Isabelle_DOF/src/DOF/RegExpInterface.thy

242 lines
11 KiB
Plaintext
Executable File

(*************************************************************************
* Copyright (C)
* 2019 The University of Exeter
* 2018-2019 The University of Paris-Saclay
* 2018 The University of Sheffield
*
* License:
* This program can be redistributed and/or modified under the terms
* of the 2-clause BSD-style license.
*
* SPDX-License-Identifier: BSD-2-Clause
*************************************************************************)
chapter\<open>The High-Level Interface to the Automata-Library\<close>
theory RegExpInterface
imports "Functional-Automata.Execute"
keywords
"reflect_ML_exports" :: thy_decl
begin
text\<open> The implementation of the monitoring concept follows the following design decisions:
\<^enum> We re-use generated code from the AFP submissions @{theory "Regular-Sets.Regular_Set"} and
@{theory "Functional-Automata.Automata"}, converted by the code-generator into executable SML code
(ports to future Isabelle versions should just reuse future versions of these)
\<^enum> Monitor-Expressions are regular expressions (in some adapted syntax)
over Document Class identifiers; they denote the language of all possible document object
instances belonging to these classes
\<^enum> Instead of expanding the sub-class relation (and building the product automaton of all
monitor expressions), we convert the monitor expressions into automata over class-id's
executed in parallel, in order to avoid blowup.
\<^enum> For efficiency reasons, the class-ids were internally abstracted to integers; the
encoding table is called environment \<^verbatim>\<open>env\<close>.
\<^enum> For reusability reasons, we did NOT abstract the internal state representation in the
deterministic automata construction (lists of lists of bits - sic !) by replacing them
by unique keys via a suitable coding-table; rather, we opted for keeping the automatas small
(no products, no subclass-expansion).
\<close>
section\<open>Monitor Syntax over RegExp - constructs\<close>
notation Star ("\<lbrace>(_)\<rbrace>\<^sup>*" [0]100)
notation Plus (infixr "||" 55)
notation Times (infixr "~~" 60)
notation Atom ("\<lfloor>_\<rfloor>" 65)
definition rep1 :: "'a rexp \<Rightarrow> 'a rexp" ("\<lbrace>(_)\<rbrace>\<^sup>+")
where "\<lbrace>A\<rbrace>\<^sup>+ \<equiv> A ~~ \<lbrace>A\<rbrace>\<^sup>*"
definition opt :: "'a rexp \<Rightarrow> 'a rexp" ("\<lbrakk>(_)\<rbrakk>")
where "\<lbrakk>A\<rbrakk> \<equiv> A || One"
value "Star (Conc(Alt (Atom(CHR ''a'')) (Atom(CHR ''b''))) (Atom(CHR ''c'')))"
text\<open>or better equivalently:\<close>
value "\<lbrace>(\<lfloor>CHR ''a''\<rfloor> || \<lfloor>CHR ''b''\<rfloor>) ~~ \<lfloor>CHR ''c''\<rfloor>\<rbrace>\<^sup>*"
section\<open>Some Standard and Derived Semantics\<close>
text\<open> This is just a reminder - already defined in @{theory "Regular-Sets.Regular_Exp"}
as @{term lang}.\<close>
text\<open>In the following, we give a semantics for our regular expressions, which so far have
just been a term language (i.e. abstract syntax). The semantics is a ``denotational semantics'',
i.e. we give a direct meaning for regular expressions in some universe of ``denotations''.
