516 lines
18 KiB
Text
516 lines
18 KiB
Text
/-
|
||
Copyright (c) 2018 Microsoft Corporation. All rights reserved.
|
||
Released under Apache 2.0 license as described in the file LICENSE.
|
||
Authors: Leonardo de Moura, Sebastian Ullrich
|
||
|
||
Implementation for the parsec parser combinators described in the
|
||
paper:
|
||
https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/parsec-paper-letter.pdf
|
||
-/
|
||
prelude
|
||
import init.data.to_string init.data.string.basic init.data.list.basic init.control.except
|
||
import init.data.repr init.lean.name init.data.dlist init.control.monad_fail init.control.combinators
|
||
namespace lean
|
||
namespace parser
|
||
@[reducible] def position : Type := nat
|
||
|
||
structure message :=
|
||
(pos : position := 0)
|
||
(unexpected : string := "") -- unexpected input
|
||
(expected : dlist string := dlist.empty) -- expected productions
|
||
|
||
open string (iterator)
|
||
|
||
def expected.to_string : list string → string
|
||
| [] := ""
|
||
| [e] := e
|
||
| [e1, e2] := e1 ++ " or " ++ e2
|
||
| (e::es) := e ++ ", " ++ expected.to_string es
|
||
|
||
protected def message.to_string (input : string) (msg : message) : string :=
|
||
let (line, col) := input.line_column msg.pos in
|
||
"error at line " ++ to_string line ++ ", column " ++ to_string col ++ ":\n" ++
|
||
"unexpected " ++ msg.unexpected ++ "\n" ++
|
||
let ex_list := msg.expected.to_list in
|
||
if ex_list = [] then "" else "expected " ++ expected.to_string ex_list
|
||
|
||
def message.repr (msg : message) : string :=
|
||
"{pos := " ++ repr msg.pos ++ ", " ++
|
||
"unexpected := " ++ repr msg.unexpected ++ ", " ++
|
||
"expected := dlist.of_list " ++ repr msg.expected.to_list ++ "}"
|
||
|
||
instance message_has_repr : has_repr message :=
|
||
⟨message.repr⟩
|
||
|
||
/-
|
||
Remark: we store expected "error" messages in `ok_eps` results.
|
||
They contain the error that would have occurred if a
|
||
successful "epsilon" alternative was not taken.
|
||
-/
|
||
inductive result (α : Type)
|
||
| ok (a : α) (it : iterator) : result
|
||
| ok_eps (a : α) (it : iterator) (expected : dlist string) : result
|
||
| error {} (msg : message) (consumed : bool) : result
|
||
|
||
@[inline] def result.mk_eps {α : Type} (a : α) (it : iterator) : result α :=
|
||
result.ok_eps a it dlist.empty
|
||
|
||
open result
|
||
|
||
def parser_t (m : Type → Type) (α : Type) :=
|
||
iterator → m (result α)
|
||
|
||
abbreviation parser (α : Type) := parser_t id α
|
||
|
||
namespace parser_t
|
||
variables {m : Type → Type} [monad m] {α β : Type}
|
||
|
||
def run (p : parser_t m α) (s : string) (fname := "") : m (except message α) :=
|
||
do r ← p s.mk_iterator,
|
||
pure $ match r with
|
||
| ok a _ := except.ok a
|
||
| ok_eps a _ _ := except.ok a
|
||
| error msg _ := except.error msg
|
||
|
||
protected def pure (a : α) : parser_t m α :=
|
||
λ it, pure $ mk_eps a it
|
||
|
||
def eps : parser_t m unit :=
|
||
parser_t.pure ()
|
||
|
||
protected def failure : parser_t m α :=
|
||
λ it, pure $ error { unexpected := "failure", pos := it.offset } ff
|
||
|
||
def merge (msg₁ msg₂ : message) : message :=
|
||
{ expected := msg₁.expected ++ msg₂.expected, ..msg₁ }
|
||
|
||
/--
|
||
The `bind p q` combinator behaves as follows:
|
||
1- If `p` fails, then it fails.
