Separate parsing status from actual errors

This commit is contained in:
Brandon Dyck 2024-09-11 10:25:45 -06:00
parent c1eae9fa34
commit d0460e71e3
3 changed files with 140 additions and 109 deletions

View File

@ -46,18 +46,18 @@ func Regexp(str string) gigaparsec.Parser[byte, []byte] {
}
re := regexp.MustCompile(str)
expected := fmt.Sprintf("match `%s`", str)
return func(input gigaparsec.State[byte]) (consumed bool, result gigaparsec.Result[byte, []byte], err error) {
return func(input gigaparsec.State[byte]) (gigaparsec.Result[byte, []byte], error) {
r := NewRuneReader(input.Cursor())
idx := re.FindReaderIndex(r)
// TODO Check error from r; this requires an Error() method on cursor.RuneReader.
if idx == nil {
return false, gigaparsec.Result[byte, []byte]{}, gigaparsec.ParseError{
return gigaparsec.Fail[byte, []byte](false, gigaparsec.Message{
Pos: input.Pos(),
Expected: []string{expected},
// TODO Not having a Got is unsatisfactory, but how do I extract useful information?
// Maybe just read a fixed number of bytes or to the end, whichever comes first?
// I could add extra methods to cursor.RuneReader to figure out how much it had read.
}
}), nil
}
// Alas, this is a little wasteful because a Regexp can only return indices
// when searching a RuneReader.
@ -65,13 +65,9 @@ func Regexp(str string) gigaparsec.Parser[byte, []byte] {
n, _, err := input.Cursor().Read(dst)
if err != nil {
// If we can't access those same bytes again, something is wrong.
return false, gigaparsec.Result[byte, []byte]{}, fmt.Errorf("Regex: unexpected error: %w", err)
return gigaparsec.Result[byte, []byte]{}, fmt.Errorf("Regex: unexpected error: %w", err)
}
result = gigaparsec.Result[byte, []byte]{
State: input.At(input.Pos() + n),
Value: dst,
Message: gigaparsec.MessageOK(input.Pos()),
}
return true, result, nil
next := input.At(input.Pos() + n)
return gigaparsec.Succeed(true, dst, next, gigaparsec.MessageOK(input.Pos())), nil
}
}

