Unit and Functional Testing

Testing Base Julia

Julia is under rapid development and has an extensive test suite to verify functionality across multiple platforms. If you build Julia from source, you can run this test suite with maketest. In a binary install, you can run the test suite using Base.runtests().

runtests([tests=["all"][, numcores=iceil(CPU_CORES/2)]])

Run the Julia unit tests listed in tests, which can be either a string or an array of strings, using numcores processors. (not exported)

Test Framework

The Test module contains macros and functions related to testing. A default handler is provided to run the tests, and a custom one can be provided by the user by using the registerhandler() function.

To use the default handler, the macro @test() can be used directly:

julia>usingBase.Testjulia>@test1==1julia>@test1==0ERROR:testfailed:1==0inerroraterror.jl:21indefault_handlerattest.jl:19indo_testattest.jl:39julia>@testerror("This is what happens when a test fails")ERROR:testerrorduringerror("This is what happens when a test fails")Thisiswhathappenswhenatestfailsinerroraterror.jl:21inanonymousattest.jl:62indo_testattest.jl:37

As seen in the examples above, failures or errors will print the abstract syntax tree of the expression in question.

Another macro is provided to check if the given expression throws an exception of type extype, @test_throws():

julia>@test_throwsErrorExceptionerror("An error")ErrorException("An error")julia>@test_throwsBoundsErrorerror("An error")ERROR:testfailed:error("An error")inerroraterror.jl:21indefault_handlerattest.jl:19indo_test_throwsattest.jl:55julia>@test_throwsDomainErrorthrow(DomainError())DomainError()julia>@test_throwsDomainErrorthrow(EOFError())ERROR:testfailed:throw(EOFError())inerroraterror.jl:21indefault_handlerattest.jl:19indo_test_throwsattest.jl:55

As floating-point values can be imprecise, you can perform approximate equality checks using either @testab (where , typed via tab completion of \approx, is the isapprox function) or use the macros @test_approx_eq macro (which differs from isapprox in that it treats NaN values as equal and has a smaller default tolerance) or @test_approx_eq_eps (which takes an extra argument indicating the relative tolerance):

julia>@test10.999999999julia>@test10.999999ERROR:testfailed:1isapprox0.999999inexpression:10.999999inerroraterror.jl:21indefault_handlerattest.jl:30indo_testattest.jl:53julia>@test_approx_eq1.0.999999999ERROR:assertionfailed:|1.0-0.999999999|<2.220446049250313e-121.0=1.00.999999999=0.999999999intest_approx_eqattest.jl:75intest_approx_eqattest.jl:80julia>@test_approx_eq1.0.9999999999999julia>@test_approx_eq_eps1.0.9991e-2julia>@test_approx_eq_eps1.0.9991e-3ERROR:assertionfailed:|1.0-0.999|<=0.0011.0=1.00.999=0.999difference=0.0010000000000000009>0.001inerroraterror.jl:22intest_approx_eqattest.jl:68

Handlers

A handler is a function defined for three kinds of arguments: Success, Failure, Error:

# An example definition of a test handlertest_handler(r::Success)=nothingtest_handler(r::Failure)=error("test failed: $(r.expr)")test_handler(r::Error)=rethrow(r)

A different handler can be used for a block (with with_handler()):

julia>usingBase.Testjulia>custom_handler(r::Test.Success)=println("Success on $(r.expr)")custom_handler(genericfunction with1method)julia>custom_handler(r::Test.Failure)=error("Error on custom handler: $(r.expr)")custom_handler(genericfunction with2methods)julia>custom_handler(r::Test.Error)=rethrow(r)custom_handler(genericfunction with3methods)julia>Test.with_handler(custom_handler)do@test1==1@test1!=1endSuccesson:((1==1))ERROR:Erroroncustomhandler::((1!=1))inerroraterror.jl:21incustom_handleratnone:1indo_testattest.jl:39inanonymousatnofile:3intask_local_storageattask.jl:28inwith_handlerattest.jl:24

The Success and Failure types include an additonal field, resultexpr, which is a partially evaluated expression. For example, in a comparison it will contain an expression with the left and right sides evaluated.

Macros

@test(ex)

Test the expression ex and calls the current handler to handle the result.

@test_throws(extype, ex)

Test that the expression ex throws an exception of type extype and calls the current handler to handle the result.

@test_approx_eq(a, b)

Test two floating point numbers a and b for equality taking in account small numerical errors.

@test_approx_eq_eps(a, b, tol)

Test two floating point numbers a and b for equality taking in account a margin of tolerance given by tol.

@inferred f(x)

Tests that the call expression f(x) returns a value of the same type inferred by the compiler. It’s useful to check for type stability.

f(x) can be any call expression. Returns the result of f(x) if the types match, and an ErrorResult if it finds different types.

julia>usingBase.Testjulia>f(a,b,c)=b>1?1:1.0f(genericfunction with1method)julia>typeof(f(1,2,3))Int64julia>@code_warntypef(1,2,3)Variables:a::Int64b::Int64c::Int64Body:begin# none, line 1:unless(Base.slt_int)(1,b::Int64)::Boolgoto0return10:return1.0end::UNION{FLOAT64,INT64}julia>@inferredf(1,2,3)ERROR:returntype Int64doesnotmatchinferredreturntype Union{Float64,Int64}inerrorat./error.jl:21julia>@inferredmax(1,2)2

Functions

with_handler(f, handler)

Run the function f using the handler as the handler.