adf-obdd/lib/src/adfbiodivine.rs

//! This module describes the abstract dialectical framework
//! utilising the biodivine-lib-bdd (see <https://github.com/sybila/biodivine-lib-bdd>) BDD implementation to compute various semantics.
//!
//! These are currently the
//!  - grounded
//!  - stable
//!  - complete
//! semantics of ADFs.

use crate::{
    datatypes::{
        adf::{
            PrintDictionary, PrintableInterpretation, ThreeValuedInterpretationsIterator,
            TwoValuedInterpretationsIterator, VarContainer,
        },
        Term,
    },
    parser::AdfParser,
};

use biodivine_lib_bdd::{boolean_expression::BooleanExpression, Bdd, BddVariableSet};
use derivative::Derivative;

#[derive(Derivative)]
#[derivative(Debug)]
/// Representation of an ADF, with an ordering and dictionary which relates statements to numbers, a binary decision diagram, and a list of acceptance functions in biodivine  representation together with a variable-list (needed by biodivine).
///
/// To be compatible with results from the own implementation of the Bdd-based [`Adf`][crate::adf::Adf], we use the [`Term`][crate::datatypes::Term]-based representation for the various computed models.
pub struct Adf {
    ordering: VarContainer,
    ac: Vec<Bdd>,
    vars: Vec<biodivine_lib_bdd::BddVariable>,
    #[derivative(Debug = "ignore")]
    varset: BddVariableSet,
    rewrite: Option<Bdd>, // stable rewrite
}

impl Adf {
    /// Instantiates a new ADF, based on the parser-data.
    pub fn from_parser(parser: &AdfParser) -> Self {
        log::info!("[Start] instantiating BDD");
        let mut bdd_var_builder = biodivine_lib_bdd::BddVariableSetBuilder::new();
        let namelist = parser
            .namelist()
            .read()
            .expect("ReadLock on namelist failed")
            .clone();
        let slice_vec: Vec<&str> = namelist.iter().map(<_>::as_ref).collect();
        bdd_var_builder.make_variables(&slice_vec);
        let bdd_variables = bdd_var_builder.build();
        let mut result = Self {
            ordering: parser.var_container(),
            ac: vec![bdd_variables.mk_false(); parser.dict_size()],
            vars: bdd_variables.variables(),
            varset: bdd_variables,
            rewrite: None,
        };
        log::trace!("variable order: {:?}", result.vars);
        log::debug!("[Start] adding acs");
        parser
            .formula_order()
            .iter()
            .enumerate()
            .for_each(|(insert_order, new_order)| {
                log::trace!(
                    "Pos {}/{} formula {}, {:?}",
                    insert_order + 1,
                    parser.formula_count(),
                    new_order,
                    parser.ac_at(insert_order)
                );
                result.ac[*new_order] = result
                    .varset
                    .eval_expression(&parser.ac_at(insert_order).expect("Insert order needs to exist, as all the data originates from the same parser object").to_boolean_expr());
                log::trace!("instantiated {}", result.ac[*new_order]);
            });
        log::info!("[Success] instantiated");
        result
    }

    /// Instantiates a new ADF and prepares a rewriting for the stable model computation based on the parser-data.
    pub fn from_parser_with_stm_rewrite(parser: &AdfParser) -> Self {
        let mut result = Self::from_parser(parser);
        log::debug!("[Start] rewriting");
        result.stm_rewriting(parser);
        log::debug!("[Done] rewriting");
        result
    }

    pub(crate) fn stm_rewriting(&mut self, parser: &AdfParser) {
        let expr = parser.formula_order().iter().enumerate().fold(
            biodivine_lib_bdd::boolean_expression::BooleanExpression::Const(true),
            |acc, (insert_order, new_order)| {
                BooleanExpression::And(
                    Box::new(acc),
                    Box::new(BooleanExpression::Iff(
                        Box::new(BooleanExpression::Variable(
                            self.ordering
                                .name(crate::datatypes::Var(*new_order))
                                .expect("Variable should exist"),
                        )),
                        Box::new(parser.ac_at(insert_order).expect("Insert order needs to exist, as all the data originates from the same parser object").to_boolean_expr()),
                    )),
                )
            },
        );
        log::trace!("{:?}", expr);
        self.rewrite = Some(self.varset.eval_expression(&expr));
    }

