TODO.txt

@@ -1,6 +1,4 @@
 Write Repeat tests
-Clean up cursor tests
-Combine Cursor with State
 Think about not requiring so much Pos() when making messages
 Rename Seq2 to Seq
 Document Seq

bytes/regexp_test.go

@@ -4,7 +4,6 @@ package bytes_test
 
 import (
 	"bytes"
-	"strings"
 	"testing"
 
 	"git.codemonkeysoftware.net/b/gigaparsec"
@@ -29,7 +28,7 @@ func TestRegexp(t *testing.T) {
 
 	}))
 	t.Run("basically works", func(t *testing.T) {
-		result, err := pbytes.Regexp("a")(gigaparsec.MakeState(cursor.NewReaderAt(strings.NewReader("a"))))
+		result, err := pbytes.Regexp("a")(gigaparsec.MakeState(cursor.NewSlice([]byte("a"))))
 		must.NoError(t, err)
 		success, value, _ := result.Status()
 		test.True(t, success, test.Sprint(result.Message()))
@@ -39,8 +38,8 @@ func TestRegexp(t *testing.T) {
 }
 
 func TestRuneReader(t *testing.T) {
-	var s = "abcdefghijklmnopqrstuvwxyz"
-	rr := pbytes.NewRuneReader(cursor.NewReaderAt(strings.NewReader(s)))
+	var s = []byte("abcdefghijklmnopqrstuvwxyz")
+	rr := pbytes.NewRuneReader(cursor.NewSlice(s))
 	for i, b := range s {
 		r, n, err := rr.ReadRune()
 		test.NoError(t, err)
@@ -68,7 +67,7 @@ func TestMatchString(t *testing.T) {
 		notPrefix := func(b []byte) bool { return !bytes.HasPrefix(input, b) }
 		s := string(bgen.Filter(notPrefix).Draw(t, "s"))
 
-		result, err := pbytes.MatchString(s)(gigaparsec.MakeState(cursor.NewReaderAt(bytes.NewReader(input))))
+		result, err := pbytes.MatchString(s)(gigaparsec.MakeState(cursor.NewSlice(input)))
 		test.NoError(t, err)
 		success, _, _ := result.Status()
 		test.False(t, success)
@@ -78,7 +77,7 @@ func TestMatchString(t *testing.T) {
 		input := rapid.SliceOfN(rapid.Byte(), 1, 100).Draw(t, "input")
 		slen := rapid.IntRange(0, len(input)).Draw(t, "slen")
 		s := string(input[:slen])
-		result, err := pbytes.MatchString(s)(gigaparsec.MakeState(cursor.NewReaderAt(bytes.NewReader(input))))
+		result, err := pbytes.MatchString(s)(gigaparsec.MakeState(cursor.NewSlice(input)))
 		must.NoError(t, err)
 		success, value, next := result.Status()
 		must.True(t, success)

cursor/cursor.go

@@ -3,7 +3,6 @@
 package cursor
 
 import (
-	"errors"
 	"io"
 )
 
@@ -33,36 +32,50 @@ type Cursor[Datum any] interface {
 	At(pos uint64) Cursor[Datum]
 }
 
-type ReaderAt[T any] interface {
-	ReadAt(p []T, off int64) (n int, err error)
+type SliceCursor[Datum any] struct {
+	data   []Datum
+	offset uint64
 }
 
-type SliceReaderAt[T any] []T
+func NewSlice[Datum any](data []Datum) SliceCursor[Datum] {
+	return SliceCursor[Datum]{
+		data:   data,
+		offset: 0,
+	}
+}
 
-func (s SliceReaderAt[T]) ReadAt(dst []T, off int64) (n int, err error) {
-	if off < 0 {
-		return 0, errors.New("SliceReaderAt.ReadAt: negative offset")
+func (sc SliceCursor[Datum]) Read(dst []Datum) (n uint64, next Cursor[Datum], err error) {
+	if sc.offset == uint64(len(sc.data)) {
+		return 0, sc, io.EOF
 	}
-	if off >= int64(len(s)) {
-		return 0, io.EOF
-	}
-	n = copy(dst, s[off:])
-	if n < len(dst) {
+	copied := copy(dst, sc.data[sc.offset:])
+	if copied < len(dst) {
 		err = io.EOF
 	}
-	return n, err
+	n = uint64(copied)
+	sc.offset += n
+	return n, sc, err
 }
 
-type ReaderAtCursor[T any] struct {
-	r   ReaderAt[T]
+func (sc SliceCursor[Datum]) Pos() uint64 {
+	return sc.offset
+}
+
+func (sc SliceCursor[Datum]) At(pos uint64) Cursor[Datum] {
+	sc.offset = pos
+	return sc
+}
+
+type ReaderAtCursor struct {
+	r   io.ReaderAt
 	pos uint64
 }
 
-func NewReaderAt[T any](r ReaderAt[T]) ReaderAtCursor[T] {
-	return ReaderAtCursor[T]{r: r}
+func NewReaderAt(r io.ReaderAt) ReaderAtCursor {
+	return ReaderAtCursor{r: r}
 }
 
