Models

If a context is satisfiable, Yices can build a model of the context’s assertions. Functions are provided to extract the values of terms in a model. Atomic values (e.g., integer or bitvector constants) can be obtained directly. Non-atomic values—that is, tuples or functions—are represented internally as nodes in a DAG. The API includes functions to explore this DAG and get the values of tuples or functions.

Model Construction

model_t* yices_get_model(context_t* ctx, int32_t keep_subst)

Builds a model from a satisfiable context.

If there’s an error, the function returns NULL. Otherwise, it constructs a model of the assertions in ctx and returns a pointer to this model. The model must be deleted when it is no longer used by calling yices_free_model.

Parameters

  • ctx: context
  • keep_subst: flag to indicates whether the model should include eliminated variables

The context’s status must be either STATUS_SAT or STATUS_UNKNOWN.

When assertions are added to a context, the simplification procedures may eliminate variables by substitution (see yices_context_enable_option). If keep_subst is true (i.e. non-zero), then yices_get_model keeps track of eliminated variables so that their values can be computed. If keep_subst if false (i.e., zero), then the model does not include the eliminated variables.

It is better to set keep_susbt to true. The only reason not to do it is to save the memory and computation cost of storing the eliminated variables. This cost is usually low but there are exceptions.

Error report

Note

The returned model captures a snapshot of the context’s current state. Future operations on ctx (including deleting or resetting ctx) do not change the model.

model_t* yices_model_from_map(uint32_t n, const term_t var[], const term_t map[])

Builds a model from a term-to-term mapping.

Parameters

  • n: number of terms in arrays var and map
  • var: array of n uninterpreted terms
  • map: array of n constant terms

The two arrays var and map define the mapping. Every element of array var must be an uninterpreted term; every element of array map must be a constant term. The constant map[i] specifies the value of var[i] in the resulting model. There must not be duplicates in array var, and the type of term map[i] must be a subtype of var[i]’s type.

This function returns NULL if something goes wrong. It allocates and creates a model otherwise, and returns a pointer to this model. When the model is no longer used, it must be deleted by calling yices_free_model.

Currently, function types are not supported. Every term in map[i] must either be an atomic constant, or a tuple of atomic constants, or a tuple of tuples of constants, and so forth. The term var[i] cannot have a function type.

Error report

  • if a term in var or map is not valid:

    – error code: INVALID_TERM

    – term1 := the invalid term

  • if map[i]’s type is not a subtype of var[i]’s type:

    – error code: TYPE_MISMATCH

    – term1 := map[i]

    – type1 := type of var[i] (expected type)

  • if var[i] is not an uninterpreted term:

    – error code: MDL_UNINT_REQUIRED

  • if map[i] is not a constant:

    – error code: MDL_CONSTANT_REQUIRED

  • if var contains duplicate terms:

    – error code: MDL_DUPLICATE_VAR

  • if map[i] has function type or has a function subterm:

    – error code: MDL_FTYPE_NOT_ALLOWED

  • if the construction fails for some other reason:

    – error code: MDL_CONSTRUCTION_FAILED

void yices_model_collect_defined_terms(model_t *mdl, term_vector_t *v)

Collect all the uninterpreted terms that have a value in mdl and store them in vector v.

The set of terms is returned as follows:

  • v->size is the number of uninterpreted terms
  • v->data[0 … v->size - 1] contains the terms

Here is an example use of this function:

yices_assert_formula(ctx, f);
if (yices_check(ctx, ...) == STATUS_SAT) {
   term_vector_t v;
   model_t *m = yices_get_model(ctx, true);
   yices_init_term_vector(&v);
   yices_model_collect_defined_terms(m, &v);
   ....
}

At the end of this code fragment, vector v typically contains all the uninterpreted terms that occur in f. In some rare cases, terms may be eliminated during preprocessing and assertion simplifications. These terms will not be defined in model m and will not be stored in vector v.

void yices_free_model(model_t* mdl)

Deletes a model.

This function deletes model mdl, which must be a pointer returned by either yices_get_model or yices_model_from_map.

Note

If this function is not called, Yices will automatically free the model on a call to yices_exit or yices_reset.

Value of a Term in a Model

The following functions compute the value of a term in a model. For terms of atomic types, the value can be extracted directly. Non-atomic values (i.e., tuples or functions) can be extracted by traversing the model’s DAG.