    /// returns `true` if the stable rewriting for this ADF exists.
    pub fn has_stm_rewriting(&self) -> bool {
        self.rewrite.is_some()
    }

    pub(crate) fn var_container(&self) -> &VarContainer {
        &self.ordering
    }

    pub(crate) fn ac(&self) -> &[Bdd] {
        &self.ac
    }
    /// Computes the grounded extension and returns it as a list.
    pub fn grounded(&self) -> Vec<Term> {
        log::info!("[Start] grounded");
        let ac = &self.ac.clone();
        let result = self
            .grounded_internal(ac)
            .iter()
            .map(|elem| elem.into())
            .collect();
        log::info!("[Done] grounded");
        result
    }

    fn var_list(&self, interpretation: &[Bdd]) -> Vec<(biodivine_lib_bdd::BddVariable, bool)> {
        self.vars
            .iter()
            .enumerate()
            .filter(|(idx, _elem)| interpretation[*idx].is_truth_value())
            .map(|(idx, elem)| (*elem, interpretation[idx].is_true()))
            .collect::<Vec<(biodivine_lib_bdd::BddVariable, bool)>>()
    }

    fn var_list_from_term(
        &self,
        interpretation: &[Term],
    ) -> Vec<(biodivine_lib_bdd::BddVariable, bool)> {
        self.vars
            .iter()
            .enumerate()
            .filter(|(idx, _elem)| interpretation[*idx].is_truth_value())
            .map(|(idx, elem)| (*elem, interpretation[idx].is_true()))
            .collect::<Vec<(biodivine_lib_bdd::BddVariable, bool)>>()
    }

    pub(crate) fn grounded_internal(&self, interpretation: &[Bdd]) -> Vec<Bdd> {
        let mut new_interpretation: Vec<Bdd> = interpretation.into();
        loop {
            let mut truth_extention: bool = false;
            // let var_list: Vec<(biodivine_lib_bdd::BddVariable, bool)> = self
            //     .vars
            //     .iter()
            //     .enumerate()
            //     .filter(|(idx, _elem)| new_interpretation[*idx].is_truth_value())
            //     .map(|(idx, elem)| (*elem, new_interpretation[idx].is_true()))
            //     .collect();
            let var_list = self.var_list(&new_interpretation);

            log::trace!("var-list: {:?}", var_list);

            for ac in new_interpretation
                .iter_mut()
                .filter(|elem| !elem.is_truth_value())
            {
                log::trace!("checking ac: {}", ac);
                *ac = ac.restrict(&var_list);
                if ac.is_truth_value() {
                    truth_extention = true;
                }
            }
            if !truth_extention {
                break;
            }
        }
        new_interpretation
    }

    /// Computes the complete models.
    /// Returns an [Iterator][std::iter::Iterator] which contains all the complete models.
    pub fn complete<'a, 'b>(&'a self) -> impl Iterator<Item = Vec<Term>> + 'b
    where
        'a: 'b,
    {
        ThreeValuedInterpretationsIterator::from_bdd(&self.grounded_internal(&self.ac)).filter(
            |terms| {
                let var_list = self.var_list_from_term(terms);
                self.ac
                    .iter()
                    .enumerate()
                    .all(|(idx, ac)| terms[idx].cmp_information(&ac.restrict(&var_list)))
            },
        )
    }

