ITF1788 – Interval Test Framework for IEEE Std 1788-2015

Create portable unit tests in an easy domain specific language
Share unit tests between several (compatible) libraries
Extend with plugins for additional interval arithmetic libraries

Overview

This framework can compile unit tests for a wide range of interval arithmetic libraries. Test cases can be written in a simple domain-specific language using polish prefix notation and interval literals. This way unit tests must not be written separately over and over again, but can be shared between libraries (where semantics and data types would allow it).

To give an example, the following simple unit test is compiled for several interval arithmetic libraries.

testcase example {
	disjoint [3.0, 4.0] [1.0, 2.0] = true;
}

C++: libieep1788

https://github.com/nehmeier/libieeep1788

template<typename T>
using I = p1788::infsup::interval<T, p1788::flavor::infsup::setbased::mpfr_bin_ieee754_flavor>;

BOOST_AUTO_TESTCASE(EXAMPLE)
{
	BOOST_CHECK(disjoint(I<double>(3.0, 4.0), I<double>(1.0, 2.0)));
}

GNU Octave: interval package

https://octave.sourceforge.io/interval/

## example

%!test
%! assert (disjoint (infsup (3.0, 4.0), infsup (1.0, 2.0)));

Julia: ValidatedNumerics

https://github.com/JuliaIntervals/IntervalArithmetic.jl/
Note: IntervalArithmetic.jl is not conforming to IEEE Std 1788-2015 yet, but development is on its way.

using IntervalArithmetic
using Base.Test

@testset "example" begin
    @test isdisjoint(Interval(3.0, 4.0), Interval(1.0, 2.0))
end

C++: Gaol

https://sourceforge.net/projects/gaol/
Note: Gaol is not fully conforming to IEEE Std 1788-2015. It does not support decorated intervals.

using namespace gaol;

BOOST_AUTO_TESTCASE(EXAMPLE)
{
	BOOST_CHECK(interval(3.0, 4.0).set_disjoint(interval(1.0, 2.0)));
}

Python: pyIbex

https://github.com/benEnsta/pyIbex
Note: pyIbex is not fully conforming to IEEE Std 1788-2015. It does not support decorated intervals.

import unittest
import pyIbex
from pyIbex import Interval

class TestCase_example(unittest.TestCase):
    def test_1(self):
       self.assertTrue(Interval(3.0, 4.0).is_disjoint(Interval(1.0, 2.0)))

C++: C-XSC

http://www2.math.uni-wuppertal.de/~xsc/xsc/cxsc/
Note: C-XSC is not conforming to IEEE Std 1788-2015. It does not support decorated intervals. Empty intervals trigger an exception, which renders such tests useless. Several standard functions are not supported.

using namespace cxsc;

BOOST_AUTO_TESTCASE(EXAMPLE)
{
	BOOST_CHECK(Disjoint(interval(3.0, 4.0), interval(1.0, 2.0)));
}

C++: Ibex

http://www.ibex-lib.org/
Note: Ibex is not fully conforming to IEEE Std 1788-2015. It does not support decorated intervals.

using namespace ibex;

BOOST_AUTO_TESTCASE(EXAMPLE)
{
	BOOST_CHECK(Interval(3.0, 4.0).is_disjoint(Interval(1.0, 2.0)));
}

GNU Octave: INTLAB toolbox

http://www.ti3.tu-harburg.de/rump/intlab/
Note: INTLAB is not conforming to IEEE Std 1788-2015. It does not support decorated or empty intervals. Several standard functions are not supported.

## example

%!test
%! assert (emptyintersect (infsup (3.0, 4.0), infsup (1.0, 2.0)));

ITL – Interval Test Libraries

The itl subfolder contains several interval arithmetic unit tests from various libraries, which are based on intervals with boundaries of type binary64. If you are interested in sharing test libraries, please provide them as a pull request.

Function names are based on IEEE Std 1788-2015, IEEE standard for interval arithmetic, and support for more functions can easily be added in plugins (see below). Operands and results are either intervals, numbers, boolean values or arrays. Intervals may be given in hexadecimal notation. Both, bare and decorated intervals are supported.

Simple test cases use a syntax of the type function and operands = expected result. Advanced test cases can be written as function and operands = tight result <= accurate result, where a tight result would be desired, but it suffices if the actual result is included in a wider, still accurate result. Tests for IEEE Std 1788-2015 exceptions can be written as function and operands = expected result signal expected exception.

For details read

M. Kiesner, M. Nehmeier, J. Wolff von Gudenberg (2015). “ITF1788: An Interval Testframework for IEEE 1788.” https://se2.informatik.uni-wuerzburg.de/publications/tr495.pdf

Plugins

Support for programming languages, test frameworks and interval arithmetic libraries is added via plugins. The src/plugin subfolder already contains support for some and more can be added using the template yaml files.

Plugins are independent of each other and are subdivided into (1) programming language, (2) testing framework and (3) interval arithmetic library. So, if you want to add support for another C++ interval library, you only need to customize the library's part, that is how constructors and arithmetic functions are used.

Usage

For global installation: Use python3 setup.py install (probably requires root privileges). Alternatively, add the project workspace to the environment variable PYTHONPATH.

Then you can use python3 -m itf1788 --help to show the list of available options and usage examples.

oheim / ITF1788