Many functions in this section attempt to evaluate a term t in a model. If the value can’t be computed, these functions return -1 and report one of the following errors:

Other error codes are possible, depending on the function.

Atomic Values

int32_t yices_get_bool_value(model_t *mdl, term_t t, int32_t *val)

Value of a Boolean term.

This function computes the value of term t in mdl and stores the result in *val as either 0 (for false) or 1 (for true).

If t’s value can’t be computed or if t is not a Boolean term, the function leaves *val unchanged, updates the error report, and returns -1. Otherwise, it returns 0.

Error report

  • if t is not a Boolean term:

    – error code: TYPE_MISMATCH

    – term1 := t

    – type1 := Bool type

See also yices_formula_true_in_model and yices_formulas_true_in_model.

int32_t yices_get_int32_value(model_t *mdl, term_t t, int32_t *val)

Value of an integer (32 bits).

This function computes the value of t in model mdl and stores it in *val. It fails and returns -1 if t’s value can’t be computed, if it is not an integer, or if it is too large or too small to be represented as a signed 32bit integer.

Error report

  • If t is not an arithmetic term:

    – error code: ARITHTERM_REQUIRED

    – term1 := t

  • If t’s value is not an integer or does not fit in 32 bits:

    – error code: EVAL_OVERFLOW

int32_t yices_get_int64_value(model_t *mdl, term_t t, int64_t *val)

Value as an integer (64 bits).

This function is similar to yices_get_int32_value but it succeeds if t’s value can be represented as a signed 64bit integer.

int32_t yices_get_rational32_value(model_t *mdl, term_t t, int32_t *num, uint32_t *den)

Value as a rational (32 bits).

This function computes the value of t in mdl and returns it as a rational number. The numerator is stored in *num and the denominator is stored in *den. If t’s value can’t be computed or does not fit in this representation, the function returns -1 and leaves both *num and *den unchanged.

Error report

  • If t is not an arithmetic term:

    – error code: ARITHTERM_REQUIRED

    – term1 := t

  • If t’s value can’t be represented as a 32bit numerator divided by a 32bit numerator:

    – error code: EVAL_OVERFLOW

int32_t yices_get_rational64_value(model_t *mdl, term_t t, int64_t *num, uint64_t *den)

Value as a rational (64 bits).

This function is similar to yices_get_rational32_value except that it uses a 64bit numerator and a 64bit denominator.

int32_t yices_get_double_value(model_t *mdl, term_t t, double *val)

Value as a floating-point number.

This function computes the value of t in mdl, converts it to a floating-point number, and stores the result in variable *val. It fails (and returns -1) if t’s value can’t be computed or if t is not an arithmetic term. It returns 0 otherwise.

Error report
int32_t yices_get_mpz_value(model_t *mdl, term_t t, mpz_t val)

Value as a GMP integer.

This function stores t’s value in the GMP integer val. The variable val must be initialized (see the GMP documentation). This function fails if t’s value can’t be computed or if it’s not an integer.

Error report

Note

This function is not declared unless you include gmp.h before yices.h in your code, as in:

#include <gmp.h>
#include <yices.h>
int32_t yices_get_mpq_value(model_t *mdl, term_t t, mpq_t val)

Value as a GMP rational.

This function stores t’s value in the GMP rational val. The variable val must be initialized (see the GMP documentation). This function fails if t’s value can’t be computed or if t is not an arithmetic term.

Error report

Note

Like yices_get_mpz_value, this function is declared if header file gmp.h is included before yices.h.

int32_t yices_get_algebraic_number_value(model_t *mdl, term_t t, lp_algebraic_number_t *a)

Value as an algebraic number.

This function stores t’s value in the structure *a. The type lp_algebraic_number_t is defined in the libpoly library, and represents an algebraic number.

This function fails if t’s value cannot be computed, or if it is not an algebraic number. It will also fail if the Yices library was compiled without support for MCSAT (cf. MCSAT Support).