View File

@ -10,9 +10,57 @@ import (
)
type Result[In, Out any] struct {
Value Out
State[In]
Message
consumed, failed bool
Value Out
next State[In]
Message Message
}
func Fail[In, Out any](consumed bool, msg Message) Result[In, Out] {
return Result[In, Out]{
consumed: consumed,
failed: true,
Message: msg,
}
}
func (r Result[In, Out]) Failed() (ok, consumed bool, msg Message) {
ok = r.failed
if ok {
consumed = r.consumed
msg = r.Message
}
return
}
func Succeed[In, Out any](consumed bool, value Out, next State[In], msg Message) Result[In, Out] {
return Result[In, Out]{
failed: false,
Value: value,
consumed: consumed,
next: next,
Message: msg,
}
}
func (r Result[In, Out]) Succeeded() (ok, consumed bool, value Out, next State[In], msg Message) {
ok = !r.failed
if ok {
consumed = r.consumed
value = r.Value
next = r.next
msg = r.Message
}
return
}
func (r Result[In, Out]) Consumed() bool {
return r.consumed
}
func (r Result[In, Out]) Consume(consumed bool) Result[In, Out] {
r.consumed = consumed
return r
}
type Message struct {
@ -30,11 +78,11 @@ func MessageOK(pos uint64) Message { return Message{Pos: pos} }
func MessageEnd(pos uint64) Message { return Message{Pos: pos, Got: "end of input"} }
type ParseError Message
// type ParseError Message
func (pe ParseError) Error() string {
return fmt.Sprintf("parse error: %d: %s", pe.Pos, pe.Got)
}
// func (pe ParseError) Error() string {
// return fmt.Sprintf("parse error: %d: %s", pe.Pos, pe.Got)
// }
func MakeState[In any](c cursor.Cursor[In]) State[In] {
return State[In]{cursor: c}
@ -61,125 +109,115 @@ func (s State[In]) At(pos uint64) State[In] {
return State[In]{cursor: s.cursor.At(pos)}
}
type Parser[In, Out any] func(State[In]) (consumed bool, reply Result[In, Out], err error)
type Parser[In, Out any] func(State[In]) (Result[In, Out], error)
func Return[In, Out any](value Out) Parser[In, Out] {
return func(state State[In]) (bool, Result[In, Out], error) {
return false, Result[In, Out]{
Value: value,
State: state,
Message: MessageOK(state.Pos()),
}, nil
return func(state State[In]) (Result[In, Out], error) {
return Succeed(false, value, state, MessageOK(state.Pos())), nil
}
}
func Satisfy[T any](pred func(T) bool) Parser[T, T] {
return func(state State[T]) (bool, Result[T, T], error) {
return func(state State[T]) (Result[T, T], error) {
token := make([]T, 1)
n, next, err := state.Read(token)
if errors.Is(err, io.EOF) {
return false, Result[T, T]{}, ParseError(MessageEnd(state.Pos()))
return Fail[T, T](false, MessageEnd(state.Pos())), nil
}
if err != nil {
return false, Result[T, T]{}, err
return Result[T, T]{}, err
}
if n != 1 {
panic(fmt.Sprintf("expected 1 element from Read, but got %d", n))
}
if pred(token[0]) {
return true, Result[T, T]{
Value: token[0],
State: next,
Message: MessageOK(state.Pos()),
}, nil
return Succeed(true, token[0], next, MessageOK(state.Pos())), nil
}
return false, Result[T, T]{}, ParseError{
return Fail[T, T](false, Message{
Pos: state.Pos(),
Got: fmt.Sprint(token),
}
}), nil
}
}
func Slice[T comparable](s []T) Parser[T, []T] {
expected := fmt.Sprint(s)
return func(state State[T]) (consumed bool, reply Result[T, []T], err error) {
return func(state State[T]) (reply Result[T, []T], err error) {
token := make([]T, len(s))
_, next, err := state.Read(token)
if errors.Is(err, io.EOF) {
return false, Result[T, []T]{}, ParseError(MessageEnd(state.Pos()))
return Fail[T, []T](false, MessageEnd(state.Pos())), nil
}
if err != nil {
return false, Result[T, []T]{}, err
return Result[T, []T]{}, err
}
if !slices.Equal(s, token) {
return false, Result[T, []T]{}, ParseError{
return Fail[T, []T](false, Message{
Pos: state.Pos(),
Got: fmt.Sprint(token),
Expected: []string{expected},
}
}), nil
}
return true, Result[T, []T]{
Value: token,
State: next,
Message: MessageOK(state.Pos()),
}, nil
return Succeed(true, token, next, MessageOK(state.Pos())), nil
}
}
func Bind[In, A, B any](p Parser[In, A], f func(A) Parser[In, B]) Parser[In, B] {
return func(input State[In]) (bool, Result[In, B], error) {
consumed, resultA, err := p(input)
return func(input State[In]) (Result[In, B], error) {
resultA, err := p(input)
if err != nil {
return false, Result[In, B]{}, err
return Result[In, B]{}, err
}
consumed2, replyB, err := f(resultA.Value)(resultA.State)
return consumed || consumed2, replyB, err
if ok, consumed, msg := resultA.Failed(); ok {
return Fail[In, B](consumed, msg), nil
}
_, consumedA, valueA, next, _ := resultA.Succeeded()
resultB, err := f(valueA)(next)
if err != nil {
return Result[In, B]{}, err
}
return resultB.Consume(consumedA || resultB.Consumed()), nil
}
}
func Choose[In, Out any](p Parser[In, Out], ps ...Parser[In, Out]) Parser[In, Out] {
// TODO Check this against the Parsec paper again, and simplify it.
all := append([]Parser[In, Out]{p}, ps...)
return func(input State[In]) (bool, Result[In, Out], error) {
return func(input State[In]) (Result[In, Out], error) {
expecteds := make([][]string, 0, len(all))
var value Out
var got string
var gotGot bool
var failed bool
for _, q := range all {
consumed, result, err := q(input)
if consumed {
return consumed, result, err
result, err := q(input)
if err != nil {
return Result[In, Out]{}, err
}
if result.Consumed() {
return result, nil
}
var qMsg Message
if err != nil {
var parseErr ParseError
if !errors.As(err, &parseErr) {
// It broke. Give up.
return consumed, result, err
}
failed = failed && true
qMsg = Message(parseErr)
if isFailure, _, msg := result.Failed(); isFailure {
qMsg = msg
failed = true
} else {
_, _, qValue, _, msg := result.Succeeded()
if failed {
value = result.Value
value = qValue
failed = false
}
qMsg = result.Message
qMsg = msg
}
if !gotGot {
if got == "" {
got = qMsg.Got
gotGot = true
}
}
msg := Message{Pos: input.Pos(), Got: got, Expected: slices.Concat(expecteds...)}
if failed {
return false, Result[In, Out]{}, ParseError(msg)
return Fail[In, Out](false, msg), nil
}
return false, Result[In, Out]{
Value: value,
State: input,
Message: msg,
}, nil
return Succeed(false, value, input, msg), nil
}
}
@ -189,31 +227,29 @@ func Choose[In, Out any](p Parser[In, Out], ps ...Parser[In, Out]) Parser[In, Ou
// parser consumed nothing, Try will allow backing out of a complex
// parser that did partially succeeded.
func Try[In, Out any](p Parser[In, Out]) Parser[In, Out] {
return func(input State[In]) (bool, Result[In, Out], error) {
consumed, reply, err := p(input)
return func(input State[In]) (Result[In, Out], error) {
result, err := p(input)
if err != nil {
return false, Result[In, Out]{}, err
return result, err
}
return consumed, reply, err
if failed, _, msg := result.Failed(); failed {
return Fail[In, Out](false, msg), nil
}
return result, nil
}
}
func Label[In, Out any](p Parser[In, Out], l string) Parser[In, Out] {
return func(input State[In]) (consumed bool, reply Result[In, Out], err error) {
consumed, reply, err = p(input)
if consumed {
return
return func(input State[In]) (reply Result[In, Out], err error) {
result, err := p(input)
if err != nil || result.Consumed() {
return result, err
}
if err == nil {
reply.Message = reply.Message.expect(l)
return
if succeeded, _, value, next, msg := result.Succeeded(); succeeded {
return Succeed(false, value, next, msg.expect(l)), nil
}
var parseErr ParseError
if errors.As(err, &parseErr) {
err = ParseError(Message(parseErr).expect(l))
return
}
return
_, _, msg := result.Failed()
return Fail[In, Out](false, msg.expect(l)), nil
}
}
@ -223,21 +259,17 @@ func Map[In, Out1, Out2 any](p Parser[In, Out1], f func(Out1) Out2) Parser[In, O
})
}
func End[In any](s State[In]) (consumed bool, reply Result[In, struct{}], err error) {
func End[In any](s State[In]) (reply Result[In, struct{}], err error) {
_, _, err = s.cursor.Read([]In{})
if errors.Is(err, io.EOF) {
reply := Result[In, struct{}]{
State: s,
Message: MessageOK(s.Pos()),
}
return true, reply, nil
return Succeed(true, struct{}{}, s, MessageOK(s.Pos())), nil
}
if err != nil {
return false, Result[In, struct{}]{}, fmt.Errorf("End: unexpected error: %w", err)
return Result[In, struct{}]{}, fmt.Errorf("End: unexpected error: %w", err)
}
return false, Result[In, struct{}]{}, ParseError{
return Fail[In, struct{}](false, Message{
Pos: s.Pos(),
Got: "",
Expected: []string{"end of input"},
}
}), nil
}