This universe of denotations is in our concrete case:\<close>
text\<open>Now the denotational semantics for regular expression can be defined on a post-card:\<close>
fun L :: "'a rexp => 'a lang"
where L_Emp : "L Zero = {}"
|L_One: "L One = {[]}"
|L_Atom: "L (\<lfloor>a\<rfloor>) = {[a]}"
|L_Un: "L (el || er) = (L el) \<union> (L er)"
|L_Conc: "L (el ~~ er) = {xs@ys | xs ys. xs \<in> L el \<and> ys \<in> L er}"
|L_Star: "L (Star e) = Regular_Set.star(L e)"
text\<open>A more useful definition is the sub-language - definition\<close>
fun L\<^sub>s\<^sub>u\<^sub>b :: "'a::order rexp => 'a lang"
where L\<^sub>s\<^sub>u\<^sub>b_Emp: "L\<^sub>s\<^sub>u\<^sub>b Zero = {}"
|L\<^sub>s\<^sub>u\<^sub>b_One: "L\<^sub>s\<^sub>u\<^sub>b One = {[]}"
|L\<^sub>s\<^sub>u\<^sub>b_Atom: "L\<^sub>s\<^sub>u\<^sub>b (\<lfloor>a\<rfloor>) = {z . \<forall>x. x \<le> a \<and> z=[x]}"
|L\<^sub>s\<^sub>u\<^sub>b_Un: "L\<^sub>s\<^sub>u\<^sub>b (el || er) = (L\<^sub>s\<^sub>u\<^sub>b el) \<union> (L\<^sub>s\<^sub>u\<^sub>b er)"
|L\<^sub>s\<^sub>u\<^sub>b_Conc: "L\<^sub>s\<^sub>u\<^sub>b (el ~~ er) = {xs@ys | xs ys. xs \<in> L\<^sub>s\<^sub>u\<^sub>b el \<and> ys \<in> L\<^sub>s\<^sub>u\<^sub>b er}"
|L\<^sub>s\<^sub>u\<^sub>b_Star: "L\<^sub>s\<^sub>u\<^sub>b (Star e) = Regular_Set.star(L\<^sub>s\<^sub>u\<^sub>b e)"
definition XX where "XX = (rexp2na example_expression)"
definition YY where "YY = na2da(rexp2na example_expression)"
(* reminder from execute *)
value "NA.accepts (rexp2na example_expression) [0,1,1,0,0,1]"
value "DA.accepts (na2da (rexp2na example_expression)) [0,1,1,0,0,1]"
section\<open>HOL - Adaptions and Export to SML\<close>
definition enabled :: "('a,'\<sigma> set)da \<Rightarrow> '\<sigma> set \<Rightarrow> 'a list \<Rightarrow> 'a list"
where "enabled A \<sigma> = filter (\<lambda>x. next A x \<sigma> \<noteq> {}) "
definition zero where "zero = (0::nat)"
definition one where "one = (1::nat)"
export_code zero one Suc Int.nat nat_of_integer int_of_integer (* for debugging *)
example_expression (* for debugging *)
Zero One Atom Plus Times Star (* regexp abstract syntax *)
rexp2na na2da enabled (* low-level automata interface *)
NA.accepts DA.accepts
in SML module_name RegExpChecker
subsection\<open>Infrastructure for Reflecting exported SML code\<close>
ML\<open>
fun reflect_local_ML_exports args trans = let
fun eval_ML_context ctxt = let
fun is_sml_file f = String.isSuffix ".ML" (Path.implode (#path f))
val files = (map (Generated_Files.check_files_in (Context.proof_of ctxt)) args)
val ml_files = filter is_sml_file (map #1 (maps Generated_Files.get_files_in files))
val ml_content = map (fn f => Syntax.read_input (#content f)) ml_files
fun eval ml_content = fold (fn sml => (ML_Context.exec
(fn () => ML_Context.eval_source ML_Compiler.flags sml)))
ml_content
in
(eval ml_content #> Local_Theory.propagate_ml_env) ctxt
end
in
Toplevel.generic_theory eval_ML_context trans
end
val files_in_theory =
(Parse.underscore >> K [] || Scan.repeat1 Parse.path_binding) --
Scan.option (\<^keyword>\<open>(\<close> |-- Parse.!!! (\<^keyword>\<open>in\<close>
|-- Parse.theory_name --| \<^keyword>\<open>)\<close>));
val _ =
Outer_Syntax.command \<^command_keyword>\<open>reflect_ML_exports\<close>
"evaluate generated Standard ML files"
(Parse.and_list1 files_in_theory >> (fn args => reflect_local_ML_exports args));
\<close>
reflect_ML_exports _