Error report

  • If MCSAT is not supported:

    – error code: EVAL_NOT_SUPPORTED

  • If t is not an arithmetic term:

    – error code: ARITH_TERM_REQUIRED

  • If t is an arithmetic term and has a rational value (i.e., not an algebraic number):

    – error code: EVAL_CONVERSION_FAILED

Note

This function is not declared unless you include the libpoly header algebraic_number.h before yices.h in your code:

#include <poly/algebraic_number.h>
#include <yices.h>

Example

The following code fragment prints the value of a term t, using libpoly’s functions. This works if t has an algebraic value in model mdl:

static void show_algebraic_value(model_t *mdl, term_t t) {
  lp_algebraic_number_t n;
  int32_t code;

  code = yices_get_algebraic_number_value(mdl, t, &n);
  if (code < 0) {
    yices_print_error(stderr);
  } else {
    lp_algebraic_number_print(&n, stdout);
    fflush(stdout);
    lp_algebraic_number_destruct(&n);
  }
}
int32_t yices_get_bv_value(model_t *mdl, term_t t, int32_t val[])

Value of a bitvector term.

This function computes t’s value in mdl and stores it in array val. The value is returned using the little-endian convention: the least significant bit is stored in val[0] and the most significant bit is stored in val[n-1] (where n is the number of bits). The array val must be large enough to store these n bits.

The number of bits in t can be found by calling yices_term_bitsize.

The function fails if t’s value can’t be computed or if t is not a bitvector term. It leaves val unchanged and returns -1 in this case. Otherwise, it returns 0.

Error report

int32_t yices_get_scalar_value(model_t *mdl, term_t t, int32_t *val)

Value of a scalar or uninterpreted term.

The value of t is returned as an integer index in *val. This index has the same meaning as in function yices_constant:

  • If t has type tau and tau is a scalar type of cardinality n then *val is an integer between 0 and n-1.
  • If t has an uninterpreted type, then the returned index can be any non-negative integer.

The returned index is a unique identifier. If two terms t1 and t2 are not equal in the model mdl then their values are distinct integer indices.

The function returns -1 if there’s an error or 0 otherwise.

Error report

  • If t does not have a scalar or uninterpreted type:

    – error code: SCALAR_TERM_REQUIRED

    – term1 := t

Value of Formulas

int32_t yices_formula_true_in_model(model_t* mdl, term_t f)

Checks whether a formula is true in a model.

This function checks whether f is true in mdl. It returns 1 if f is true in the model, 0 if f is false in the model, or -1 if f is not a Boolean term or if it can’t be evaluated.

This function reports the same error codes as yices_get_bool_value.

int32_t yices_formulas_true_in_model(model_t* mdl, uint32_t n, const term_t f[])

Checks whether an array of formulas is true in a model.

This function checks whether the formulas f[0] to f[n-1] are all true in mdl. It returns 1 if all f[i]*s are true, 0 if an *f[i] is false, or -1 of an f[i] is not Boolean or can’t be evaluated.

This function reports the same errors as yices_get_bool_value.

It is more efficient to call this function that to call yices_formula_true_in_model n times.

General Values

The preceding functions are sufficient to extract atomic values from a model, but the value of a term may be a tuple or a function. To deal with the general case, Yices provides functions to access values by exploring the model DAG. Function yices_get_value evaluates a term and returns a node descriptor from which the term value can be constructed.

Within a model, each node has an integer identifier and a tag that specifies the node’s type. All DAG-exploration functions store this information in records of type yval_t defined as follows:

typedef struct yval_s {
  int32_t node_id;
  yval_tag_t node_tag;
} yval_t;

Distinct nodes in a model have distinct node_ids.

The type yval_tag_t lists the possible tags of a node. Leaf nodes represent atomic values. They can have the following tags:

The following tags are used for non-leaf nodes:

A node of tag YVAL_TUPLE has n children nodes, each representing the value of a tuple component. For example, tuple (1, true, 0b00) is represented by a node with three children. The first child is an atomic node with tag YVAL_RATIONAL and value 1; the second child has tag YVAL_BOOL and value true; and the third child has tag YVAL_BV and value 0b00.

A node of tag YVAL_FUNCTION represents a function. In a model, all functions have a simple form. They are defined by a finite enumeration of mappings that specify the function value at distinct points in its domain, and a default value for all other points. Each mapping in the enumeration is represented by a node of tag YVAL_MAPPING. For a function f with n arguments, a mapping is a tuple of n+1 nodes (written [ k1 … kn ↦ v ]). The n nodes k1 … kn define a point in f’s domain, and the value of f at this point is represented by node v.

For example, consider the function f such that:

f(0, 0) = 0
f(3, 1) = 1
f(x, y) = -2 in all other cases.

