Cleanups to handling ml predicates and substitutions (#4625)

~Blocked on: #4631
~Blocked on: #4630
~Blocked on: #4633

While reviewing and going over
#4621 with @Stevengre , it
became somewhat clear that how we handle turning substitions into ML
predicates is a bit dirty. This attempts to clean this up a bit. Where
potentially breaking changes to API are introduced here, I've checked if
it affects the following repos when I mention "downstream" below:
`evm-semantics kontrol wasm-semantics riscv-semantics mir-semantics`.

In particular:

- The function `CTerm.anti_unify` has a simplification where it reuses a
function from `kast.manip` instead of reimplementing it.
- The functions `CSubst.from_pred` and `CSubst.pred` are added, as
replacements for `Subst.ml_pred`. This is because `Subst.ml_pred`
doesn't have a good way to produce correctly sorted predicates, because
it's in module `kast.inner`.
- `Subst.ml_pred` is removed, and tests are updated to use the new
`CSubst` variant. None of the downstream repositories use
`Subst.ml_pred` directly.
- The new `CSubst.pred` correctly sorts the generated `#Equals` clauses,
defaulting to `K` sort or if a `KDefinition` is supplied using it to do
sort inference. It also provides options for controlling whether we
include the substitution or the constraints in the generated predicate.
- A test is added for a `CSubst.pred` case which caused a bug in the
integration tests dealing with identity substitutions.
- The `CTermSymbolic.implies` function is updated to reuse
`CSubst.from_pred` instead of reimplementing it.
- On the case of duplicate entries, the first is kept and the latter are
made as predicates.

Author: ehildenb

pyk/src/pyk/cterm/cterm.py

@@ -6,11 +6,12 @@
 from typing import TYPE_CHECKING
 
 from ..kast import KInner
-from ..kast.inner import KApply, KRewrite, KToken, Subst, bottom_up
+from ..kast.inner import KApply, KRewrite, KToken, KVariable, Subst, bottom_up
 from ..kast.manip import (
     abstract_term_safely,
     build_claim,
     build_rule,
+    extract_subst,
     flatten_label,
     free_vars,
     ml_pred_to_bool,
@@ -20,9 +21,9 @@
     split_config_and_constraints,
     split_config_from,
 )
-from ..prelude.k import GENERATED_TOP_CELL
+from ..prelude.k import GENERATED_TOP_CELL, K
 from ..prelude.kbool import andBool, orBool
-from ..prelude.ml import is_bottom, is_top, mlAnd, mlBottom, mlEqualsTrue, mlImplies, mlTop
+from ..prelude.ml import is_bottom, is_top, mlAnd, mlBottom, mlEquals, mlEqualsTrue, mlImplies, mlTop
 from ..utils import unique
 
 if TYPE_CHECKING:
@@ -217,17 +218,7 @@ def anti_unify(
             if KToken('true', 'Bool') not in [disjunct_lhs, disjunct_rhs]:
                 new_cterm = new_cterm.add_constraint(mlEqualsTrue(orBool([disjunct_lhs, disjunct_rhs])))
 
-        new_constraints = []
-        fvs = new_cterm.free_vars
-        len_fvs = 0
-        while len_fvs < len(fvs):
-            len_fvs = len(fvs)
-            for constraint in common_constraints:
-                if constraint not in new_constraints:
-                    constraint_fvs = free_vars(constraint)
-                    if any(fv in fvs for fv in constraint_fvs):
-                        new_constraints.append(constraint)
-                        fvs = fvs | constraint_fvs
+        new_constraints = remove_useless_constraints(common_constraints, new_cterm.free_vars)
 
         for constraint in new_constraints:
             new_cterm = new_cterm.add_constraint(constraint)
@@ -341,6 +332,26 @@ def from_dict(dct: dict[str, Any]) -> CSubst:
         constraints = (KInner.from_dict(c) for c in dct['constraints'])
         return CSubst(subst=subst, constraints=constraints)
 
+    @staticmethod
+    def from_pred(pred: KInner) -> CSubst:
+        """Extract from a boolean predicate a CSubst."""
+        subst, pred = extract_subst(pred)
+        return CSubst(subst=subst, constraints=flatten_label('#And', pred))
+
+    def pred(self, sort_with: KDefinition | None = None, subst: bool = True, constraints: bool = True) -> KInner:
+        """Return an ML predicate representing this substitution."""
+        _preds: list[KInner] = []
+        if subst:
+            for k, v in self.subst.minimize().items():
+                sort = K
+                if sort_with is not None:
+                    _sort = sort_with.sort(v)
+                    sort = _sort if _sort is not None else sort
+                _preds.append(mlEquals(KVariable(k, sort=sort), v, arg_sort=sort))
+        if constraints:
+            _preds.extend(self.constraints)
+        return mlAnd(_preds)
+
     @property
     def constraint(self) -> KInner:
         """Return the set of constraints as a single flattened constraint using `mlAnd`."""

