feat(frontends/lean): remove tactic hints for specific classes
The idea is to separate class-instance resolution and tactic framework as two independent engines.
This commit is contained in:
parent
dc0e6861b7
commit
90ece4dd1b
8 changed files with 19 additions and 89 deletions
|
@ -14,7 +14,6 @@ Author: Leonardo de Moura
|
|||
#include "library/num.h"
|
||||
#include "frontends/lean/parser.h"
|
||||
#include "frontends/lean/util.h"
|
||||
#include "frontends/lean/tactic_hint.h"
|
||||
#include "frontends/lean/tokens.h"
|
||||
|
||||
#ifndef LEAN_INSTANCE_DEFAULT_PRIORITY
|
||||
|
@ -229,7 +228,7 @@ void register_class_cmds(cmd_table & r) {
|
|||
/** \brief If the constant \c e is a class, return its name */
|
||||
optional<name> constant_is_ext_class(environment const & env, expr const & e) {
|
||||
name const & cls_name = const_name(e);
|
||||
if (is_class(env, cls_name) || !empty(get_tactic_hints(env, cls_name))) {
|
||||
if (is_class(env, cls_name)) {
|
||||
return optional<name>(cls_name);
|
||||
} else {
|
||||
return optional<name>();
|
||||
|
|
|
@ -15,7 +15,6 @@ Author: Leonardo de Moura
|
|||
#include "library/error_handling/error_handling.h"
|
||||
#include "frontends/lean/util.h"
|
||||
#include "frontends/lean/class.h"
|
||||
#include "frontends/lean/tactic_hint.h"
|
||||
#include "frontends/lean/local_context.h"
|
||||
#include "frontends/lean/choice_iterator.h"
|
||||
|
||||
|
@ -55,15 +54,7 @@ pair<expr, constraint> mk_placeholder_elaborator(std::shared_ptr<placeholder_con
|
|||
In this case, \c fn will do the following:
|
||||
1) if the elaborated type of ?m is a 'class' C, then the stream will start with
|
||||
a) all local instances of class C (if elaborator.local_instances == true)
|
||||
b) solutions produced by tactic_hints for class C
|
||||
|
||||
2) if the elaborated type of ?m is not a class, then the stream will only contain
|
||||
the solutions produced by tactic_hints.
|
||||
|
||||
The unifier only process delayed choice constraints when there are no other kind
|
||||
of constraint to be processed.
|
||||
|
||||
This is a helper class for implementing this choice function.
|
||||
b) all global instances of class C
|
||||
*/
|
||||
struct placeholder_elaborator : public choice_iterator {
|
||||
std::shared_ptr<placeholder_context> m_C;
|
||||
|
@ -77,22 +68,14 @@ struct placeholder_elaborator : public choice_iterator {
|
|||
// global declaration names that are class instances.
|
||||
// This information is retrieved using #get_class_instances.
|
||||
list<name> m_instances;
|
||||
// Tactic hints for the class
|
||||
list<tactic_hint_entry> m_tactics;
|
||||
// result produce by last executed tactic.
|
||||
proof_state_seq m_tactic_result;
|
||||
justification m_jst;
|
||||
|
||||
placeholder_elaborator(std::shared_ptr<placeholder_context> const & C,
|
||||
expr const & meta, expr const & meta_type,
|
||||
list<expr> const & local_insts, list<name> const & instances,
|
||||
list<tactic_hint_entry> const & tacs,
|
||||
justification const & j):
|
||||
choice_iterator(), m_C(C),
|
||||
m_meta(meta), m_meta_type(meta_type),
|
||||
m_local_instances(local_insts), m_instances(instances),
|
||||
m_tactics(tacs),
|
||||
m_jst(j) {
|
||||
choice_iterator(), m_C(C), m_meta(meta), m_meta_type(meta_type),
|
||||
m_local_instances(local_insts), m_instances(instances), m_jst(j) {
|
||||
}
|
||||
|
||||
constraints mk_constraints(constraint const & c, buffer<constraint> const & cs) {
|
||||
|
@ -157,21 +140,6 @@ struct placeholder_elaborator : public choice_iterator {
|
|||
}
|
||||
}
|
||||
|
||||
optional<constraints> get_next_tactic_result() {
|
||||
while (auto next = m_tactic_result.pull()) {
|
||||
m_tactic_result = next->second;
|
||||
if (!empty(next->first.get_goals()))
|
||||
continue; // has unsolved goals
|
||||
substitution subst = next->first.get_subst();
|
||||
expr const & mvar = get_app_fn(m_meta);
|
||||
bool relax = m_C->m_relax;
|
||||
constraints cs = metavar_closure(m_meta_type).mk_constraints(subst, m_jst, relax);
