Public interface for a predicate-extractor class. More...
#include <PredicateExtractor.h>
Related Functions | |
(Note that these are not member functions.) | |
| void | PredicateExtractor_compute_preds (PredicateExtractor_ptr self, node_ptr expr) |
| The function computes and collects the predicates of a given expression. | |
| void | PredicateExtractor_compute_preds_from_hierarchy (PredicateExtractor_ptr self, FlatHierarchy_ptr fh) |
| This function applies PredicateExtractor_compute_preds to every element of an hierarchy. | |
| PredicateExtractor_ptr | PredicateExtractor_create (SymbTable_ptr st, boolean use_approx) |
| The constructor creates a predicate-extractor. | |
| void | PredicateExtractor_destroy (PredicateExtractor_ptr self) |
| Class PredicateExtractor destructor. | |
| Set_t | PredicateExtractor_get_all_clusters (const PredicateExtractor_ptr self) |
| Returns the set of clusters for all so far collected predicates. | |
| Set_t | PredicateExtractor_get_all_preds (const PredicateExtractor_ptr self) |
| Returns the set of predicates computed so far. | |
| Set_t | PredicateExtractor_get_preds_of_a_cluster (const PredicateExtractor_ptr self, Set_t cluster) |
| Returns a set of predicates responsible for a given cluster. | |
| Set_t | PredicateExtractor_get_var_cluster (const PredicateExtractor_ptr self, node_ptr var) |
| Given a fully resolved var name the function returns a cluster the variable belongs to. | |
| void | PredicateExtractor_print (const PredicateExtractor_ptr self, FILE *stream, boolean printPredicates, boolean printClusters) |
| The function prints out the predicates collected so far and clusters computed. | |
Public interface for a predicate-extractor class.
Thus if normalization of the input expression is not required it is more efficient to use this class.
A few definitions:
predicate -- is a boolean expression which have only scalar (not boolean) subexpressions. See PredicateNormaliser.h for more info about predicate normalization.
cluster -- is a set of variables met in one predicates. If a variable is met in 2 different predicates then their clusters are united in one. Implicitly, clusters divide predicates in groups, i.e. every group is a set of predicates that caused this cluster.
Note that from the definitions both clusters and predicates can be only over scalar (not boolean) variables.
This class allows computation of only predicate or both predicates and cluster.
Initially, I ([AT]) tried to implement an option to compute clusters only without predicates but this did not work. The problem is that 1) it is necessary to memoize the results, 2) clusters may disappear during computation (i.e. be merged with others). Because of 1) it is necessary to hash expr->clusters-in-it. Because of 2) it is necessary to hash cluster->expr-where-is-came-from and then any merge of clusters may require to update huge number of elements in the both above hashes. Right now a hash expr->predicate-subparts-in-it is created. This allows to get clusters through getting dependencies. Any other solution I through of was of about the same efficiency. Thus I decided to use the most straightforward one.
This is a stand-alone class. This class needs only a type checker -- to get the type of input expression and type check the generated (returned) expressions.
Preicate Extractor class
| void PredicateExtractor_compute_preds | ( | PredicateExtractor_ptr | self, | |
| node_ptr | expr | |||
| ) | [related] |
The function computes and collects the predicates of a given expression.
See PredicateExtractor.h for more info on predicates and clusters.
Note: that normalization of the input expression is not done. Only predicates are computed (the lesser things are done the lesser time/memory is spent). See class PredicateNormaliser if predicate-normalized expressions are required.
To additionally get/compute clusters PredicateExtractor_get_all_clusters can be used.
Input expressions may/may not be expanded/normalized/flattened, whereas the collected predicates are flattened, expanded and created with find_node, in particular all identifiers fully resolved.
WARNING: memoization is done. Providing the same expression a second times does not produce any additional predicate.
Collected clusters/predicates are stored internally and can be obtained with PredicateExtractor_get_all_preds and PredicateExtractor_get_all_clusters.
| void PredicateExtractor_compute_preds_from_hierarchy | ( | PredicateExtractor_ptr | self, | |
| FlatHierarchy_ptr | fh | |||
| ) | [related] |
This function applies PredicateExtractor_compute_preds to every element of an hierarchy.
Note that symbol table in self has to correspond to the hierarchy, i.e. contains all the required symbols
| PredicateExtractor_ptr PredicateExtractor_create | ( | SymbTable_ptr | st, | |
| boolean | use_approx | |||
| ) | [related] |
The constructor creates a predicate-extractor.
See PredicateExtractor.h for more info on predicates and clusters. The parameter 'checker' is a type checker used during predicate extraction and subsequent type checking of generated expressions.
Parameter use_approx can be used to make the extractor give up when dealing with too-large expressions. This is currently used by the heuristics which extract the variable ordering out of the fsm.
NOTE that the type checker remembers the type of checked expressions (free or reuse nodes with care).
| void PredicateExtractor_destroy | ( | PredicateExtractor_ptr | self | ) | [related] |
Class PredicateExtractor destructor.
| Set_t PredicateExtractor_get_all_clusters | ( | const PredicateExtractor_ptr | self | ) | [related] |
Returns the set of clusters for all so far collected predicates.
This function computes and returns clusters for all so far computed predicates.
Returned result is Set_t of Set_t of fully resolved variables. Everything returned belongs to self.
Note that this function perform computation and may take some time (though intermediate results are remembered between calls).
It is possible to get a group of predicates responsible for a given cluster with PredicateExtractor_get_preds_of_a_cluster.
NOTE: subsequent call of PredicateExtractor_compute_preds makes any data returned by this function invalid.
| Set_t PredicateExtractor_get_all_preds | ( | const PredicateExtractor_ptr | self | ) | [related] |
Returns the set of predicates computed so far.
Predicates are fully expanded and resolved expressions created with find_node, i.e. no freeing or modifications are allowed. Returned Set_t belongs to self.
| Set_t PredicateExtractor_get_preds_of_a_cluster | ( | const PredicateExtractor_ptr | self, | |
| Set_t | cluster | |||
| ) | [related] |
Returns a set of predicates responsible for a given cluster.
Given a cluster (Set_t of vars) returned by PredicateExtractor_get_all_clusters this function returns a set of predicates which caused the given cluster.
Returned result is not-empty Set_t of fully expanded/resolved expressions and belongs to self.
NOTE: subsequent call of PredicateExtractor_compute_preds makes any data returned by this function or PredicateExtractor_get_all_clusters invalid.
| Set_t PredicateExtractor_get_var_cluster | ( | const PredicateExtractor_ptr | self, | |
| node_ptr | var | |||
| ) | [related] |
Given a fully resolved var name the function returns a cluster the variable belongs to.
If clusters were not computed before this function triggers the cluster computation.
Returned result is Set_t of fully resolved variables. Everything returned belongs to self.
If a var was not met in any of predicates then NULL is returned. (This is always so for boolean vars since boolean vars cannot be in predicates).
NOTE: subsequent call of PredicateExtractor_compute_preds makes any data returned by this function invalid.
| void PredicateExtractor_print | ( | const PredicateExtractor_ptr | self, | |
| FILE * | stream, | |||
| boolean | printPredicates, | |||
| boolean | printClusters | |||
| ) | [related] |
The function prints out the predicates collected so far and clusters computed.
Options printPredicates and printClusters control what should be printed. At least one of them has to be set up.
If only predicates are printed, then they are printed in the order they were obtained.
Otherwise, clusters are printed and if additionally printPredicates is up then the after every cluster its predicates are printed.
Note that if clusters were not computed so far but asked to be printed, they will be computed.
1.6.1