|
||
2- If `p` succeeds and consumes input, then execute `q`
|
||
3- If `q` succeeds but does not consume input, then execute `q`
|
||
and merge error messages if both do not consume any input.
|
||
-/
|
||
protected def bind (p : parser_t m α) (q : α → parser_t m β) : parser_t m β :=
|
||
λ it, do r ← p it,
|
||
match r with
|
||
| ok a it :=
|
||
do r ← q a it,
|
||
pure (match r with
|
||
| ok_eps b it msg₂ := ok b it
|
||
| error msg ff := error msg tt
|
||
| other := other)
|
||
| ok_eps a it ex₁ :=
|
||
do r ← q a it,
|
||
pure (match r with
|
||
| ok_eps b it ex₂ := ok_eps b it (ex₁ ++ ex₂)
|
||
| error msg₂ ff := error { expected := ex₁ ++ msg₂.expected, .. msg₂ } ff
|
||
| other := other)
|
||
| error msg c := pure $ error msg c
|
||
|
||
instance : monad (parser_t m) :=
|
||
{ bind := λ _ _, parser_t.bind, pure := λ _, parser_t.pure }
|
||
|
||
instance : monad_fail (parser_t m) :=
|
||
{ fail := λ _ s it, pure $ error { unexpected := s, pos := it.offset } ff }
|
||
|
||
instance : has_monad_lift m (parser_t m) :=
|
||
{ monad_lift := λ _ x it, do a ← x,
|
||
pure (mk_eps a it) }
|
||
|
||
def expect (msg : message) (exp : string) : message :=
|
||
{expected := dlist.singleton exp, ..msg}
|
||
|
||
@[inline] def label (p : parser_t m α) (ex : string) : parser_t m α :=
|
||
λ it, do r ← p it,
|
||
pure $ match r with
|
||
| ok_eps a it _ := ok_eps a it (dlist.singleton ex)
|
||
| error msg ff := error (expect msg ex) ff
|
||
| other := other
|
||
|
||
/--
|
||
`try p` behaves like `p`, but it pretends `p` hasn't
|
||
consumed any input when `p` fails.
|
||
|
||
It is useful for implementing infinite lookahead.
|
||
The parser `try p <|> q` will try `q` even when
|
||
`p` has consumed input.
|
||
|
||
It is also useful for specifying both the lexer and parser
|
||
together.
|
||
```
|
||
(do try (ch 'l' >> ch 'e' >> ch 't'), whitespace, ...)
|
||
<|>
|
||
...
|
||
```
|
||
Without the `try` combinator we will not be able to backtrack on the `let` keyword.
|
||
-/
|
||
def try (p : parser_t m α) : parser_t m α :=
|
||
λ it, do r ← p it,
|
||
pure $ match r with
|
||
| error msg _ := error msg ff
|
||
| other := other
|
||
|
||
/--
|
||
The `orelse p q` combinator behaves as follows:
|
||
1- If `p` consumed input, then return result produced by `p`
|
||
even if it produced an error.
|
||
Recall that the `try p` combinator can be used to
|
||
pretend that `p` did not consume any input, and
|
||
simulate infinite lookahead.
|
||
2- If `p` did not consume any input, and `q` consumed
|
||
input, then return result produced by `q`.
|
||
Note that, `q`'s result is returned even if
|
||
`p` succeeded without consuming input.
|
||
3- If `p` and `q` did not consume any input, then
|
||
it combines their error messages (even if one of
|
||
them succeeded).