|
||||
constraint c = mk_eq_cnstr(mvar, subst.instantiate(mvar), m_jst, relax);
|
||||
return some(cons(c, cs));
|
||||
}
|
||||
return optional<constraints>();
|
||||
}
|
||||
|
||||
virtual optional<constraints> next() {
|
||||
while (!empty(m_local_instances)) {
|
||||
expr inst = head(m_local_instances);
|
||||
|
@ -187,18 +155,6 @@ struct placeholder_elaborator : public choice_iterator {
|
|||
if (auto cs = try_instance(inst))
|
||||
return cs;
|
||||
}
|
||||
if (auto cs = get_next_tactic_result())
|
||||
return cs;
|
||||
while (!empty(m_tactics)) {
|
||||
tactic const & tac = head(m_tactics).get_tactic();
|
||||
m_tactics = tail(m_tactics);
|
||||
proof_state ps(goals(goal(m_meta, m_meta_type)), substitution(), m_C->m_ngen.mk_child());
|
||||
try {
|
||||
m_tactic_result = tac(m_C->env(), m_C->ios(), ps);
|
||||
if (auto cs = get_next_tactic_result())
|
||||
return cs;
|
||||
} catch (exception &) {}
|
||||
}
|
||||
return optional<constraints>();
|
||||
}
|
||||
};
|
||||
|
@ -207,9 +163,7 @@ struct placeholder_elaborator : public choice_iterator {
|
|||
constraint mk_placeholder_cnstr(std::shared_ptr<placeholder_context> const & C, expr const & m) {
|
||||
environment const & env = C->env();
|
||||
justification j = mk_failed_to_synthesize_jst(env, m);
|
||||
auto choice_fn = [=](expr const & meta, expr const & meta_type, substitution const & s,
|
||||
name_generator const & /* ngen */) {
|
||||
expr const & mvar = get_app_fn(meta);
|
||||
auto choice_fn = [=](expr const & meta, expr const & meta_type, substitution const &, name_generator const &) {
|
||||
if (auto cls_name_it = is_ext_class(C->tc(), meta_type)) {
|
||||
name cls_name = *cls_name_it;
|
||||
list<expr> const & ctx = C->m_ctx.get_data();
|
||||
|
@ -217,13 +171,10 @@ constraint mk_placeholder_cnstr(std::shared_ptr<placeholder_context> const & C,
|
|||
if (C->use_local_instances())
|
||||
local_insts = get_local_instances(C->tc(), ctx, cls_name);
|
||||
list<name> insts = get_class_instances(env, cls_name);
|
||||
list<tactic_hint_entry> tacs;
|
||||
if (!s.is_assigned(mvar))
|
||||
tacs = get_tactic_hints(env, cls_name);
|
||||
if (empty(local_insts) && empty(insts) && empty(tacs))
|
||||
if (empty(local_insts) && empty(insts))
|
||||
return lazy_list<constraints>(); // nothing to be done
|
||||
// we are always strict with placeholders associated with classes
|
||||
return choose(std::make_shared<placeholder_elaborator>(C, meta, meta_type, local_insts, insts, tacs, j));
|
||||
return choose(std::make_shared<placeholder_elaborator>(C, meta, meta_type, local_insts, insts, j));
|
||||
} else {
|
||||
// do nothing, type is not a class...
|
||||
return lazy_list<constraints>(constraints());
|
||||
|
@ -313,7 +264,7 @@ constraint mk_placeholder_root_cnstr(std::shared_ptr<placeholder_context> const
|
|||
}
|
||||
|
||||
/** \brief Create a metavariable, and attach choice constraint for generating
|
||||
solutions using class-instances and tactic-hints.
|
||||
solutions using class-instances
|
||||
*/
|
||||
pair<expr, constraint> mk_placeholder_elaborator(
|
||||
environment const & env, io_state const & ios, local_context const & ctx,
|
||||
|
|
|
@ -21,8 +21,6 @@ namespace lean {
|
|||
typedef list<tactic_hint_entry> tactic_hint_entries;
|
||||
|
||||
struct tactic_hint_state {
|
||||
typedef name_map<tactic_hint_entries> class_tactics;
|
||||
class_tactics m_class_tactics;
|
||||
tactic_hint_entries m_tactics;
|
||||
};
|
||||
|
||||
|
@ -39,12 +37,7 @@ struct tactic_hint_config {
|
|||
}
|
||||
|
||||
static void add_entry_core(state & s, entry const & e) {
|
||||
if (e.m_class.is_anonymous())
|
||||
s.m_tactics = add(s.m_tactics, e);
|
||||
else if (auto it = s.m_class_tactics.find(e.m_class))
|
||||
s.m_class_tactics.insert(e.m_class, add(*it, e));
|
||||
else
|
||||
s.m_class_tactics.insert(e.m_class, entries(e));
|
||||
s.m_tactics = add(s.m_tactics, e);
|
||||
}
|
||||
|
||||
static void add_entry(environment const & env, io_state const &, state & s, entry const & e) {