Then the node that represents this function has three children. Two children are mapping nodes for [ 0, 0 ↦ 0 ] and [ 3, 1 ↦ 1 ], and the third child is an atomic node representing the default value -2.

To support partially-defined values, the following tag is also defined:

A node with this tag may be returned as the default value of a function f. In such a case, the model stores a finite list of mappings for f but does not assign a value to f outside the points listed in the mappings.

The DAG-exploration API includes a function that returns the node for a term value, functions that provide information about a node, functions that return the value of leaf nodes, and functions that return the children of non-leaf nodes.

int32_t yices_get_value(model_t *mdl, term_t t, yval_t *val)

Value as a node reference.

This function evaluates t in model mdl. If this succeeds, it stores a node descriptor for t’s value in the record *val:

  • val->node_id contains the node id
  • val->node_tag contains the tag

The function returns 0 in this case.

If t’s evaluation fails, the function returns -1 and leaves *val unchanged.

int32_t yices_val_is_int32(model_t *mdl, const yval_t *v)

Checks whether a node’s value is a signed 32bit integer.

This function returns true (i.e., 1) if v->node_tag is YVAL_RATIONAL and the node value is an integer that can be represented using 32 bits. It returns false (i.e., 0) if the tag is not YVAL_RATIONAL or if the value does not fit into a signed 32bit integer.

See also yices_val_get_int32.

int32_t yices_val_is_int64(model_t *mdl, const yval_t *v)

Checks whether a node’s value is a signed 64bit integer.

This function is similar to yices_val_is_int32 except that it returns true if the node has tag YVAL_RATIONAL and its value is an integer that can be represented using 64 bits.

See also yices_val_get_int64.

int32_t yices_val_is_rational32(model_t *mdl, const yval_t *v)

Checks whether a node’s value is a 32bit rational.

This function returns true if the node described by *v has tag YVAL_RATIONAL and if its value can be written as a 32bit signed numerator divided by a 32bit unsigned denominator. If returns false otherwise.

See also yices_val_get_rational32.

int32_t yices_val_is_rational64(model_t *mdl, const yval_t *v)

Checks whether a node’s value is a 64bit rational.

This function returns true if the node described by *v has tag YVAL_RATIONAL and if its value can be written as a 64bit signed numerator divided by a 64bit unsigned denominator. If returns false otherwise.

See also yices_val_get_rational64.

int32_t yices_val_is_integer(model_t *mdl, const yval_t *v)

Checks whether a node’s value is an integer.

This function returns true if the node described by *v has tag YVAL_RATIONAL and the node’s value is an integer. It returns false otherwise.

See also yices_val_get_mpz.

uint32_t yices_val_bitsize(model_t *mdl, const yval_t *v)

Number of bits in a bitvector constant node.

If the node described by *v has tag YVAL_BV, then this function returns the number of bits in the node’s value. Otherwise, it returns 0.

See also yices_val_get_bv.

uint32_t yices_val_tuple_arity(model_t *mdl, const yval_t *v)

Number of components in a tuple node.

If the node described by *v has tag YVAL_TUPLE, then this function returns the number of components in the tuple (i.e., the number of children of the node). It returns 0 otherwise.

See also yices_val_expand_tuple.

uint32_t yices_val_function_arity(model_t *mdl, const yval_t *v)

Arity of a function node.

If the node described by *v has tag YVAL_FUNCTION, then this function returns the function’s arity (i.e., number of parameters to the function). If the node has another tag, this function returns 0.

uint32_t yices_val_mapping_arity(model_t *mdl, const yval_t *v)

Arity of a mapping node.

If the node described by *v has tag YVAL_MAPPING, then this function returns the node’s arity (i.e, if the node stores a mapping [ k1 … kn ↦ v ], then the function returns n). Otherwise, it returns 0.

See also yices_val_expand_mapping.

type_t yices_val_function_type(model_t *mdl, const yval_t *v)

Type of a function node.

If the node described by *v has tag YVAL_FUNCTION, then this function returns the type of the node. This is a function type. Otherwise, the function returns -1 (i.e., NULL_TYPE).

Error report

int32_t yices_val_get_bool(model_t *mdl, const yval_t *v, int32_t *val)

Value of a Boolean node.

This function checks whether node v has tag YVAL_BOOL. If so, it stores the node’s value into *val and returns 0. Otherwise, it returns -1 and leaves *val unchanged.