-func (rac ReaderAtCursor[T]) Read(dst []T) (uint64, Cursor[T], error) {
+func (rac ReaderAtCursor) Read(dst []byte) (uint64, Cursor[byte], error) {
 	n, err := rac.r.ReadAt(dst, int64(rac.pos))
 	if n > 0 {
 		rac.pos += uint64(n)
@@ -70,11 +83,11 @@ func (rac ReaderAtCursor[T]) Read(dst []T) (uint64, Cursor[T], error) {
 	return uint64(n), rac, err
 }
 
-func (rac ReaderAtCursor[T]) Pos() uint64 {
+func (rac ReaderAtCursor) Pos() uint64 {
 	return rac.pos
 }
 
-func (rac ReaderAtCursor[T]) At(pos uint64) Cursor[T] {
+func (rac ReaderAtCursor) At(pos uint64) Cursor[byte] {
 	rac.pos = pos
 	return rac
 }

cursor/cursor_test.go

@@ -100,8 +100,12 @@ func testCursor[C cursor.Cursor[byte]](t *testing.T, makeCursor func([]byte) C)
 	}))
 }
 
+func TestSliceCursor(t *testing.T) {
+	testCursor(t, cursor.NewSlice[byte])
+}
+
 func TestReaderAtCursor(t *testing.T) {
-	testCursor(t, func(b []byte) cursor.ReaderAtCursor[byte] {
+	testCursor(t, func(b []byte) cursor.ReaderAtCursor {
 		return cursor.NewReaderAt(bytes.NewReader(b))
 	})
 	t.Run("Read returns an error if the ReaderAt fails", rapid.MakeCheck(func(t *rapid.T) {

cursor/helper.go

@@ -0,0 +1,26 @@
+// SPDX-License-Identifier: Unlicense
+
+package cursor
+
+import (
+	"io"
+)
+
+// BufferedReaderAt uses a buffer to supplement an io.Reader
+// with limited backward seeking.
+type BufferedReaderAt struct{}
+
+func NewBufferedReaderAt(r io.Reader, minBuffer uint64) *BufferedReaderAt {
+	return nil
+}
+
+// ReadAt reads bytes from the underlying reader. If the offset is after
+// the end of the buffer, ReadAt will first read and ignore bytes from the
+// underlying reader until it reaches the offset. If the offset is
+// before the start of the buffer, ReadAt will return an error.
+//
+// If your parser needs unlimited lookahead, you should probably
+// just read the whole input into a slice and use BytesCursor.
+func (b *BufferedReaderAt) ReadAt(dst []byte, offset int64) (int, error) {
+	return 0, nil
+}

cursor/helper_test.go

@@ -0,0 +1,9 @@
+// SPDX-License-Identifier: Unlicense
+
+package cursor_test
+
+import "testing"
+
+func TestBufferedReader(t *testing.T) {
+	Todo(t)
+}

parser_test.go

@@ -30,7 +30,7 @@ func hasPrefix(prefix []byte) func([]byte) bool {
 func TestSlice(t *testing.T) {
 	assertParseFails := func(t rapid.TB, input []byte, p gigaparsec.Parser[byte, []byte]) {
 		t.Helper()
-		start := gigaparsec.MakeState(cursor.NewReaderAt(bytes.NewReader(input)))
+		start := gigaparsec.MakeState(cursor.NewSlice(input))
 		result, err := p(start)
 		must.NoError(t, err)
 		success, _, _ := result.Status()
@@ -66,7 +66,7 @@ func TestSlice(t *testing.T) {
 		input := rapid.SliceOfN(rapid.Byte(), 1, -1).Draw(t, "input")
 		sLen := rapid.IntRange(0, len(input)).Draw(t, "sLen")
 		s := input[:sLen]
-		start := gigaparsec.MakeState(cursor.NewReaderAt(bytes.NewReader(input)))
+		start := gigaparsec.MakeState(cursor.NewSlice(input))
 
 		result, err := gigaparsec.MatchSlice(s)(start)
 		must.NoError(t, err)
@@ -108,7 +108,7 @@ func TestBind(t *testing.T) {
 		p := makeParser(pConsume)
 		q := func(struct{}) gigaparsec.Parser[byte, struct{}] { return makeParser(qConsume) }
 
-		result, err := gigaparsec.Bind(p, q)(gigaparsec.MakeState(cursor.NewReaderAt(bytes.NewReader(nil))))
+		result, err := gigaparsec.Bind(p, q)(gigaparsec.MakeState(cursor.NewSlice([]byte{})))
 		must.NoError(t, err)
 		must.EqOp(t, pConsume || qConsume, result.Consumed())
 	}))
@@ -143,6 +143,4 @@ func TestRepeat(t *testing.T) {
 	t.Run("succeeds when number of successes is greater than minCount", Todo)
 	t.Run("consumes iff at least one application consumes", Todo)
 	t.Run("fails on error", Todo)
-	t.Run("position is unchanged on failure", Todo)
-	t.Run("position follows last success on overall success", Todo)
 }