    /// Shifts the representation and allows to use the naive approach.
    ///
    /// The grounded interpretation is computed by the [biodivine library](https://github.com/sybila/biodivine-lib-bdd) first.
    pub fn hybrid_step(&self) -> crate::adf::Adf {
        crate::adf::Adf::from_biodivine_vector(
            self.var_container(),
            &self.grounded_internal(self.ac()),
        )
    }

    /// Shifts the representation and allows to use the naive approach.
    ///
    /// `bio_grounded` will compute the grounded, based on [biodivine](https://github.com/sybila/biodivine-lib-bdd), first.
    pub fn hybrid_step_opt(&self, bio_grounded: bool) -> crate::adf::Adf {
        if bio_grounded {
            self.hybrid_step()
        } else {
            crate::adf::Adf::from_biodivine_vector(self.var_container(), self.ac())
        }
    }

    /// Computes the stable models.
    /// Returns an [Iterator][std::iter::Iterator] which contains all the stable models.
    pub fn stable<'a, 'b>(&'a self) -> impl Iterator<Item = Vec<Term>> + 'b
    where
        'a: 'b,
    {
        TwoValuedInterpretationsIterator::from_bdd(&self.grounded_internal(&self.ac)).filter(
            |terms| {
                let reduction_list = self
                    .vars
                    .iter()
                    .enumerate()
                    .filter_map(|(idx, elem)| {
                        if terms[idx].is_truth_value() && !terms[idx].is_true() {
                            Some((*elem, false))
                        } else {
                            None
                        }
                    })
                    .collect::<Vec<(biodivine_lib_bdd::BddVariable, bool)>>();
                let reduct = self
                    .ac
                    .iter()
                    .map(|ac| ac.restrict(&reduction_list))
                    .collect::<Vec<_>>();
                let grounded = self.grounded_internal(&reduct);
                terms
                    .iter()
                    .zip(grounded.iter())
                    .all(|(left, right)| left.cmp_information(right))
            },
        )
    }

    /// Computes the stable models.
    /// This variant returns all stable models and utilises a rewrite of the ADF as one big conjunction of equalities (`if and only if`).
    pub fn stable_bdd_representation(&self) -> Vec<Vec<Term>> {
        let smc = self.stable_model_candidates();
        log::debug!("[Start] checking for stability");
        smc.into_iter()
            .filter(|terms| {
                let reduction_list = self
                    .vars
                    .iter()
                    .enumerate()
                    .filter_map(|(idx, elem)| {
                        if terms[idx].is_truth_value() && !terms[idx].is_true() {
                            Some((*elem, false))
                        } else {
                            None
                        }
                    })
                    .collect::<Vec<(biodivine_lib_bdd::BddVariable, bool)>>();
                let reduct = self
                    .ac
                    .iter()
                    .map(|ac| ac.restrict(&reduction_list))
                    .collect::<Vec<_>>();
                let grounded = self.grounded_internal(&reduct);
                terms
                    .iter()
                    .zip(grounded.iter())
                    .all(|(left, right)| left.cmp_information(right))
            })
            .collect::<Vec<Vec<Term>>>()
    }

    pub(crate) fn stable_model_candidates(&self) -> Vec<Vec<Term>> {
        let internal_rewriting: Bdd;
        let sr: &Bdd;
        if self.has_stm_rewriting() {
            sr = self.rewrite.as_ref().expect("self.rewrite has to exist, because it has been checked just before calling this reference");
        } else {
            internal_rewriting = self.stable_representation();
            sr = &internal_rewriting;
        }
        log::debug!("[Start] construct stable model candidates");
        sr.sat_valuations()
            .map(|valuation| {
                self.vars
                    .iter()
                    .map(|var| {
                        if valuation.value(*var) {
                            Term::TOP
                        } else {
                            Term::BOT
                        }
                    })
                    .collect::<Vec<Term>>()
            })
            .collect::<Vec<Vec<Term>>>()
    }