pyk/src/pyk/cterm/symbolic.py

@@ -25,7 +25,7 @@
     kore_server,
 )
 from ..prelude.k import GENERATED_TOP_CELL, K_ITEM
-from ..prelude.ml import is_top, mlEquals
+from ..prelude.ml import mlAnd
 
 if TYPE_CHECKING:
     from collections.abc import Iterable, Iterator
@@ -267,19 +267,8 @@ def implies(
             raise ValueError('Received empty predicate for valid implication.')
         ml_subst = self.kore_to_kast(result.substitution)
         ml_pred = self.kore_to_kast(result.predicate)
-        ml_preds = flatten_label('#And', ml_pred)
-        if is_top(ml_subst):
-            csubst = CSubst(subst=Subst({}), constraints=ml_preds)
-            return CTermImplies(csubst, (), None, result.logs)
-        subst_pattern = mlEquals(KVariable('###VAR'), KVariable('###TERM'))
-        _subst: dict[str, KInner] = {}
-        for subst_pred in flatten_label('#And', ml_subst):
-            m = subst_pattern.match(subst_pred)
-            if m is not None and type(m['###VAR']) is KVariable:
-                _subst[m['###VAR'].name] = m['###TERM']
-            else:
-                raise AssertionError(f'Received a non-substitution from implies endpoint: {subst_pred}')
-        csubst = CSubst(subst=Subst(_subst), constraints=ml_preds)
+        ml_subst_pred = mlAnd(flatten_label('#And', ml_subst) + flatten_label('#And', ml_pred))
+        csubst = CSubst.from_pred(ml_subst_pred)
         return CTermImplies(csubst, (), None, result.logs)
 
     def assume_defined(self, cterm: CTerm, module_name: str | None = None) -> CTerm:

pyk/src/pyk/kast/inner.py

@@ -749,20 +749,6 @@ def from_pred(pred: KInner) -> Subst:
                     raise ValueError(f'Invalid substitution predicate: {conjunct}')
         return Subst(subst)
 
-    @property
-    def ml_pred(self) -> KInner:
-        """Turn this `Subst` into a matching logic predicate using `{_#Equals_}` operator."""
-        items = []
-        for k in self:
-            if KVariable(k) != self[k]:
-                items.append(KApply('#Equals', [KVariable(k), self[k]]))
-        if len(items) == 0:
-            return KApply('#Top')
-        ml_term = items[0]
-        for _i in items[1:]:
-            ml_term = KApply('#And', [ml_term, _i])
-        return ml_term
-
     @property
     def pred(self) -> KInner:
         """Turn this `Subst` into a boolean predicate using `_==K_` operator."""

pyk/src/pyk/kcfg/show.py

@@ -469,7 +469,9 @@ def dump(self, cfgid: str, cfg: KCFG, dump_dir: Path, dot: bool = False) -> None
             cover_file = covers_dir / f'config_{cover.source.id}_{cover.target.id}.txt'
             cover_constraint_file = covers_dir / f'constraint_{cover.source.id}_{cover.target.id}.txt'
 
-            subst_equalities = flatten_label('#And', cover.csubst.subst.ml_pred)
+            subst_equalities = flatten_label(
+                '#And', cover.csubst.pred(sort_with=self.kprint.definition, constraints=False)
+            )
 
             if not cover_file.exists():
                 cover_file.write_text('\n'.join(self.kprint.pretty_print(se) for se in subst_equalities))

pyk/src/pyk/kcfg/tui.py

@@ -309,7 +309,12 @@ def _cterm_text(cterm: CTerm) -> tuple[str, str]:
                 term_str, constraint_str = _cterm_text(crewrite)
 
             elif type(self._element) is KCFG.Cover:
-                subst_equalities = map(_boolify, flatten_label('#And', self._element.csubst.subst.ml_pred))
+                subst_equalities = map(
+                    _boolify,
+                    flatten_label(
+                        '#And', self._element.csubst.pred(sort_with=self._kprint.definition, constraints=False)
+                    ),
+                )
                 constraints = map(_boolify, flatten_label('#And', self._element.csubst.constraint))
                 term_str = '\n'.join(self._kprint.pretty_print(se) for se in subst_equalities)
                 constraint_str = '\n'.join(self._kprint.pretty_print(c) for c in constraints)
@@ -320,7 +325,10 @@ def _cterm_text(cterm: CTerm) -> tuple[str, str]:
                     term_strs.append('')
                     term_strs.append(f'  - {shorten_hashes(target_id)}')
                     if len(csubst.subst) > 0:
-                        subst_equalities = map(_boolify, flatten_label('#And', csubst.subst.ml_pred))
+                        subst_equalities = map(
+                            _boolify,
+                            flatten_label('#And', csubst.pred(sort_with=self._kprint.definition, constraints=False)),
+                        )
                         term_strs.extend(f'    {self._kprint.pretty_print(cline)}' for cline in subst_equalities)
                     if len(csubst.constraints) > 0:
                         constraints = map(_boolify, flatten_label('#And', csubst.constraint))

