ccti

ccti (Curry Concolic Testing Interpreter) is a tool for automated generation of test cases for the functional logic language Curry.

To learn more about Curry take a look at the Curry report.

News

18.12.2017: A first release of a Curry package including a prototype version of ccti is now available. You can install it using the Curry Package Manager.

Installation

Installation of z3

ccti makes use of the z3 SMT solver. Since ccti uses version 2.6 of the SMT-LIB standard which is not yet supported in the latest release version of z3 (4.5.0), you have to install the latest version from the git repository. You can specify the installation directory for z3 using the --prefix option:

git clone https://github.com/Z3Prover/z3.git
cd z3
python scripts/mk_make.py --prefix=$HOME/z3/master
cd build
make
make install

After having installed z3 please make sure that you add the z3 binary to your path.

Installation of PAKCS (Version 2)

In order to install ccti you need the PAKCS Curry compiler (with type classes). You can either download the development release here or you can build PAKCS from source by checking out the latest version from gitlab using the following command:

git clone https://git.ps.informatik.uni-kiel.de/curry/pakcs.git

For installation of PAKCS please refer to the instructions provided in the pakcs folder (INSTALL.txt for the tarball and GITINSTALL.txt in case you build from source).

For Ubuntu/Debian systems there are also DEB packages of PAKCS. Here you can find out how to install them.

Installation of ccti

After PAKCS has been successfully installed, you can install ccti via the Curry Package Manager (CPM). In order to build ccti please make sure that the binary of the Curry Package Manager is in your path, e.g. by executing cypm --help.

First, you should initialize/update your local package index using the command cypm update. Afterwards you can install ccti by executing cypm install ccti. The ccti binary will be installed to $HOME/.cpm/bin.

Alternatively, you can check out the ccti package with cypm checkout ccti, change into the corresponding directory, e.g. cd $HOME/ccti and install ccti via the Makefile provided in the package by executing make install. Afterwards you will find the binary of ccti in the bin folder. In order to make it globally accessible you should add $HOME/ccti/bin to your path.

Removing ccti

To remove ccti simply execute cypm uninstall ccti. In case you checked out the package and installed ccti locally you can simply remove the ccti folder, e.g. assuming you installed it in your $HOME directory:

rm -rf $HOME/ccti

Usage of ccti

In order to apply ccti for the automated generation of test cases for a Curry function you need to call its binary with Curry module. This Curry module has to include a main function. The body of the main function has to be a call (with concrete input data) to the function which shall be concolically tested. This call is required to initialize the concolic testing interpreter.

For instance, take a look at the following Curry module:

data Nat = IHi
         | O Nat
         | I Nat
 deriving Eq

add :: Nat -> Nat -> Nat
add IHi   y     = succ y
add (O x) IHi   = I x
add (O x) (O y) = O (add x y)
add (O x) (I y) = I (add x y)
add (I x) IHi   = O (succ x)
add (I x) (O y) = I (add x y)
add (I x) (I y) = O (add (succ x) y)

succ :: Nat -> Nat
succ IHi   = O IHi
succ (O x) = I x
succ (I x) = O (succ x)

main :: Nat
main = add IHi (I (O IHi))

This module includes a representation of positive binary numbers in Curry as well as an operation add to add two binaries. In order to generate test cases for add we provide a main function with the initial call add IHi (I (O IHi)).

Applying ccti on this module produces the following output including six test cases for add:

examples> ccti Nat.curry
========================================================================================================================
Generating FlatCurry code
[1 of 2] Skipping  Prelude          ( /local/jrt/pakcs/lib/Prelude.curry, /local/jrt/pakcs/lib/.curry/Prelude.fcy )
[2 of 2] Compiling Nat              ( Nat.curry, .curry/Nat.fcy )
========================================================================================================================
Reading FlatCurry file(s)
========================================================================================================================
Annotating case expressions with fresh identifiers
========================================================================================================================
Generating SMT-LIB declarations for FlatCurry types
========================================================================================================================
Beginning with concolic search
(Nat.add (Nat.O Nat.IHi) Nat.IHi = {Nat.I Nat.IHi})
(Nat.add (Nat.I Nat.IHi) (Nat.O Nat.IHi) = {Nat.I (Nat.O Nat.IHi)})
(Nat.add (Nat.I Nat.IHi) Nat.IHi = {Nat.O (Nat.O Nat.IHi)})
(Nat.add (Nat.I Nat.IHi) (Nat.I (Nat.O Nat.IHi)) = {Nat.O
                                                   (Nat.O (Nat.O Nat.IHi))})
(Nat.add (Nat.O Nat.IHi) (Nat.I (Nat.O Nat.IHi)) = {Nat.I (Nat.I Nat.IHi)})
(Nat.add Nat.IHi (Nat.I (Nat.O Nat.IHi)) = {Nat.O (Nat.I Nat.IHi)})

You can find the Nat example as well as some other simple examples in the examples folder of the ccti package which you can checkout via CPM by cypm checkout ccti.

Options for ccti

There are several options for ccti. You get an overview by executing ccti --help.

For instance, you can change the coverage criterion used during concolic execution. Per default branch coverage is used. But there are examples in which this criterion is not sufficient, e.g. Perm.curry. You can switch to function coverage by running the following command:

ccti --cover=function Perm.curry

If you want to take a look at the SMT-LIB scripts generated by ccti and sent to the z3 SMT solver, you can produce a dump of these scripts by running examples> ccti --dump-smt Nat.curry. The dump will be saved in a file in the hidden directory examples/.smt/.