    /// compute the stable representation
    fn stable_representation(&self) -> Bdd {
        log::debug!("[Start] stable representation rewriting");
        self.ac.iter().enumerate().fold(
            self.varset.eval_expression(
                &biodivine_lib_bdd::boolean_expression::BooleanExpression::Const(true),
            ),
            |acc, (idx, formula)| {
                acc.and(
                    &formula.iff(
                        &self.varset.eval_expression(
                            &biodivine_lib_bdd::boolean_expression::BooleanExpression::Variable(
                                self.ordering
                                    .name(crate::datatypes::Var(idx))
                                    .expect("Variable should exist"),
                            ),
                        ),
                    ),
                )
            },
        )
    }

    /// Creates a [PrintableInterpretation] for output purposes.
    pub fn print_interpretation<'a, 'b>(
        &'a self,
        interpretation: &'b [Term],
    ) -> PrintableInterpretation<'b>
    where
        'a: 'b,
    {
        PrintableInterpretation::new(interpretation, &self.ordering)
    }

    /// Creates a [PrintDictionary] for output purposes.
    pub fn print_dictionary(&self) -> PrintDictionary {
        PrintDictionary::new(&self.ordering)
    }
}

/// Provides ADF-Specific operations on truth valuations.
pub trait AdfOperations {
    /// Returns `true` if the roBDD is either valid or unsatisfiable.
    fn is_truth_value(&self) -> bool;

    /// Compares whether the information between two given roBDDs are the same.
    fn cmp_information(&self, other: &Self) -> bool;
}

impl AdfOperations for Bdd {
    fn is_truth_value(&self) -> bool {
        self.is_false() || self.is_true()
    }

    fn cmp_information(&self, other: &Self) -> bool {
        self.is_truth_value() == other.is_truth_value() && self.is_true() == other.is_true()
    }
}

/// Implementations of the restrict-operations on roBDDs.
pub trait BddRestrict {
    /// Provides an implementation of the restrict-operation on roBDDs for one variable.
    fn var_restrict(&self, variable: biodivine_lib_bdd::BddVariable, value: bool) -> Self;
    /// Provides an implementation of the restrict-operation on a set of variables.
    fn restrict(&self, variables: &[(biodivine_lib_bdd::BddVariable, bool)]) -> Self;
}

impl BddRestrict for Bdd {
    fn var_restrict(&self, variable: biodivine_lib_bdd::BddVariable, value: bool) -> Bdd {
        self.var_select(variable, value).var_exists(variable)
    }

    fn restrict(&self, variables: &[(biodivine_lib_bdd::BddVariable, bool)]) -> Bdd {
        let mut variablelist: Vec<biodivine_lib_bdd::BddVariable> = Vec::new();
        variables
            .iter()
            .for_each(|(var, _val)| variablelist.push(*var));
        self.select(variables).exists(&variablelist)
    }
}

impl ThreeValuedInterpretationsIterator {
    fn from_bdd(bdd: &[Bdd]) -> Self {
        let terms = bdd.iter().map(|value| value.into()).collect::<Vec<_>>();
        Self::new(&terms)
    }
}

impl TwoValuedInterpretationsIterator {
    fn from_bdd(bdd: &[Bdd]) -> Self {
        let terms = bdd.iter().map(|value| value.into()).collect::<Vec<_>>();
        Self::new(&terms)
    }
}

#[cfg(test)]
mod test {
    use super::*;
    use test_log::test;

    use biodivine_lib_bdd::*;

