Theory Automatic_Refinement.Autoref_Translate
theory Autoref_Translate
imports Autoref_Fix_Rel
begin
subsection ‹Definitions›
subsubsection ‹APP and ABS Rules›
lemma autoref_ABS:
"⟦ ⋀x x'. (x,x')∈Ra ⟹ (c x, a x')∈Rr ⟧ ⟹ (c, λ'x. a x)∈Ra→Rr"
by auto
lemma autoref_APP:
"⟦ (c,a)∈Ra→Rr; (x,x')∈Ra ⟧ ⟹ (c$x, a $ x')∈Rr"
by (auto dest: fun_relD)
lemma autoref_beta:
assumes "(c,a x)∈R"
shows "(c,(λ'x. a x)$x)∈R"
using assms by auto
lemmas dflt_trans_rules = autoref_beta autoref_ABS autoref_APP
subsubsection ‹Side Conditions›
text ‹
Rules can have prefer and defer side-conditions. Prefer conditions must
be solvable in order for the rule to apply, and defer conditions must
hold after the rule has been applied and the recursive translations have been
performed. Thus, prefer-conditions typically restrict on the abstract
expression, while defer conditions restrict the translated expression.
In order to solve the actual side conditions, we use the
‹Tagged_Solver›-infrastructure. The solvers are applied after
the ‹PREFER›/‹DEFER› tag has been removed.
›
text ‹
Tag to remove internal stuff from term.
Before a prefer/defer side condition is evaluated, all terms inside these
tags are purged from autoref-specific annotations, i.e., operator-annotations,
relator annotations, and tagged applications.
›
definition [simp, autoref_tag_defs]: "REMOVE_INTERNAL x ≡ x"
text ‹Useful abbreviation to require some property that is not related
to teh refinement framework›
abbreviation "PREFER nt Φ ≡ PREFER_tag (nt (REMOVE_INTERNAL Φ))"
abbreviation "DEFER nt Φ ≡ DEFER_tag (nt (REMOVE_INTERNAL Φ))"
definition [simp]: "REMOVE_INTERNAL_EQ a b ≡ a=b"
lemma REMOVE_INTERNAL_EQI: "REMOVE_INTERNAL_EQ a a" by simp
lemma autoref_REMOVE_INTERNAL_EQ:
assumes "(c,a)∈R"
assumes "REMOVE_INTERNAL_EQ c c'"
shows "(c',a)∈R"
using assms by simp
ML ‹
signature AUTOREF_TACTICALS = sig
val is_prefer_cond: int -> thm -> bool
val is_defer_cond: int -> thm -> bool
val REPEAT_INTERVAL: tactic' -> itactic
val COND'': (int -> thm -> bool) -> tactic' -> tactic' -> tactic'
val REPEAT_ON_SUBGOAL: tactic' -> tactic'
val IF_SOLVED: tactic' -> tactic' -> tactic' -> tactic'
end
structure Autoref_Tacticals :AUTOREF_TACTICALS = struct
fun is_prefer_cond i st =
case Logic.concl_of_goal (Thm.prop_of st) i of
@{mpat "Trueprop (PREFER_tag _)"} => true
| _ => false
fun is_defer_cond i st =
case Logic.concl_of_goal (Thm.prop_of st) i of
@{mpat "Trueprop (DEFER_tag _)"} => true
| _ => false
fun REPEAT_INTERVAL tac i j st = let
val n = Thm.nprems_of st - (j - i)
fun r st = (
COND
(has_fewer_prems n)
(all_tac)
(tac i THEN_ELSE (r,all_tac))
) st
in
r st
end
fun COND'' cond tac1 tac2 = IF_EXGOAL (fn i => fn st =>
if cond i st then tac1 i st else tac2 i st
)
fun REPEAT_ON_SUBGOAL tac i = REPEAT_INTERVAL tac i i
fun IF_SOLVED tac tac1 tac2 i st = let
val n = Thm.nprems_of st
in
(tac i THEN COND (has_fewer_prems n) (tac1 i) (tac2 i)) st
end
end
signature AUTOREF_TRANSLATE = sig
type trans_net = (int * thm) Net.net
val add_post_rule: thm -> Context.generic -> Context.generic
val delete_post_rule: thm -> Context.generic -> Context.generic
val get_post_rules: Proof.context -> thm list
val compute_trans_net: Autoref_Fix_Rel.thm_pairs -> Proof.context
-> trans_net
val side_tac: Proof.context -> tactic'
val side_dbg_tac: Proof.context -> tactic'
val trans_step_only_tac: Proof.context -> tactic'
val trans_dbg_step_tac: Proof.context -> tactic'
val trans_step_tac: Proof.context -> tactic'
val trans_tac: Proof.context -> itactic
val trans_analyze: Proof.context -> int -> int -> thm -> bool
val trans_pretty_failure: Proof.context -> int -> int -> thm -> Pretty.T
val trans_phase: Autoref_Phases.phase
val setup: theory -> theory
end
structure Autoref_Translate :AUTOREF_TRANSLATE = struct
type trans_net = (int * thm) Net.net
structure autoref_post_simps = Named_Thms (
val name = @{binding autoref_post_simps}