Error report

int32_t yices_val_get_int32(model_t *mdl, const yval_t *v, int32_t *val)

Value of an integer node (32 bits).

This function checks whether node v has tag YVAL_RATIONAL and whether its value can be represented as a 32bit signed integer. If so, it stores the node’s value into *val and returns 0. Otherwise, it returns -1 and leaves *val unchanged.

Error report

int32_t yices_val_get_int64(model_t *mdl, const yval_t *v, int64_t *val)

Value of an integer node (64 bits).

This function is similar to yices_val_get_int32 except that it succeeds if the node’s value fits in a 64bit signed integer.

int32_t yices_val_get_rational32(model_t *mdl, const yval_t *v, int32_t *num, uint32_t *den)

Value of a rational node (32 bits).

This function checks whether node v has tag YVAL_RATIONAL and whether its value can be represented as a 32bit numerator divided by a 32bit denominator. If so, it stores the numerator in *num and the denominator in *den and returns 0. Otherwise, it returns -1 and leaves *num and *den unchanged.

Error report

int32_t yices_val_get_rational64(model_t *mdl, const yval_t *v, int64_t *num, uint64_t *den)

Value of a rational node (64 bits).

This function is similar to yices_val_get_rational32 except that it succeeds if the node’s value can be represented as a 64bit numerator divided by a 64bit denominator.

int32_t yices_val_get_double(model_t *mdl, const yval_t *v, double *val)

Node value as a floating-point number.

This function checks whether node *v has tag YVAL_RATIONAL or YVAL_ALGEBRAIC. If so it converts the node’s value to a double-precision floating point number and stores it in *val. The function returns 0 in this case, If *v has a different tag, the function returns -1 and leaves *val unchanged.

Error report

int32_t yices_val_get_mpz(model_t *mdl, const yval_t *v, mpz_t val)

Node value as a GMP integer.

This function checks whether node *v has tag YVAL_RATIONAL and if its value is an integer. If so, it copies the node value in the GMP integer val and returns 0. Otherwise, it leaves val unchanged and returns -1.

The variable val must be initialized (see the GMP documentation).

Error report

Note

This function is not declared unless you include gmp.h before yices.h in your code.

int32_t yices_val_get_mpq(model_t *mdl, const yval_t *v, mpq_t val)

Node value as a GMP rational.

This function checks whether node v has tag YVAL_RATIONAL. If so, it copies the node’s value in the GMP rational val and returns 0. Otherwise, it leaves val unchanged and returns -1.

The variable val must be initialized (see the GMP documentation).

Error report

Note

This function is not declared unless you include gmp.h before yices.h in your code.

int32_t yices_val_get_algebraic_number(model_t *mdl, const yval_t *v, lp_algebraic_number_t *a)

Node value as an algebraic number.

This function checks whether node v has tag YVAL_ALGEBRAIC. If so, it copies the node’s value in the structure *a and returns 0. Otherwise, it leaves the structure unchanged and returns -1.

This function fails if the Yices library was compiled without support for MCSAT (cf. MCSAT Support).

Error report

Note

This function is not declared unless you include the libpoly header algebraic_number.h before yices.h in your code:

#include <poly/algebraic_number.h>
#include <yices.h>
int32_t yices_val_get_bv(model_t *mdl, const yval_t *v, int32_t val[])

Value of a bitvector node.

This function checks whether node v has tag YVAL_BV. If so, it copies the node value into array val and returns 0. Otherwise, it leaves val unchanged and returns -1.

The array val must be large enough to store n integers, where n is the number of bits in the bitvector value. The number of bits can be found by calling function yices_val_bitsize. The value is returned using the little-endian convention: the least-significant bit is stored in val[0] and the most significant bit is in val[n-1].

Error report

int32_t yices_val_get_scalar(model_t *mdl, const yval_t *v, int32_t *val, type_t *tau)

Value of a scalar node.

This function checks whether v has tag YVAL_SCALAR. If so, it copies the node’s value (an integer index) into *val and the node’s type into *tau. The function returns 0 in this case. It returns -1 and leaves *val and *tau unchanged otherwise.

Error report

int32_t yices_val_expand_tuple(model_t *mdl, const yval_t *v, yval_t child[])

Components of a tuple node.

If node v has tag YVAL_TUPLE, then this function copies the node’s children into array child. The array must be large enough to store all the children descriptors. The number of children is given by function yices_val_tuple_arity. The first child is stored in child[0] and the last child is in child[n-1] where n is the node’s arity.