|
||||
|
@ -67,16 +60,16 @@ struct tactic_hint_config {
|
|||
}
|
||||
|
||||
static void write_entry(serializer & s, entry const & e) {
|
||||
s << e.m_class << e.m_pre_tactic;
|
||||
s << e.m_pre_tactic;
|
||||
}
|
||||
|
||||
static entry read_entry(deserializer & d) {
|
||||
entry e;
|
||||
d >> e.m_class >> e.m_pre_tactic;
|
||||
d >> e.m_pre_tactic;
|
||||
return e;
|
||||
}
|
||||
static optional<unsigned> get_fingerprint(entry const & e) {
|
||||
return some(hash(e.m_class.hash(), e.m_pre_tactic.hash()));
|
||||
return some(e.m_pre_tactic.hash());
|
||||
}
|
||||
};
|
||||
|
||||
|
@ -95,11 +88,6 @@ void finalize_tactic_hint() {
|
|||
delete g_class_name;
|
||||
}
|
||||
|
||||
list<tactic_hint_entry> get_tactic_hints(environment const & env, name const & c) {
|
||||
tactic_hint_state const & s = tactic_hint_ext::get_state(env);
|
||||
return ptr_to_list(s.m_class_tactics.find(c));
|
||||
}
|
||||
|
||||
list<tactic_hint_entry> get_tactic_hints(environment const & env) {
|
||||
tactic_hint_state const & s = tactic_hint_ext::get_state(env);
|
||||
return s.m_tactics;
|
||||
|
@ -119,15 +107,9 @@ expr parse_tactic_name(parser & p) {
|
|||
}
|
||||
|
||||
environment tactic_hint_cmd(parser & p) {
|
||||
name cls;
|
||||
if (p.curr_is_token(get_lbracket_tk())) {
|
||||
p.next();
|
||||
cls = get_class_name(p.env(), p.parse_expr());
|
||||
p.check_token_next(get_rbracket_tk(), "invalid tactic hint, ']' expected");
|
||||
}
|
||||
expr pre_tac = parse_tactic_name(p);
|
||||
tactic tac = expr_to_tactic(p.env(), pre_tac, nullptr);
|
||||
return tactic_hint_ext::add_entry(p.env(), get_dummy_ios(), tactic_hint_entry(cls, pre_tac, tac));
|
||||
return tactic_hint_ext::add_entry(p.env(), get_dummy_ios(), tactic_hint_entry(pre_tac, tac));
|
||||
}
|
||||
|
||||
void register_tactic_hint_cmd(cmd_table & r) {
|
||||
|
|
|
@ -11,18 +11,16 @@ Author: Leonardo de Moura
|
|||
namespace lean {
|
||||
class tactic_hint_entry {
|
||||
friend struct tactic_hint_config;
|
||||
name m_class;
|
||||
expr m_pre_tactic;
|
||||
tactic m_tactic;
|
||||
bool m_compiled;
|
||||
public:
|
||||
tactic_hint_entry():m_compiled(false) {}
|
||||
tactic_hint_entry(name const & c, expr const & pre_tac, tactic const & tac):
|
||||
m_class(c), m_pre_tactic(pre_tac), m_tactic(tac), m_compiled(true) {}
|
||||
tactic_hint_entry(expr const & pre_tac, tactic const & tac):
|
||||
m_pre_tactic(pre_tac), m_tactic(tac), m_compiled(true) {}
|
||||
expr const & get_pre_tactic() const { return m_pre_tactic; }
|
||||
tactic const & get_tactic() const { return m_tactic; }
|
||||
};
|
||||
list<tactic_hint_entry> get_tactic_hints(environment const & env, name const & cls_name);
|
||||
list<tactic_hint_entry> get_tactic_hints(environment const & env);
|
||||
class parser;
|
||||
expr parse_tactic_name(parser & p);
|
||||
|
|
|
@ -17,7 +17,7 @@ definition my_tac3 := fixpoint (λ f, [apply @or.intro_left; f |
|
|||
apply @or.intro_right; f |
|
||||
assumption])
|
||||
|
||||
tactic_hint [or] my_tac3
|
||||
tactic_hint my_tac3
|
||||
|
||||
theorem T3 {a b c : Prop} (Hb : b) : a ∨ b ∨ c
|
||||
:= _
|
||||
|
|
|
@ -15,5 +15,5 @@ definition my_tac3 := fixpoint (λ f, [apply @or.intro_left; f |
|
|||
apply @or.intro_right; f |
|
||||
assumption])
|
||||
|
||||
tactic_hint [or] my_tac3
|
||||
tactic_hint my_tac3
|
||||
theorem T3 {a b c : Prop} (Hb : b) : a ∨ b ∨ c
|
||||
|
|
|
@ -17,7 +17,7 @@ definition my_tac := fixpoint (λ t, [ apply @inl_inhabited; t
|
|||
| apply @num.is_inhabited
|
||||
])
|
||||
|
||||
tactic_hint [inhabited] my_tac
|
||||
tactic_hint my_tac
|
||||
|
||||
theorem T : inhabited (sum false num)
|
||||
|
||||
|
|
|
@ -20,7 +20,7 @@ definition my_tac := repeat (trace "iteration"; state;
|
|||
.. apply @num.is_inhabited; trace "used num")) ; now
|
||||
|
||||
|
||||
tactic_hint [inhabited] my_tac
|
||||
tactic_hint my_tac
|
||||
|
||||
theorem T : inhabited (sum false num)
|
||||
|
||||
|
|
Loading…
Reference in a new issue