9.3 KiB
Tests
About testing in the hledger project, as of 201809.
Kinds of tests
- Isolated — tests should return the same results regardless of the order in which they are run.
- Composable — if tests are isolated, then I can run 1 or 10 or 100 or 1,000,000 and get the same results.
- Fast — tests should run quickly.
- Inspiring — passing the tests should inspire confidence
- Writable — tests should be cheap to write relative to the cost of the code being tested.
- Readable — tests should be comprehensible for reader, invoking the motivation for writing this particular test.
- Behavioral — tests should be sensitive to changes in the behavior of the code under test. If the behavior changes, the test result should change.
- Structure-insensitive — tests should not change their result if the structure of the code changes.
- Automated — tests should run without human intervention.
- Specific — if a test fails, the cause of the failure should be obvious.
- Deterministic — if nothing changes, the test result shouldn’t change.
- Predictive — if the tests all pass, then the code under test should be suitable for production." --Kent Beck
-
Unit tests
Unit tests exercise small chunks of functionality. In hledger, that means a function. So, many of our functions have one or more unit tests. These are mostly in hledger-lib, with a few in hledger.
Our unit tests use the tasty test runner, tasty-hunit HUnit-style tests, and some helpers from Hledger.Utils.Test, such as:
tests
andtest
aliases fortestGroup
andtestCase
assert*
helpers for constructing various kinds of assertions
We would like our unit tests to be:
- easy to read (clear, concise)
- easy to write (low boilerplate, low cognitive load)
- easy to maintain (easy to edit, easy to refactor, robust)
- easy to associate with the code under test (easy to view/jump between code & test, easy to estimate coverage)
- and scalable (usable for all devs, easy to run and select, suitable for small/large modules/packages).
Here's the current pattern (let us know if you see a better way):
module Foo ( ... tests_Foo -- export this module's and submodules' tests ) where import Hledger -- provides Hledger.Utils.Test helpers import Bar -- submodules, providing tests_Bar etc. import Baz functionA = ... functionB = ... functionC = ... functionD = ... tests_Foo = tests "Foo" [ -- define tests at the end of each module -- a group of several named tests for functionA tests "functionA" [ test "a basic test" $ assertBool "" SOMEBOOL ,test "a pretty equality test" $ SOMEEXPR @?= EXPECTEDVALUE ,test "a pretty parsing test" $ assertParseEq PARSER INPUT EXPECTEDRESULT ,test "a multiple assertions test" $ do A @?= B doSomeIO C @?= D ] -- a single test containing multiple unnamed assertions for functionB ,test "functionB" $ do assertBool "" BOOL EXPR @?= VALUE ,tests_Foo -- aggregate submodule tests ,tests_Bar ]
Here are some real-world examples.
The unit tests are shipped as part of the hledger executable, and can always be run via the test command (
hledger test
).Here's the quick way to run unit tests while developing:
make ghcid-test
ormake ghcid-test-Some.Module
. -
Doc tests
Like unit tests, but defined inside functions' haddock documentation, in the style of a GHCI transcript. These test functionality, provide usage examples in the API docs, and test those examples, all at once. They are a bit more finicky and slower than unit tests. See doctest for more.
doctests do not work on Mac with GHC 8.4+, out of the box. See ghc#15105 for current status and a workaround.
-
Functional tests
Functional tests test the overall functioning of the program. For hledger, that means running
hledger
with various inputs and options and checking for the expected output. This exercises functionality in the hledger and hledger-lib packages. We do this with shelltestrunner. Tests are defined in files named*.test
under hledger/test/, grouped by component (command or topic name). For more about these, see the README there. -
Code tests
We have some tests aimed at testing eg code quality, generally defined as make rules, such as:
make haddocktest
can haddock process all code docs without errormake buildtest
does all code build warning free with the default GHC version & stackage snapshotmake buildtestall
does the code build warning free with all supported GHC versions/stackage snapshots
See below for examples.
-
Package test suites
Haskell tools like stack and cabal recognise test suites defined in a package's cabal file (or package.yaml file). These can be run via
stack test
,cabal test
etc., and they are required to build and pass by services like Stackage. Here are the currently hledger package test suites:
package test suite what it runs hledger-lib doctests doctests hledger-lib easytests unit tests hledger test builtin test command (hledger's + hledger-lib's unit tests) hledger-ui
hledger-web
Coverage
This means how thoroughly the code is tested - both in breadth (are all parts of the code tested at least a little ?) and in depth (are all possible code paths, states, situations tested ?).
Our current test coverage can be summarised like so:
package unit doc functional
hledger-lib X X X
hledger X X
hledger-ui
hledger-web
There are ways to generate detailed coverage reports for haskell unit tests, at least. It would be useful to set this up for hledger.
How to run tests
Run unit tests:
$ make unittest
Run doctests:
$ make doctest
Run functional tests (and unit tests, now):
$ stack install shelltestrunner
$ make functest
Run the package tests (unit tests, maybe doctests, but not functional tests) of all or selected packages.
$ stack test [PKG]
Run "default tests: package plus functional tests":
$ make test
Test generation of haddock docs:
$ make haddocktest
Thorough test for build issues with current GHC:
$ make buildtest
Thorough test for build issues with all supported GHC versions:
$ make buildtestall
Run built-in hledger/hledger-lib unit tests via hledger command:
$ hledger test # test installed hledger
$ stack build hledger && stack exec -- hledger test # test just-built hledger
$ hledger test --help
test [TESTPATTERN] [SEED]
Run the unit tests built in to hledger-lib and hledger,
printing results on stdout and exiting with success or failure.
Tests are run in two batches: easytest-based and hunit-based tests.
If any test fails or gives an error, the exit code will be non-zero.
If a pattern argument (case sensitive) is provided, only easytests
in that scope and only hunit tests whose name contains it are run.
If a numeric second argument is provided, it will set the randomness
seed for easytests.
Rebuild and rerun hledger/hledger-lib unit tests via ghcid:
$ make ghcid-test
Rebuild and rerun only some tests via ghcid (see hledger test --help):
$ make ghcid-test-TESTPATTERN
See all test-related make rules:
$ make help-test