|
||
-/
|
||
protected def orelse (p q : parser_t m α) : parser_t m α :=
|
||
λ it, do r ← p it,
|
||
match r with
|
||
| ok_eps a it' ex₁ :=
|
||
do r' ← q it,
|
||
pure (match r' with
|
||
| ok_eps _ _ ex₂ := ok_eps a it' (ex₁ ++ ex₂)
|
||
| error msg₂ ff := ok_eps a it' (ex₁ ++ msg₂.expected)
|
||
| other := other)
|
||
| error msg₁ ff :=
|
||
do r' ← q it,
|
||
pure (match r' with
|
||
| ok_eps a it' ex₂ := ok_eps a it' (msg₁.expected ++ ex₂)
|
||
| error msg₂ ff := error (merge msg₁ msg₂) ff
|
||
| other := other)
|
||
| other := pure other
|
||
|
||
instance : alternative (parser_t m) :=
|
||
{ orelse := λ _, parser_t.orelse,
|
||
failure := λ _, parser_t.failure }
|
||
|
||
/-- Parse `p` without consuming any input. -/
|
||
def lookahead (p : parser_t m α) : parser_t m α :=
|
||
λ it, do r ← p it,
|
||
pure $ match r with
|
||
| ok a s' := mk_eps a it
|
||
| other := other
|
||
|
||
/-- `not_followed_by p` succeeds when parser `p` fails -/
|
||
def not_followed_by (p : parser_t m α) (msg : string := "input") : parser_t m unit :=
|
||
λ it, do r ← p it,
|
||
pure $ match r with
|
||
| ok _ _ := error { pos := it.offset, unexpected := msg } ff
|
||
| ok_eps _ _ _ := error { pos := it.offset, unexpected := msg } ff
|
||
| error _ _ := mk_eps () it
|
||
end parser_t
|
||
|
||
/- Type class for abstracting from concrete monad stacks containing a `parser_t` somewhere. -/
|
||
class monad_parser (m : Type → Type) :=
|
||
-- analogous to e.g. `monad_state.lift`
|
||
(lift {} {α : Type} : parser α → m α)
|
||
-- Analogous to e.g. `monad_reader_adapter.map` before simplification (see there).
|
||
-- Its usage seems to be way too common to justify moving it into a separate type class.
|
||
(map {} {α : Type} : (∀ {n α} [monad n], parser_t n α → parser_t n α) → m α → m α)
|
||
|
||
instance {m : Type → Type} [monad m] : monad_parser (parser_t m) :=
|
||
{ lift := λ α p it, pure $ p it,
|
||
map := λ α f x, f x }
|
||
|
||
instance monad_parser_trans {m n : Type → Type} [has_monad_lift m n] [monad_functor m m n n] [monad_parser m] : monad_parser n :=
|
||
{ lift := λ α p, monad_lift (monad_parser.lift p : m α),
|
||
map := λ α f x, monad_map (λ β x, (monad_parser.map @f x : m β)) x }
|
||
|
||
namespace monad_parser
|
||
variables {m : Type → Type} [monad m] [monad_parser m] [alternative m] {α β : Type}
|
||
|
||
@[inline] def eps : m unit :=
|
||
lift $ parser_t.eps
|
||
|
||
def left_over : m iterator :=
|
||
lift $ λ it, mk_eps it it
|
||
|
||
@[inline] def label (p : m α) (ex : string) : m α :=
|
||
map (λ _ _ inst p, @parser_t.label _ inst _ p ex) p
|
||
|
||
infixr ` <?> `:2 := label
|
||
|
||
/--
|
||
`try p` behaves like `p`, but it pretends `p` hasn't
|
||
consumed any input when `p` fails.
|
||
|
||
It is useful for implementing infinite lookahead.
|
||
The parser `try p <|> q` will try `q` even when
|
||
`p` has consumed input.
|
||
|
||
It is also useful for specifying both the lexer and parser
|
||
together.
|
||
```
|
||
(do try (ch 'l' >> ch 'e' >> ch 't'), whitespace, ...)
|
||
<|>
|
||
...
|
||
```
|
||
Without the `try` combinator we will not be able to backtrack on the `let` keyword.
