The rewrite by antecedent command applies a rewrite rule to an expression or a predicate in a goal.
The expression or predicate is selected first, and the rewrite rule second.
In the expression case, the rewrite rule must be a predicate in one of the following forms.
|pattern = replacement|
|t pattern = replacement|
In the predicate case, the rewrite rule must be a predicate in one of the following forms.
|pattern ( true replacement)|
|pattern ( false replacement)|
|t pattern replacement|
|t pattern ( true replacement)|
|t pattern ( false replacement)|
The rewrite rule must be an antecedent of the expression or predicate, meaning that it must be a conjunct of a surrounding conjunction predicate, or of the left operand of a surrounding implication predicate, or of the | part of a surrounding schema text, or of the part of a surrounding (unique) existential quantification predicate, or of an antecedent predicate of the surrounding goal. The following is an attempt to formalize this, in which r is the rewrite rule appearing as a conjunct of a larger predicate r q (not necessarily the first conjunct), and f is the selected formula appearing within a larger predicate p(f).
|r q p(f)|
|... | r q p(f)|
|ds | p(f) r q|
|1 ds | p(f) r q|
The pattern part of the rewrite rule must match the selected expression or predicate, interpreting the variables declared by any schema text t as jokers for this match.
In the case of a quantified rewrite rule, side-conditions arise from any declarations and | part of the schema text t. These side-conditions must be automatically decidable for the command to be applicable. Currently, the simplification tac command is used as this decision procedure.
The expression or predicate is replaced by the corresponding instantiation of the replacement.
Unlike rewrite by rule and rewrite by section, the rewrite rule must match the entire expression or predicate, and only one rewrite is done.
This command differs from the Leibniz command in the following ways. Rewriting is always done left-to-right, never right-to-left (see commutation). The equality or equivalence used as the rewrite rule can be within a quantified predicate. Only one formula can be rewritten by a use of this command (the side-conditions for multiple rewrites would have been too awkward to manage, but this might change in future). The selections must be within a goal, not any other form of paragraph (because there is always somewhere within a goal to put the side-conditions, again this might change in future).
A trace of the instantiated rewrites can be made to appear in the shell window by invoking cadiz with the option -dw.
"rewrite by antecedent" e p
This example rewrites expression e using predicate p.
"rewrite by antecedent" p1 p2
This example rewrites predicate p1 using predicate p2.