The function returns 0 if v’s tag is YVAL_TUPLE or -1 otherwise.

Error report

int32_t yices_val_expand_function(model_t *mdl, const yval_t *f, yval_t *def, yval_vector_t *v)

Components of a function node.

If node f has tag YVAL_FUNCTION, this function extracts the node’s components:

  • the default node for f is stored in *def
  • the set of mappings for f is copied in vector v

The vector must be initialized by calling yices_init_yval_vector. The function sets v->size to the number of mappings for f and it copies the mapping nodes into v->data[0]v->data[v->size-1]. All the nodes in the vector have tag YVAL_MAPPING and have the same arity as the function node f.

The function returns 0 if there’s no error, or -1 if f’s tag is not YVAL_FUNCTION.

Error report

int32_t yices_val_expand_mapping(model_t *mdl, const yval_t *m, yval_t tup[], yval_t *val)

Components of a mapping node.

This function checks whether m has tag YVAL_MAPPING. If so, it copies the node’s components in array tup and in val. The function returns 0 in this case. It returns -1 and leaves tup and val unchanged otherwise.

For a mapping [ k1 … kn ↦ v ], the nodes k1 … kn are copied in tup[0] to tup[n-1], and the node v is copied into *val. Array tup must be large enough to store the n node descriptors. The arity n can be found by calling yices_val_mapping_arity (or using yices_val_function_arity).

Error report

Values as Terms

term_t yices_get_value_as_term(model_t *mdl, term_t t)

Value as a constant term.

This function evaluates term t in model mdl and returns t’s value as a constant term.

For a term t of atomic types, this function is equivalent to extracting t’s value then converting it to a constant:

  • If t is a Boolean term, the returned value is either true or false, as constructed by yices_true or yices_false.
  • If t is an arithmetic term, the returned value is constant term as constructed with yices_mpq.
  • If t has uninterpreted or scalar type, the returned value is a constant term of that type as built by function yices_constant.
  • If t is a bitvector term, the returned value is a bitvector constant as built with yices_bvconst_from_array.

If t is a tuple, then the function returns a constant tuple.

Function types are not supported. The function fails and returns NULL_TERM if t has a function type. It also returns NULL_TERM if t’s value can’t be computed.

Error report

  • If t’s value can’t be converted to a constant term:

    – error code: EVAL_CONVERSION_FAILED

    This happens if t has a function type or if t has tuple type and some subterm of t has a function type.

int32_t yices_term_array_value(model_t *mdl, uint32_t n, const term_t a[], term_t b[])

Values as constant terms.

This function computes the values of terms a[0 … n-1], converts these values to constant terms, and stores the results in array b.

Parameters

  • mdl: model
  • n: size of arrays a and b
  • a: array of n terms
  • b: array to store the result as n constant terms.

This function has the same behavior as calling yices_get_value_as_term n times. It returns 0 if all values can be computed, or -1 if there’s an error. The possible error codes are the same as for yices_get_value_as_term.

Supports

Given a term t and a model M, one may be intersted in the set of uninterpreted terms on which the value assigned to t depends. We call this set a support of t in M. For example, if t is the term (ite (> x 0) (+ x z) y) then and x has positive value in M then the value of t in M does not depend on y. In this case, the support of t in M is the set { x, z }. If we have a different model R and x has a negative value in R then the support of t in R is the set { x, y }. In R, the value of t does not depend on z’s value.

The following two functions compute the support of a term or a set of terms in a model.

int32_t yices_model_term_support(model_t *mdl, term_t t, term_vector_t *v)

Get the support of t in mdl

The function computes a support for t in mdl and store the result in vector v.

Parameters

  • mdl: model
  • t: term
  • v: term vector

Vector v must be initialized by yices_init_term_vector. The support set is returned in v as follows:

  • v->size stores the number of uninterpreted terms in the support set
  • v0->data[0]v->data[n-1] store n distinct uninterpreted terms (where n = v->size).

Error report

  • if t is not a valid term:

    – error code: INVALID_TERM

    – term1 := t

int32_t yices_model_term_array_support(model_t *mdl, uint32_t n, const term_t a[], term_vector_t *v)

Get the support of n terms in mdl

Parameters

  • mdl: model
  • n: size of array a
  • a: array of n terms
  • v: term vector

This is similar to the yices_model_term_support but it computes a support for n terms stored in array a. The set is returned in vector v.

