Skip to content

Making interactive persistent using proof scripts

Mattias Ulbrich ulbrich@kit.edu, 2015

Experimental feature: Proof scripts are currently only visible in the GUI if KeY is launched with the --experimental option. Concrete syntax is subject to change.

Every KeY user knows the pain of manually redoing an interactive proof over and over again even though only a tiny bit of the program or specification have changed. NOW you can use proof scripts which will alleviate your worries with repeating proofs.

Proof scripts are textual representations of rule applications, settings changes and macro invocations. They are notated in linear order. The target of a script command is usually the first open goal in the proof tree, i.e., the first reached when traversing the proof tree (not necessarily the first in the Goal pane in the GUI).

Syntax of proof scripts

Proof commands start with an identifier followed by optional arguments:

command argName="argument" "argument without name" ;

Every command is terminated with a semicolon. There are named arguments in the form argName="argument" and unnamed argument without name. Single '...' and double quotes "..." can both be used. If "argument" is an identifier itself, the quotes can optionally be dropped, like in

rule andRight ;

The character # begins a comment which extends to the end of the line. Currently, there are no multi-line comments in proof scripts. Proof scripts are case sensitive.

How can scripts be loaded

  1. In the GUI: Open a proof obligation. Currently proof scripts work only reliably if applied to the root of a fresh proof obligation (no rule applications, prune steps). The open the context menu (sequent or proof tree), choose "Strategy macros" and then "Run proof script ...". In the file dialog that appears, choose the script file to load. The script is automatically applied. You can follow the process in a logging window.

  2. From within key input files: At the very end of key files INSTEAD of a \proof object, you can now attach a proof script directly in double quotes following the keyword \proofScript. An very simple example reads:

\predicates {a; b;}
\problem { a & b -> a }

\proofScript "
   rule 'impRight';
   rule 'andLeft';
   rule 'close';"

Since the script is no enclosed in "...", only single quotes '...' can be used in the script. You can also load a script from an external file by writing

\proofScript "script 'some_filename.script';"

Commands in scripts

This list of available script commands is subject to change and to extension. If you write your own script commands (s. below), please add an explanation to this list here. The list is sorted alphabetically.

  • auto

Apply the automatic KeY strategy on the current goal. Optinally you can specify the number of steps to run.

auto steps=30000; # run at most 30000 steps automatically.
  • cut

Performs a cut and thus splits the sequent into two goals. The unnamed argument is the formula to cut with

cut "value1 = value2";
  • exit

Terminate the script prematurely at that point. Used mainly for debug purposes.

exit;
  • instantiate

    Quantifier instantiation is a task that often occurs. Instead of specifying the entire formula, it suffices here to name the variable that is to be instantiated. If that is not unique, the number of the occurrence of that quantified variable can be specified as well.

    Instantiate the third instantiateable formula whose bound variable is called x with the value 42

    instantiate var="x" occ="3" with="42"
    
    The quantified formula can also be specified if wanted. This here for the antecedent.
    instantiate formula="\forall int x; f(x) = 42" with="23"
    
    Existentially quantified variables can be instantiated if they occur on the succedent side.
    instantiate formula="\exists int x; f(x) = 42" with="23"
    
    Instantiate x and hide the quantified formula:
    instantiate hide var=x value="x_0"
    
    * leave

    Mark the currently active goal as non-interactive (the orange hand symbol in the GUI). It is then excluded from further analysis by scripts. This is good for debugging unfinished proof scripts.

  • macro

    Invoke a macro on the current goal. The names of available macros include:

    • autopilot -- full autopilot
    • autopilot-prep -- preparation only autopilot
    • split-prop -- propositional expansion w/ splits
    • nosplit-prop -- propositional expansion w/o splits
    • simp-upd -- update simplification
    • simp-heap -- heap simplification
    macro autopilot-prep;
    

    Note

    Future version may drop the macro keyword and allow macro invocations directly.

  • rule

    Apply a single rule onto the current sequent. As unnamed argument add the name of the taclet to be applied. If the taclet matches only once on the entire sequent, the rule is applied. If it matches more than once you need to specify more. In that case you can first specify the sequence formula and then the number of the occurrence in the formula or the specific subterm via the 'on' keyword. If a rule has schema variables which must be instantiated manually, such instantiations can be provided as arguments. A schema variable named sv can be instantiated by setting the argument sv="..." or by setting inst_sv="..." (the latter works also for conflict cases like inst_occ="...").

    If there is only one matching spot on the sequent

    rule andRight;
    
    Changes sequent a=b ==> c=d to b=a ==> c=d. rule eqSymm; alone would have been ambiguous.
    rule eqSymm formula="a=b";
    
    Changes sequent a=b->c=d ==>" to "a=b->d=c ==>. The occurrence number is needed since there are two possible applications on the formula.
    rule eqSymm formula="a=b->c=d" occ=2;
    

    changes sequent "a=b->c=d ==>" to "a=b->d=c ==>". same as above, but using the option to specify a subterm instead of an occurrence number.

    rule eqSymm formula="a=b->c=d" on="c=d";
    

    Almost the same as cut "x>y":

    rule cut cutFormula="x > y";
    

  • script

    Invoke another script which resides in an external file.

    script '/path/to/other/file.script';
    

  • select

    Unlike most other commands, this command does not change the proof but chooses the goal on which the next step operates. Currently you can specify a formula. The goal is chosen such that the formula appears (toplevel) on the sequent (antecedent or succedent). You can limit the search to antecedent or succedent.

    Search for the formula anywhere

    select formula="{ x:=1 }y < x";
    

    Search only the succedent for the formula:

    select succedent formula="wellFormed(someHeap)";
    

  • smt

    Invoke an external SMT solver. That solver must be adequately configured outside the script mechanism. By default, Z3 is invoked, but that can be chosen.

    Invoke Z3

    smt;
    

    A comma separated list of solvers can be specified.

    smt solver="Z3,yices";
    

  • tryclose

    Unlike other commands this command operates on ALL open goals and effectively applies the "try provable goals below" macro to all of them. A number of steps can optionally be given.

    tryclose;
    

    Spend 2000 steps on each open goal

    tryclose steps=2000;
    

Write your on proof script commands

to be done. Contact Mattias, if you are interested.