val description = "Refinement Framework: " ^
"Automatic refinement post simplification rules"
);
val add_post_rule = autoref_post_simps.add_thm
val delete_post_rule = autoref_post_simps.del_thm
val get_post_rules = autoref_post_simps.get
fun compute_trans_net thm_pairs _ =
thm_pairs
|> map #2
|> (fn thms => thms @ @{thms dflt_trans_rules})
|> Tactic.build_net
structure trans_netD = Autoref_Data (
type T = trans_net
fun compute ctxt =
compute_trans_net (Autoref_Fix_Rel.thm_pairsD_get ctxt) ctxt
val prereq = [Autoref_Fix_Rel.thm_pairsD_init]
)
val REMOVE_INTERNAL_conv =
Conv.top_sweep_conv (
fn ctxt => fn ct => (case Thm.term_of ct of
@{mpat "REMOVE_INTERNAL _"} =>
Conv.rewr_conv @{thm REMOVE_INTERNAL_def}
then_conv Autoref_Tagging.untag_conv ctxt
| _ => Conv.no_conv) ct
)
fun REMOVE_INTERNAL_tac ctxt = CONVERSION (REMOVE_INTERNAL_conv ctxt)
fun side_tac ctxt =
resolve_tac ctxt @{thms SIDEI}
THEN' REMOVE_INTERNAL_tac ctxt
THEN' Tagged_Solver.solve_tac ctxt
fun side_dbg_tac ctxt = let
val ctxt = Config.put Tagged_Solver.cfg_keep true ctxt
in
resolve_tac ctxt @{thms SIDEI}
THEN' REMOVE_INTERNAL_tac ctxt
THEN' TRY o Tagged_Solver.solve_tac ctxt
end
local
open Autoref_Tacticals
fun trans_rule_tac ctxt net = resolve_from_net_tac ctxt net
THEN_ALL_NEW (TRY o match_tac ctxt [@{thm PRIO_TAGI}])
in
fun trans_step_only_tac ctxt = let
val net = trans_netD.get ctxt
in
(
COND'' is_defer_cond
(K no_tac)
(assume_tac ctxt ORELSE' trans_rule_tac ctxt net)
)
end
fun trans_dbg_step_tac ctxt = let
val net = trans_netD.get ctxt
val s_tac = side_tac ctxt
in DETERM o (
COND'' is_defer_cond
(SOLVED' s_tac)
(
assume_tac ctxt
ORELSE'
(trans_rule_tac ctxt net
THEN_ALL_NEW_FWD
COND'' is_prefer_cond
(TRY o DETERM o SOLVED' s_tac)
(K all_tac)
)
)
)
end
fun trans_step_tac ctxt = let
val net = trans_netD.get ctxt
val s_tac = side_tac ctxt
in DETERM o (
COND'' is_defer_cond
(SOLVED' s_tac)
(
assume_tac ctxt
ORELSE'
(trans_rule_tac ctxt net
THEN_ALL_NEW_FWD
COND'' is_prefer_cond
(DETERM o SOLVED' s_tac)
(K all_tac)
)
)
)
end
fun trans_tac ctxt = let
val ss = put_simpset HOL_basic_ss ctxt
addsimps @{thms APP_def PROTECT_def ANNOT_def}
addsimps get_post_rules ctxt
val trans_opt_tac =
resolve_tac ctxt @{thms autoref_REMOVE_INTERNAL_EQ}
THEN'
IF_SOLVED (REPEAT_ON_SUBGOAL (trans_step_tac ctxt))
(simp_tac ss THEN' resolve_tac ctxt @{thms REMOVE_INTERNAL_EQI})
(K all_tac)
in
Seq.INTERVAL trans_opt_tac
end
end
fun trans_analyze _ i j _ = j < i
fun trans_pretty_failure ctxt i j st =
if j < i then Pretty.str "No failure"
else let
val goal = Logic.get_goal (Thm.prop_of st) i
val concl = strip_all_body goal |> Logic.strip_imp_concl
val msg = case concl of
@{mpat "Trueprop (DEFER_tag _)"}
=> Pretty.str "Could not solve defered side condition"
| @{mpat "Trueprop ((_,_)∈_)"}
=> Pretty.str "Could not refine subterm"
| _ => Pretty.str "Internal: Unexpected goal"
in
Pretty.block [
msg,
Pretty.fbrk,
Syntax.pretty_term ctxt goal
]
end
val trans_phase = {
init = trans_netD.init,
tac = trans_tac,
analyze = trans_analyze,
pretty_failure = trans_pretty_failure
}
val setup = I
#> autoref_post_simps.setup
end
›
setup Autoref_Translate.setup
subsubsection ‹Standard side-tactics›
text ‹Preconditions›
definition [simp, autoref_tag_defs]: "PRECOND_tag P == P"
lemma PRECOND_tagI: "P ==> PRECOND_tag P" by simp
abbreviation "SIDE_PRECOND P == PREFER PRECOND_tag P"
declaration ‹
Tagged_Solver.declare_solver @{thms PRECOND_tagI} @{binding PRECOND}
"Refinement: Solve preconditions"
( fn ctxt => SOLVED' (
SELECT_GOAL (auto_tac ctxt)
)
)
›
text ‹Optional preconditions›
definition [simp, autoref_tag_defs]: "PRECOND_OPT_tag P == P"
lemma PRECOND_OPT_tagI: "P ==> PRECOND_OPT_tag P" by simp
abbreviation "SIDE_PRECOND_OPT P == PREFER PRECOND_OPT_tag P"
declaration ‹
Tagged_Solver.declare_solver @{thms PRECOND_OPT_tagI} @{binding PRECOND_OPT}
"Refinement: Solve optional preconditions"
( fn ctxt => SOLVED' (asm_full_simp_tac ctxt))
›
end