Implicants

int32_t yices_implicant_for_formula(model_t *mdl, term_t t, term_vector_t *v)

Computes an implicant.

This function constructs an implicant of t based on model mdl. The term t must be a Boolean term that’s true in mdl. The implicant is a list of Boolean terms a1 … an such that

  • ai is a literal (atom or negation of an atom)
  • ai is true in mdl
  • the conjunction (a1 ∧ … ∧ an) implies t

The implicant is returned in vector v, which must be initialized by yices_init_term_vector:

  • v->size stores the number of literals in the implicant (i.e., n).
  • v->data[0]v->data[n-1] store the n literals

The function returns 0 if the implicant can be computed. Otherwise, it returns -1 and resets v to the empty vector (i.e., it sets v->size to 0).

Error report

  • if t is not a valid term

    – error code: INVALID_TERM

    – term1 := t

  • if t is not a Boolean term

    – error code: TYPE_MISMATCH

    – term1 := t

    – type1 := Bool type

  • if t is false in the model

    – error code: EVAL_NO_IMPLICANT

Other errors are possible if t or a subterm of t can’t be evaluated. These are the same error codes as for any term-evaluation function.

int32_t yices_implicant_for_formulas(model_t *mdl, uint32_t n, const term_t a[], term_vector_t *v)

Implicant for an array of formulas.

This function constructs an implicant for the conjunction of formulas a[0] ∧ … ∧ a[n-1]. The result is stored in vector v as explained for yices_implicant_for_formula. This function has returns 0 if the implicant can be constructed or -1 otherwise. It has the same behavior and reports the same errors as yices_implicant_for_formula.

Model Generalization

Given a model for a formula F(X, Y), the following two functions compute a formula G(X) that generalizes the model. The formula G(X) satisfies two properties:

  1. G(X) is true in the model
  2. G(X) implies (exists y F(x, y))

Yices supports two generalization methods:

  • Generalization by substitution: This is the simplest method. It eliminates the variables Y by replacing them by their values in the model.
  • Generalization by projection: This first computes an implicant of formula F(X, Y) then eliminates the Y variables from this implicant by projection. The projection is a cheap form of quantifier elimination. It is a hybrid a Fourier-Motzkin elimination and virtual term substitution.

The two generalization functions take a parameter that specifies the generalization method. This parameter takes a value of type yices_gen_mode_t:

  • YICES_GEN_BY_SUBST selects generalization by substitution;
  • YICES_GEN_BY_PROJ selects generalization by projection;
  • YICES_GEN_DEFAULT automatically chooses the generalization method based on the type of variables to eliminate. If any variable to eliminate is an arithmetic variable, then generalization by projection is used. Otherwise, the default is generalization by substitution.
int32_t yices_generalize_model(model_t *mdl, term_t t, uint32_t nelims, const term_t elim[], yices_gen_mode_t mode, term_vector_t *v)

Model generalization for a single formula.

This function computes a generalization of mdl for formula t.

Parameters

  • mdl: model
  • t: Boolean term that’s true in mdl
  • nelims: number of variables to eliminate
  • elim: variables to eliminate
  • mode: generalization method
  • v: term vector to store the result

Every term in array elim must be an uninterpreted term (cf. yices_new_uninterpreted_term), of type Boolean, Real, or bitvector.

The generalization formula G(X) is returned in vector v. The vector must be initialized using yices_init_term_vector. G(X) is the conjunction of all formulas in v:

  • v->size: number of formulas returned
  • v->data[0]v->data[v->size-1] contain the formulas themselves

If mode is YICES_GEN_BY_PROJ then every formula in v is guaranteed to be a literal.

The function returns 0 if the generalization succeeds or -1 if there’s an error.

int32_t yices_generalize_model_array(model_t *mdl, uint32_t n, const term_t a[], uint32_t nelims, const term_t elim[], yices_gen_mode_t mode, term_vector_t *v)

Model generalization for an array of formulas.

Parameters

  • mdl: model
  • n: number of formulas in array a
  • a: array of n Boolean term that are true in mdl
  • nelims: number of variables to eliminate
  • elim: variables to eliminate
  • mode: generalization method
  • v: term vector to store the result

This function is equivalent to calling yices_generalize_model with argument (a[0] ∧ … ∧ a[n-1]).