pyk/src/pyk/proof/reachability.py

@@ -487,14 +487,16 @@ def from_spec_modules(
 
         return res
 
-    def path_constraints(self, final_node_id: NodeIdLike) -> KInner:
+    def path_constraints(self, final_node_id: NodeIdLike, sort_with: KDefinition | None = None) -> KInner:
         path = self.shortest_path_to(final_node_id)
         curr_constraint: KInner = mlTop()
         for edge in reversed(path):
             if type(edge) is KCFG.Split:
                 assert len(edge.targets) == 1
                 csubst = edge.splits[edge.targets[0].id]
-                curr_constraint = mlAnd([csubst.subst.minimize().ml_pred, csubst.constraint, curr_constraint])
+                curr_constraint = mlAnd(
+                    [csubst.pred(sort_with=sort_with, constraints=False), csubst.constraint, curr_constraint]
+                )
             if type(edge) is KCFG.Cover:
                 curr_constraint = mlAnd([edge.csubst.constraint, edge.csubst.subst.apply(curr_constraint)])
         return mlAnd(flatten_label('#And', curr_constraint))

pyk/src/tests/unit/kast/test_subst.py

@@ -7,7 +7,6 @@
 
 from pyk.kast.inner import KApply, KLabel, KVariable, Subst
 from pyk.kast.manip import extract_subst
-from pyk.prelude.kbool import TRUE
 from pyk.prelude.kint import INT, intToken
 from pyk.prelude.ml import mlAnd, mlEquals, mlEqualsTrue, mlOr, mlTop
 
@@ -108,25 +107,6 @@ def test_unapply(term: KInner, subst: dict[str, KInner], expected: KInner) -> No
     assert actual == expected
 
 
-ML_PRED_TEST_DATA: Final = (
-    ('empty', Subst({}), KApply('#Top')),
-    ('singleton', Subst({'X': TRUE}), KApply('#Equals', [KVariable('X'), TRUE])),
-    (
-        'double',
-        Subst({'X': TRUE, 'Y': intToken(4)}),
-        KApply(
-            '#And',
-            [KApply('#Equals', [KVariable('X'), TRUE]), KApply('#Equals', [KVariable('Y'), intToken(4)])],
-        ),
-    ),
-)
-
-
-@pytest.mark.parametrize('test_id,subst,pred', ML_PRED_TEST_DATA, ids=[test_id for test_id, *_ in ML_PRED_TEST_DATA])
-def test_ml_pred(test_id: str, subst: Subst, pred: KInner) -> None:
-    assert subst.ml_pred == pred
-
-
 _0 = intToken(0)
 _EQ = KLabel('_==Int_')
 EXTRACT_SUBST_TEST_DATA: Final[tuple[tuple[KInner, dict[str, KInner], KInner], ...]] = (

pyk/src/tests/unit/test_cterm.py

@@ -9,9 +9,10 @@
 from pyk.kast import Atts, KAtt
 from pyk.kast.inner import KApply, KLabel, KRewrite, KSequence, KSort, KVariable, Subst
 from pyk.kast.outer import KClaim
-from pyk.prelude.k import GENERATED_TOP_CELL
+from pyk.prelude.k import GENERATED_TOP_CELL, K
+from pyk.prelude.kbool import TRUE
 from pyk.prelude.kint import INT, intToken
-from pyk.prelude.ml import mlAnd, mlEqualsTrue
+from pyk.prelude.ml import mlAnd, mlEquals, mlEqualsTrue, mlTop
 
 from .utils import a, b, c, f, g, ge_ml, h, k, lt_ml, x, y, z
 
@@ -188,6 +189,28 @@ def test_from_kast(test_id: str, kast: KInner, expected: CTerm) -> None:
     assert cterm == expected
 
 
+ML_PRED_TEST_DATA: Final = (
+    ('empty', CSubst(Subst({})), mlTop()),
+    ('singleton', CSubst(Subst({'X': TRUE})), mlEquals(KVariable('X', sort=K), TRUE, arg_sort=K)),
+    ('identity', CSubst(Subst({'X': KVariable('X')})), mlTop()),
+    (
+        'double',
+        CSubst(Subst({'X': TRUE, 'Y': intToken(4)})),
+        mlAnd(
+            [
+                mlEquals(KVariable('X', sort=K), TRUE, arg_sort=K),
+                mlEquals(KVariable('Y', sort=K), intToken(4), arg_sort=K),
+            ]
+        ),
+    ),
+)
+
+
+@pytest.mark.parametrize('test_id,csubst,pred', ML_PRED_TEST_DATA, ids=[test_id for test_id, *_ in ML_PRED_TEST_DATA])
+def test_ml_pred(test_id: str, csubst: CSubst, pred: KInner) -> None:
+    assert csubst.pred() == pred
+
+
 APPLY_TEST_DATA: Final = (
     (CTerm.top(), CSubst(), CTerm.top()),
     (CTerm.bottom(), CSubst(), CTerm.bottom()),