|
||
-/
|
||
@[inline] def try (p : m α) : m α :=
|
||
map (λ _ _ inst p, @parser_t.try _ inst _ p) p
|
||
|
||
/-- Parse `p` without consuming any input. -/
|
||
@[inline] def lookahead (p : m α) : m α :=
|
||
map (λ _ _ inst p, @parser_t.lookahead _ inst _ p) p
|
||
|
||
-- TODO(Sebastian): `monad_functor` is too weak to lift this, probably needs something like `monad_control`
|
||
/-
|
||
/-- `not_followed_by p` succeeds when parser `p` fails -/
|
||
@[inline] def not_followed_by (p : m α) (msg : string := "input") : m unit :=
|
||
map (λ _ _ inst p, @parser_t.not_followed_by _ inst _ p msg) p
|
||
-/
|
||
|
||
/-- Faster version of `not_followed_by (satisfy p)` -/
|
||
@[inline] def not_followed_by_sat (p : char → bool) : m unit :=
|
||
lift $ λ it,
|
||
if !it.has_next then mk_eps () it
|
||
else let c := it.curr in
|
||
if p c then error { pos := it.offset, unexpected := repr c } ff
|
||
else mk_eps () it
|
||
|
||
@[inline] def eoi_error (pos : position) : result α :=
|
||
error { pos := pos, unexpected := "end of input" } ff
|
||
|
||
def curr : m char :=
|
||
lift $ λ it, mk_eps it.curr it
|
||
|
||
@[inline] def cond (p : char → bool) (t : m α) (e : m α) : m α :=
|
||
mcond (p <$> curr) t e
|
||
|
||
/--
|
||
If the next character `c` satisfies `p`, then
|
||
update position and return `c`. Otherwise,
|
||
generate error message with current position and character. -/
|
||
@[inline] def satisfy (p : char → bool) : m char :=
|
||
lift $ λ it,
|
||
if !it.has_next then eoi_error it.offset
|
||
else let c := it.curr in
|
||
if p c then ok c it.next
|
||
else error { pos := it.offset, unexpected := repr c } ff
|
||
|
||
def ch (c : char) : m char :=
|
||
satisfy (= c)
|
||
|
||
def alpha : m char :=
|
||
satisfy char.is_alpha
|
||
|
||
def digit : m char :=
|
||
satisfy char.is_digit
|
||
|
||
def upper : m char :=
|
||
satisfy char.is_upper
|
||
|
||
def lower : m char :=
|
||
satisfy char.is_lower
|
||
|
||
def any : m char :=
|
||
lift $ λ it, if !it.has_next then error { pos := it.offset, unexpected := "end of input" } ff
|
||
else ok it.curr it.next
|
||
|
||
private def str_aux (s : string) : nat → iterator → iterator → result string
|
||
| 0 _ it := ok s it
|
||
| (n+1) s_it it :=
|
||
if !it.has_next then eoi_error it.offset
|
||
else if s_it.curr = it.curr then str_aux n s_it.next it.next
|
||
else error { pos := it.offset, unexpected := repr (it.curr) } ff
|
||
|
||
/--
|
||
`str s` parses a sequence of elements that match `s`. Returns the parsed string (i.e. `s`).
|
||
This parser consumes no input if it fails (even if a partial match).
|
||
Note: The behaviour of this parser is different to that the `string` parser in the Parsec Haskell library,
|
||
as this one is all-or-nothing.
|
||
-/
|
||
def str (s : string) : m string :=
|
||
lift $ λ it, if s.is_empty then mk_eps "" it
|
||
else str_aux s s.length s.mk_iterator it
|
||
|
||
private def take_aux : nat → string → iterator → result string
|
||
| 0 r it := ok r it
|
||
| (n+1) r it :=
|
||
if !it.has_next then eoi_error it.offset
|
||
else take_aux n (r.push (it.curr)) it.next
|
||
|
||
/-- Consume `n` characters. -/
|
||
def take (n : nat) : m string :=
|
||
lift $ λ it, if n = 0 then mk_eps "" it
|
||
else take_aux n "" it
|
||
|
||
@[inline] private def mk_string_result (r : string) (it : iterator) : result string :=
|
||
if r.is_empty then mk_eps r it
|
||
else ok r it
|
||
|
||
private def take_while_aux (p : char → bool) : nat → string → iterator → result string
|
||
| 0 r it := mk_string_result r it
|
||
| (n+1) r it :=
|
||
if !it.has_next then mk_string_result r it
|
||
else let c := it.curr in
|
||
if p c then take_while_aux n (r.push c) it.next
|
||
else mk_string_result r it
|
||
|
||
/--
|
||
Consume input as long as the predicate returns `tt`, and return the consumed input.
