This tutorial shows basic use of the cluster backtracking.
The same example: expanding syntactic expression, is written using two differents ways: ABSTRACT and TREE. It is recommended to start with TREE.
In both case, just type:
se c -clean -o a_nick_name expand_expression
Each of the 2 programs performs syntaxic expansion. For example, putting the line
(Hello + Hi) the ( + great) world!
at input will produce the output:
(1) Hello the world! (2) Hello the great world! (3) Hi the world! (4) Hi the great world!
The grammar of the inputs are:
input ::= alternative; alternative ::= sequence ['+' sequence]...; sequence ::= [term]...; term ::= '(' alternative ')' | "a term";
Exercice: write a program that does the same syntaxic expansion. (trick: reuse the parser of the tutorial)
The tree example uses the most usable class of the backtracking cluster: the class BACKTRACKING.
The class BACKTRACKING make an exploration of the and/or graphes made of BACKTRACKING_NODE instances. Many usefull BACKTRACKING_NODE inheriters are defined and useables.
The syntaxic expression is transformed to a such and/or structure. Alternatives does use BACKTRACKING_NODE_OR_PAIR that make a or between two nodes. Sequences does use BACKTRACKING_NODE_AND_PAIR that make a and between to nodes. In both cases, the order of the nodes is meaningful. Two other predefined nodes are used: the_true_node (of class BACKTRACKING_NODE_TRUE) for empty terms, and, the_false_node (of class BACKTRACKING_NODE_FALSE) for errors of syntax.
To get practical results, the terms are using a local class STRING_NODE that records what to print in a local field. When explored these nodes push the string on the and call the feature continue. Not that covariance is used.
The use of that class is easy: define the features 'get_context' and 'restore_context'. These features will be called during the exploration to save the context before to explore an alternative and to restore it during the backtracking. These feature are defined as returning ANY but please make a covariant redefinition as in exemples, where the context, an INTEGER, records the count of items pushed.
The abstract example does exactly the same that tree except that it uses a more abstract class ABSTRACT_BACKTRACKING.
That class is intended to be used when structures to be explored can not be derived from BACKTRACKING_NODE.