diff --git a/gen/charmap/charmap.go b/gen/charmap/charmap.go
new file mode 100644
index 0000000..f77250f
--- /dev/null
+++ b/gen/charmap/charmap.go
@@ -0,0 +1,211 @@
+// Package charmap provides a way to read character maps.
+package charmap
+
+import (
+	"bufio"
+	"bytes"
+	"encoding/hex"
+	"errors"
+	"fmt"
+	"io"
+	"strconv"
+	"unicode"
+)
+
+// An Error indicates an error parsing a character mapping file.
+type Error struct {
+	File string
+	Line int
+	Err  error
+}
+
+func (e *Error) Error() (s string) {
+	if e.File != "" {
+		s += e.File + ":"
+	}
+	if e.Line != 0 {
+		s += strconv.Itoa(e.Line) + ":"
+	}
+	m := e.Err.Error()
+	if s == "" {
+		return m
+	}
+	return s + " " + m
+}
+
+var (
+	errBadType        = errors.New("unknown entry type")
+	errColumns        = errors.New("expected 2 columns")
+	errUnicode        = errors.New("invalid unicode sequence")
+	errCodePointRange = errors.New("code point out of range")
+)
+
+// An EntryType is a type of entry in a character mapping.
+type EntryType uint32
+
+const (
+	// EntryOneByte is an entry where a one-byte character is mapped to Unicode.
+	EntryOneByte EntryType = iota
+	// EntryTwoByte is an entry where a two-byte character is mapped to Unicode.
+	EntryTwoByte
+	// EntryDigraph is an entry where two one-byte characters are mapped to
+	// Unicode.
+	EntryDigraph
+)
+
+// A Direction is a direction context for Unicode characters.
+type Direction uint32
+
+const (
+	// DirectionAny indicates that the character can be omitted in any
+	// direction.
+	DirectionAny Direction = iota
+	// DirectionLR indicates that the character requires left-to-right context.
+	DirectionLR
+	// DirectionRL indicates that the character requires right-to-left context.
+	DirectionRL
+)
+
+// An Entry is a single entry in a character map.
+type Entry struct {
+	Type      EntryType
+	Encoded   []byte
+	Direction Direction
+	Unicode   []byte
+}
+
+// A Reader is a character map reader.
+type Reader struct {
+	name    string
+	lineno  int
+	scanner *bufio.Scanner
+	encoded [2]byte
+	unicode bytes.Buffer
+}
+
+// NewReader creates a character map reader.
+func NewReader(r io.Reader, name string) *Reader {
+	return &Reader{
+		name:    name,
+		scanner: bufio.NewScanner(r),
+	}
+}
+
+func (r *Reader) parseEntry(line []byte) (e Entry, err error) {
+	// Split into two columns.
+	i := bytes.IndexByte(line, '\t')
+	if i == -1 || i == 0 {
+		return e, errColumns
+	}
+	c1 := line[:i]
+	c2 := line[i+1:]
+	if i := bytes.IndexByte(c2, '\t'); i != -1 {
+		c2 = c2[:i]
+	}
+	if len(c2) == 0 {
+		return e, errColumns
+	}
+
+	// Parse columns.
+	switch len(c1) {
+	case 4:
+		if string(c1[0:2]) != "0x" {
+			return e, errBadType
+		}
+		if _, err := hex.Decode(r.encoded[:], c1[2:]); err != nil {
+			return e, err
+		}
+		e.Type = EntryOneByte
+		e.Encoded = r.encoded[0:1]
+	case 6:
+		if string(c1[0:2]) != "0x" {
+			return e, errBadType
+		}
+		if _, err := hex.Decode(r.encoded[:], c1[2:]); err != nil {
+			return e, err
+		}
+		e.Type = EntryTwoByte
+		e.Encoded = r.encoded[0:2]
+	case 9:
+		if string(c1[0:2]) != "0x" || string(c1[4:7]) != "+0x" {
+			return e, errBadType
+		}
+		if _, err := hex.Decode(r.encoded[0:1], c1[2:4]); err != nil {
+			return e, err
+		}
+		if _, err := hex.Decode(r.encoded[1:2], c1[7:9]); err != nil {
+			return e, err
+		}
+		e.Type = EntryDigraph
+		e.Encoded = r.encoded[0:2]
+	default:
+		return e, errBadType
+	}
+	if c2[0] == '<' {
+		i := bytes.IndexByte(c2, '+')
+		if i == -1 {
+			return e, errors.New("invalid Unicode string")
+		}
+		ctx := c2[:i]
+		c2 = c2[i+1:]
+		switch string(ctx) {
+		case "<LR>":
+			e.Direction = DirectionLR
+		case "<RL>":
+			e.Direction = DirectionRL
+		default:
+			return e, fmt.Errorf("unknown context: %q", ctx)
+		}
+	}
+	r.unicode.Reset()
+	for c2 != nil {
+		var cp []byte
+		if i := bytes.IndexByte(c2, '+'); i != -1 {
+			cp = c2[:i]
+			c2 = c2[i+1:]
+		} else {
+			cp = c2
+			c2 = nil
+		}
+		if len(cp) < 2 || string(cp[0:2]) != "0x" {
+			return e, errUnicode
+		}
+		cp = cp[2:]
+		x, err := strconv.ParseUint(string(cp), 16, 32)
+		if err != nil {
+			return e, err
+		}
+		if x > unicode.MaxRune {
+			return e, errCodePointRange
+		}
+		r.unicode.WriteRune(rune(x))
+	}
+	e.Unicode = r.unicode.Bytes()
+	return e, nil
+}
+
+// Next returns the next entry in the character map. Returns io.EOF if there are
+// no more entries.
+func (r *Reader) Next() (e Entry, err error) {
+	for {
+		if !r.scanner.Scan() {
+			if err := r.scanner.Err(); err != nil {
+				return e, err
+			}
+			return e, io.EOF
+		}
+		r.lineno++
+		line := r.scanner.Bytes()
+		// Remove comment.
+		if i := bytes.IndexByte(line, '#'); i != -1 {
+			line = line[:i]
+		}
+		if len(line) != 0 {
+			e, err = r.parseEntry(line)
+			if err != nil {
+				err = &Error{r.name, r.lineno, err}
+			}
+			return
+		}
+	}
+}