|
||
This parser does not fail. It will return an empty string if the predicate
|
||
returns `ff` on the current character. -/
|
||
def take_while (p : char → bool) : m string :=
|
||
lift $ λ it, take_while_aux p it.remaining "" it
|
||
|
||
def take_while_cont (p : char → bool) (ini : string) : m string :=
|
||
lift $ λ it, take_while_aux p it.remaining ini it
|
||
|
||
/--
|
||
Consume input as long as the predicate returns `tt`, and return the consumed input.
|
||
This parser requires the predicate to succeed on at least once. -/
|
||
def take_while1 (p : char → bool) : m string :=
|
||
lift $ λ it,
|
||
if !it.has_next then eoi_error it.offset
|
||
else let c := it.curr in
|
||
if p c
|
||
then let input := it.next in
|
||
take_while_aux p input.remaining (to_string c) input
|
||
else error { pos := it.offset, unexpected := repr c } ff
|
||
|
||
/--
|
||
Consume input as long as the predicate returns `ff` (i.e. until it returns `tt`), and return the consumed input.
|
||
This parser does not fail. -/
|
||
def take_until (p : char → bool) : m string :=
|
||
take_while (λ c, !p c)
|
||
|
||
def take_until1 (p : char → bool) : m string :=
|
||
take_while1 (λ c, !p c)
|
||
|
||
@[inline] private def mk_consumed_result (consumed : bool) (it : iterator) : result unit :=
|
||
if consumed then ok () it
|
||
else mk_eps () it
|
||
|
||
private def take_while_aux' (p : char → bool) : nat → bool → iterator → result unit
|
||
| 0 consumed it := mk_consumed_result consumed it
|
||
| (n+1) consumed it :=
|
||
if !it.has_next then mk_consumed_result consumed it
|
||
else let c := it.curr in
|
||
if p c then take_while_aux' n tt it.next
|
||
else mk_consumed_result consumed it
|
||
|
||
/-- Similar to `take_while` but it does not return the consumed input. -/
|
||
def take_while' (p : char → bool) : m unit :=
|
||
lift $ λ it, take_while_aux' p it.remaining ff it
|
||
|
||
/-- Similar to `take_while1` but it does not return the consumed input. -/
|
||
def take_while1' (p : char → bool) : m unit :=
|
||
lift $ λ it,
|
||
if !it.has_next then eoi_error it.offset
|
||
else let c := it.curr in
|
||
if p c
|
||
then let input := it.next in
|
||
take_while_aux' p input.remaining tt input
|
||
else error { pos := it.offset, unexpected := repr c } ff
|
||
|
||
/-- Consume zero or more whitespaces. -/
|
||
def whitespace : m unit :=
|
||
take_while' char.is_whitespace
|
||
|
||
/-- Shorthand for `p <* whitespace` -/
|
||
def lexeme (p : m α) : m α :=
|
||
p <* whitespace
|
||
|
||
/-- Parse a numeral in decimal. -/
|
||
def num : m nat :=
|
||
string.to_nat <$> (take_while1 char.is_digit)
|
||
|
||
/-- Return the number of characters left to be parsed. -/
|
||
def remaining : m nat :=
|
||
lift $ λ it, mk_eps it.remaining it
|
||
|
||
/-- Succeed only if there are at least `n` characters left. -/
|
||
def ensure (n : nat) : m unit :=
|
||
lift $ λ it, if n ≤ it.remaining then mk_eps () it
|
||
