuild
{:builds
{...
{:compiler {:source-map true}}}}}
Run eg's own tests expected to pass, fail, targeting Clojure, ClojureScript JVM, and ClojureScript JS.
- eftest – Eftest is a fast and pretty Clojure test runner.
- humane-test-output – Humane test output for clojure.test
- https://seespotcode.net/2018/01/13/portable-clojure-test-assert-expr/
- http://blog.nberger.com.ar/blog/2015/09/18/more-portable-complex-macro-musing
- https://github.com/clojure/clojure/blob/master/src/clj/clojure/test.clj
- https://github.com/clojure/clojurescript/blob/master/src/main/cljs/cljs/test.cljs
- https://github.com/clojure/clojurescript/blob/master/src/main/cljs/cljs/test.cljc
Copyright (c) 2020 Carlos da Cunha Fontes
The MIT License