Eftest
Eftest is a fast and pretty Clojure test runner.
Installation
To install, add the following to your project :dependencies
:
[eftest "0.6.0"]
Alternatively, if you just want to use Eftest as a lein test
replacement, add the following to your project :plugins
:
[lein-eftest "0.6.0"]
Screenshots
When all the tests pass, it looks like this:
When a test fails, it looks like:
And when a test throws an exception, it looks like:
Usage
Library
Eftest has two main functions: find-tests
and run-tests
.
The find-tests
function searches a source, which can be a namespace,
directory path, symbol, var, or a collection of any of the previous.
It returns a collection of test vars found in the source.
The run-tests
function accepts a collection of test vars and runs
them, delivering a report on the tests as it goes.
Typically these two functions are used together:
user=> (require '[eftest.runner :refer [find-tests run-tests]])
nil
user=> (run-tests (find-tests "test"))
...
The above example will run all tests found in the "test" directory.
Multithreading
By default Eftest runs all tests in parallel, which can cause issues
with tests that expect to be single-threaded. To disable this and set
all tests to be executed in serial, set the :multithread?
option to
false
:
user=> (run-tests (find-tests "test") {:multithread? false})
If you want the test vars inside a namespace to be executed in
parallel, but the namespaces themselves to be executed in serial, then
set the :multithread?
option to :vars
:
user=> (run-tests (find-tests "test") {:multithread? :vars})
If you want the vars inside a namespace to execute in serial, but the
namespaces to be executed in parallel, set the :multithread?
option
to :namespaces
:
user=> (run-tests (find-tests "test") {:multithread? :namespaces})
Alternatively, you can add the :eftest/synchronized
key as metadata
to any tests you want to force to be executed in serial:
(deftest ^:eftest/synchronized a-test
(is (= 1 1)))
Or you can synchronize the entire namespace:
(ns ^:eftest/synchronized foo.core-test
(:require [clojure.test :refer :all]
[foo.core :refer :all]))
Setting the number of threads used
When multithreading is enabled, Eftest uses a single fixed-threadpool
ExecutorService
to run all selected tests.
By default, Eftest will instantiate the threadpool with the number of processors
(cores) available to the JVM, as reported by
Runtime.availableProcessors
. (NB: in some
circumstances, such as in a CircleCI test container,
Runtime.availableProcessors
returns an erroneous value.)
Users can override the default behavior by including the key :thread-count
in the options map supplied to run-tests
with the value being any
positive integer:
user=> (run-tests (find-tests "test") {:thread-count 4})
Reporting
You can also change the reporting function used. For example, if you want a colorized reporter but without the progress bar:
user=> (run-tests (find-tests "test") {:report eftest.report.pretty/report})
Or JUnit output:
user=> (run-tests (find-tests "test") {:report eftest.report.junit/report})
Or maybe you just want the old Clojure test reporter:
user=> (run-tests (find-tests "test") {:report clojure.test/report})
If you want to redirect reporting output to a file, use the
eftest.report/report-to-file
function:
user=> (require '[eftest.report :refer [report-to-file]])
nil
user=> (require '[eftest.report.junit :as ju])
nil
user=> (run-tests (find-tests "test") {:report (report-to-file ju/report "test.xml")})
Output capturing
By default the STDOUT and STDERR from each test is captured, and displayed only if the test fails. This includes all the output generated by the test itself, and any output generated by other threads not currently running a test.
To turn off this behavior, set the :capture-output?
option to
false
:
user=> (run-tests (find-tests "test") {:capture-output? false})
Fast failure
Sometimes it's useful to end the testing on the first test failure. To
do this set the :fail-fast?
option to true
:
user=> (run-tests (find-tests "test") {:fail-fast? true})
Long test reporting
If you wish to monitor the length of time to run each test, you can
set the :test-warn-time
option to the threshold in milliseconds you
wish to warn on long test for. The measured duration includes the
running of :each-fixtures
, but not :once-fixtures
.
If you know a particular test to be slow and are ok with that, and
don't want to continually be warned about it, you can add the metadata
:eftest/slow
to either the individual test, or the entire namespace,
to prevent reporting.
Note that currently only the pretty
and progress
reporters support
logging long tests.
;; Warns for all tests that take longer than 500ms
user=> (run-tests (find-tests "test") {:test-warn-time 500})
Plugin
To use the Lein-Eftest plugin, just run:
lein eftest
You can customize the reporter and configure the concurrency settings
by adding an :eftest
key to your project map:
:eftest {:multithread? :vars
:thread-count 4
:report eftest.report.junit/report
;; You can optionally write the output to a file like so:
:report-to-file "target/junit.xml"}
Leiningen test selectors also work. With namespaces:
lein eftest foo.bar-test foo.baz-test
And with metadata keywords:
lein eftest :integration
License
Copyright © 2019 James Reeves
Distributed under the Eclipse Public License either version 1.0 or (at your option) any later version.