    #[test]
    fn biodivine_internals() {
        let mut builder = BddVariableSetBuilder::new();
        let [_a, _b, _c] = builder.make(&["a", "b", "c"]);
        let variables: BddVariableSet = builder.build();

        let a = variables.eval_expression_string("a");
        let b = variables.eval_expression_string("b");
        let c = variables.eval_expression_string("c");
        let d = variables.eval_expression_string("a & b & c");
        let e = variables.eval_expression_string("a ^ b");
        let t = variables.eval_expression(&boolean_expression::BooleanExpression::Const(true));
        let f = variables.eval_expression(&boolean_expression::BooleanExpression::Const(false));

        println!("{:?}", a.to_string());
        println!("{:?}", a.to_bytes());
        println!("{:?}", b.to_string());
        println!("{:?}", c.to_string());
        println!("{:?}", d.to_string());
        println!("{:?}", e.to_string());
        println!("{:?}", e.to_bytes());
        println!("{:?}", t.to_string());
        println!("{:?}", f.to_string());
    }
    #[test]
    fn grounded() {
        let parser = AdfParser::default();
        parser.parse()("s(a).s(b).s(c).s(d).ac(a,c(v)).ac(b,b).ac(c,and(a,b)).ac(d,neg(b)).\ns(e).ac(e,and(b,or(neg(b),c(f)))).s(f).\n\nac(f,xor(a,e)).")
            .unwrap();
        let adf = Adf::from_parser(&parser);

        let xor = adf.ac[5].clone();

        assert!(xor
            .var_restrict(adf.vars[4], false)
            .var_restrict(adf.vars[0], true)
            .var_restrict(adf.vars[1], true)
            .var_restrict(adf.vars[2], false)
            .is_true());
        let result = adf.grounded();

        assert_eq!(
            result,
            vec![
                Term::TOP,
                Term::UND,
                Term::UND,
                Term::UND,
                Term::BOT,
                Term::TOP
            ]
        );
        assert_eq!(
            format!("{}", adf.print_interpretation(&result)),
            "T(a) u(b) u(c) u(d) F(e) T(f) \n"
        );

        let parser = AdfParser::default();
        parser.parse()(
            "s(a).s(b).s(c).s(d).s(e).ac(a,c(v)).ac(b,a).ac(c,b).ac(d,neg(c)).ac(e,and(a,d)).",
        )
        .unwrap();
        let adf = Adf::from_parser(&parser);
        let result = adf.grounded();
        assert_eq!(
            result,
            vec![Term::TOP, Term::TOP, Term::TOP, Term::BOT, Term::BOT]
        );
    }

    #[test]
    fn grounded_eq_naive() {
        let parser = AdfParser::default();
        parser.parse()("s(a).s(b).s(c).s(d).ac(a,c(v)).ac(b,b).ac(c,and(a,b)).ac(d,neg(b)).\ns(e).ac(e,and(b,or(neg(b),c(f)))).s(f).\n\nac(f,xor(a,e)).")
            .unwrap();
        let adf = Adf::from_parser(&parser);

        let adf_grd = adf.grounded();
        let adf_hybrid_true = adf.hybrid_step_opt(true).grounded();
        let adf_hybrid_false = adf.hybrid_step_opt(false).grounded();
        adf_grd
            .iter()
            .zip(adf_hybrid_false.iter())
            .zip(adf_hybrid_true.iter())
            .for_each(|((a, b), c)| {
                assert!(a.compare_inf(b));
                assert!(b.compare_inf(c));
            });
    }

    #[test]
    fn complete() {
        let parser = AdfParser::default();
        parser.parse()("s(a).s(b).s(c).s(d).ac(a,c(v)).ac(b,b).ac(c,and(a,b)).ac(d,neg(b)).\ns(e).ac(e,and(b,or(neg(b),c(f)))).s(f).\n\nac(f,xor(a,e)).")
            .unwrap();
        let adf = Adf::from_parser(&parser);

        assert_eq!(
            adf.complete().next(),
            Some(vec![
                Term::TOP,
                Term::UND,
                Term::UND,
                Term::UND,
                Term::BOT,
                Term::TOP
            ])
        );

        assert_eq!(
            adf.complete().collect::<Vec<_>>(),
            [
                [
                    Term::TOP,
                    Term::UND,
                    Term::UND,
                    Term::UND,
                    Term::BOT,
                    Term::TOP
                ],
                [
                    Term::TOP,
                    Term::TOP,
                    Term::TOP,
                    Term::BOT,
                    Term::BOT,
                    Term::TOP
                ],
                [
                    Term::TOP,
                    Term::BOT,
                    Term::BOT,
                    Term::TOP,
                    Term::BOT,
                    Term::TOP
                ]
            ]
        );
    }