else error { pos := it.offset, unexpected := "end of input", expected := dlist.singleton ("at least " ++ to_string n ++ " characters") } ff
|
||
|
||
/-- Return the current position. -/
|
||
def pos : m position :=
|
||
lift $ λ it, mk_eps it.offset it
|
||
|
||
def many1_aux (p : m α) : nat → m (list α)
|
||
| 0 := do a ← p, return [a]
|
||
| (n+1) := do a ← p,
|
||
as ← (many1_aux n <|> return []),
|
||
return (a::as)
|
||
|
||
def many1 (p : m α) : m (list α) :=
|
||
do r ← remaining, many1_aux p r
|
||
|
||
def many (p : m α) : m (list α) :=
|
||
many1 p <|> return []
|
||
|
||
def many1_aux' (p : m α) : nat → m unit
|
||
| 0 := p >> return ()
|
||
| (n+1) := p >> (many1_aux' n <|> return ())
|
||
|
||
def many1' (p : m α) : m unit :=
|
||
do r ← remaining, many1_aux' p r
|
||
|
||
def many' (p : m α) : m unit :=
|
||
many1' p <|> return ()
|
||
|
||
def eoi : m unit :=
|
||
lift $ λ it, if it.remaining = 0 then mk_eps () it
|
||
else error { pos := it.offset, unexpected := repr it.curr, expected := dlist.singleton ("end of input") } ff
|
||
|
||
def sep_by1 (p : m α) (sep : m β) : m (list α) :=
|
||
(::) <$> p <*> many (sep >> p)
|
||
|
||
def sep_by (p : m α) (sep : m β) : m (list α) :=
|
||
sep_by1 p sep <|> return []
|
||
|
||
def fix_aux (f : m α → m α) : nat → m α
|
||
| 0 := failure
|
||
| (n+1) := f (fix_aux n)
|
||
|
||
def fix (f : m α → m α) : m α :=
|
||
do n ← remaining, fix_aux f (n+1)
|
||
|
||
def foldr_aux (f : α → β → β) (p : m α) (b : β) : nat → m β
|
||
| 0 := return b
|
||
| (n+1) := (f <$> p <*> foldr_aux n) <|> return b
|
||
|
||
/-- Matches zero or more occurrences of `p`, and folds the result. -/
|
||
def foldr (f : α → β → β) (p : m α) (b : β) : m β :=
|
||
do it ← left_over,
|
||
foldr_aux f p b it.remaining
|
||
|
||
def foldl_aux (f : α → β → α) (p : m β) : α → nat → m α
|
||
| a 0 := return a
|
||
| a (n+1) := (do x ← p, foldl_aux (f a x) n) <|> return a
|
||
|
||
/-- Matches zero or more occurrences of `p`, and folds the result. -/
|
||
def foldl (f : α → β → α) (a : α) (p : m β) : m α :=
|
||
do it ← left_over,
|
||
foldl_aux f p a it.remaining
|
||
|
||
def unexpected (msg : string) : m α :=
|
||
lift $ λ it, error {unexpected := msg, pos := it.offset} ff
|
||
|
||
def unexpected_at (msg : string) (pos : position) : m α :=
|
||
lift $ λ it, error {unexpected := msg, pos := pos} ff
|
||
|
||
end monad_parser
|
||
|
||
namespace parser_t
|
||
open monad_parser
|
||
variables {m : Type → Type} [monad m] {α β : Type}
|
||
|
||
def parse (p : parser_t m α) (s : string) (fname := "") : m (except message α) :=
|
||
run p s fname
|
||
|
||
def parse_with_eoi (p : parser_t m α) (s : string) (fname := "") : m (except message α) :=
|
||
run (p <* eoi) s fname
|
||
|
||
def parse_with_left_over (p : parser_t m α) (s : string) (fname := "") : m (except message (α × iterator)) :=
|
||
run (prod.mk <$> p <*> left_over) s fname
|
||
|
||
end parser_t
|
||
end parser
|
||
end lean
|