section\<open>The Abstract Interface For Monitor Expressions\<close>
text\<open>Here comes the hic : The reflection of the HOL-Automata module into an SML module
with an abstract interface hiding some generation artefacts like the internal states
of the deterministic automata ...\<close>
ML\<open>
structure RegExpInterface : sig
type automaton
type env
type cid
val alphabet : term list -> env
val ext_alphabet: env -> term list -> env
val conv : term -> env -> int RegExpChecker.rexp (* for debugging *)
val rexp_term2da: env -> term -> automaton
val enabled : automaton -> env -> cid list
val next : automaton -> env -> cid -> automaton
val final : automaton -> bool
val accepts : automaton -> env -> cid list -> bool
end
=
struct
local open RegExpChecker in
type state = bool list RegExpChecker.set
type env = string list
type cid = string
type automaton = state * ((Int.int -> state -> state) * (state -> bool))
val add_atom = fold_aterms (fn Const(c as(_,Type(@{type_name "rexp"},_)))=> insert (op=) c |_=>I);
fun alphabet termS = rev(map fst (fold add_atom termS []));
fun ext_alphabet env termS =
let val res = rev(map fst (fold add_atom termS [])) @ env;
val _ = if has_duplicates (op=) res
then error("reject and accept alphabets must be disjoint!")
else ()
in res end;
fun conv (Const(@{const_name "Regular_Exp.rexp.Zero"},_)) _ = Zero
|conv (Const(@{const_name "Regular_Exp.rexp.One"},_)) _ = Onea
|conv (Const(@{const_name "Regular_Exp.rexp.Times"},_) $ X $ Y) env = Times(conv X env, conv Y env)
|conv (Const(@{const_name "Regular_Exp.rexp.Plus"},_) $ X $ Y) env = Plus(conv X env, conv Y env)
|conv (Const(@{const_name "Regular_Exp.rexp.Star"},_) $ X) env = Star(conv X env)
|conv (Const(@{const_name "RegExpInterface.opt"},_) $ X) env = Plus(conv X env, Onea)
|conv (Const(@{const_name "RegExpInterface.rep1"},_) $ X) env = Times(conv X env, Star(conv X env))
|conv (Const (s, Type(@{type_name "rexp"},_))) env =
let val n = find_index (fn x => x = s) env
val _ = if n<0 then error"conversion error of regexp." else ()
in Atom(n) end
|conv S _ = error("conversion error of regexp:" ^ (Syntax.string_of_term (@{context})S))
val eq_int = {equal = curry(op =) : Int.int -> Int.int -> bool};
val eq_bool_list = {equal = curry(op =) : bool list -> bool list -> bool};
fun rexp_term2da env term = let val rexp = conv term env;
val nda = RegExpChecker.rexp2na eq_int rexp;
val da = RegExpChecker.na2da eq_bool_list nda;
in da end;
(* here comes the main interface of the module:
- "enabled" gives the part of the alphabet "env" for which the automatan does not
go into a final state
- next provides an automata transformation that produces an automaton that
recognizes the rest of a word after a *)
fun enabled (da as (state,(_,_))) env =
let val inds = RegExpChecker.enabled da state (0 upto (length env - 1))
in map (fn i => nth env i) inds end
fun next (current_state, (step,fin)) env a =
let val index = find_index (fn x => x = a) env
in if index < 0 then error"undefined id for monitor"
else (step index current_state,(step,fin))
end
fun final (current_state, (_,fin)) = fin current_state
fun accepts da env word = let fun index a = find_index (fn x => x = a) env
val indexL = map index word
val _ = if forall (fn x => x >= 0) indexL then ()
else error"undefined id for monitor"
in RegExpChecker.accepts da indexL end
end; (* local *)
end (* struct *)
\<close>
no_notation Atom ("\<lfloor>_\<rfloor>")
end