    #[test]
    fn complete2() {
        let parser = AdfParser::default();
        parser.parse()("s(a).s(b).s(c).s(d).ac(a,c(v)).ac(b,b).ac(c,and(a,b)).ac(d,neg(b)).")
            .unwrap();
        let adf = Adf::from_parser(&parser);
        assert_eq!(
            adf.complete().collect::<Vec<_>>(),
            [
                [Term::TOP, Term::UND, Term::UND, Term::UND],
                [Term::TOP, Term::TOP, Term::TOP, Term::BOT],
                [Term::TOP, Term::BOT, Term::BOT, Term::TOP]
            ]
        );
    }

    #[test]
    fn stable() {
        let parser = AdfParser::default();
        parser.parse()("s(a).s(b).s(c).s(d).ac(a,c(v)).ac(b,b).ac(c,and(a,b)).ac(d,neg(b)).\ns(e).ac(e,and(b,or(neg(b),c(f)))).s(f).\n\nac(f,xor(a,e)).")
            .unwrap();
        let adf = Adf::from_parser(&parser);

        let mut stable = adf.stable();

        assert_eq!(
            stable.next(),
            Some(vec![
                Term::TOP,
                Term::BOT,
                Term::BOT,
                Term::TOP,
                Term::BOT,
                Term::TOP
            ])
        );
        assert_eq!(stable.next(), None);

        let parser = AdfParser::default();
        parser.parse()("s(a).s(b).ac(a,neg(b)).ac(b,neg(a)).").unwrap();
        let adf = Adf::from_parser(&parser);

        let mut stable = adf.stable();
        assert_eq!(stable.next(), Some(vec![Term::BOT, Term::TOP]));
        assert_eq!(stable.next(), Some(vec![Term::TOP, Term::BOT]));
        assert_eq!(stable.next(), None);

        let parser = AdfParser::default();
        parser.parse()("s(a).s(b).ac(a,b).ac(b,a).").unwrap();
        let adf = Adf::from_parser(&parser);

        assert_eq!(adf.stable().next(), Some(vec![Term::BOT, Term::BOT]));

        for model in adf.stable() {
            let printer = adf.print_dictionary();
            assert_eq!(
                format!("{}", adf.print_interpretation(&model)),
                format!("{}", printer.print_interpretation(&model))
            );
        }

        let parser = AdfParser::default();
        parser.parse()("s(a).s(b).ac(a,neg(a)).ac(b,a).").unwrap();
        let adf = Adf::from_parser(&parser);
        assert_eq!(adf.stable().next(), None);
    }

    #[test]
    fn stable_version2() {
        let parser = AdfParser::default();
        parser.parse()("s(a).s(b).s(c).s(d).ac(a,c(v)).ac(b,b).ac(c,and(a,b)).ac(d,neg(b)).\ns(e).ac(e,and(b,or(neg(b),c(f)))).s(f).\n\nac(f,xor(a,e)).")
            .unwrap();
        let adf = Adf::from_parser(&parser);

        let mut stable_naive: Vec<Vec<Term>> = adf.stable().collect();
        let mut stable_v2 = adf.stable_bdd_representation();
        let adf2 = Adf::from_parser_with_stm_rewrite(&parser);
        let mut stable_v3 = adf2.stable_bdd_representation();
        stable_naive.sort();
        stable_v2.sort();
        stable_v3.sort();

        assert_eq!(stable_naive, stable_v2);
        assert_eq!(stable_v2, stable_v3);
    }
}