View File

@ -7,6 +7,7 @@ import (
"git.codemonkeysoftware.net/b/gigaparsec"
"git.codemonkeysoftware.net/b/gigaparsec/cursor"
"github.com/shoenig/test"
"github.com/shoenig/test/must"
"pgregory.net/rapid"
)
@ -23,17 +24,17 @@ func hasPrefix(prefix []byte) func([]byte) bool {
}
func TestSlice(t *testing.T) {
assertParseFails := func(t rapid.TB, input []byte, p gigaparsec.Parser[byte, []byte]) (parseErr gigaparsec.ParseError) {
assertParseFails := func(t rapid.TB, input []byte, p gigaparsec.Parser[byte, []byte]) {
t.Helper()
start := gigaparsec.MakeState(cursor.NewSlice(input))
consumed, result, err := p(start)
test.ErrorAs(t, err, &parseErr, test.Sprint("expected ParseError"))
test.False(t, consumed, test.Sprint("expected consumed to be false"))
test.SliceEmpty(t, result.Value, test.Sprint("expected result value to be empty"))
result, err := p(start)
must.NoError(t, err)
failed, consumed, _ := result.Failed()
test.True(t, failed)
test.False(t, consumed)
if t.Failed() {
t.FailNow()
}
return parseErr
}
t.Run("fails with wrong contents", rapid.MakeCheck(func(t *rapid.T) {
@ -55,11 +56,13 @@ func TestSlice(t *testing.T) {
s := input[:sLen]
start := gigaparsec.MakeState(cursor.NewSlice(input))
consumed, result, err := gigaparsec.Slice(s)(start)
test.NoError(t, err)
test.True(t, consumed, test.Sprint("expected consumed to be true"))
test.SliceEqOp(t, s, result.Value)
test.EqOp(t, uint64(len(s)), result.State.Pos())
result, err := gigaparsec.Slice(s)(start)
must.NoError(t, err)
succeeded, consumed, value, next, _ := result.Succeeded()
test.True(t, succeeded)
test.True(t, consumed)
test.SliceEqOp(t, s, value)
test.EqOp(t, uint64(len(s)), next.Pos())
}))
}