mail: address list parsing.

R=golang-dev, r, r
CC=golang-dev
https://golang.org/cl/4547084

src/pkg/mail/message.go

@@ -2,17 +2,42 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// Package mail implements parsing of mail messages according to RFC 5322.
+/*
+Package mail implements parsing of mail messages.
+
+For the most part, this package follows the syntax as specified by RFC 5322.
+Notable divergences:
+	* Obsolete address formats are not parsed, including addresses with
+	  embedded route information.
+	* Group addresses are not parsed.
+	* The full range of spacing (the CFWS syntax element) is not supported,
+	  such as breaking addresses across lines.
+*/
 package mail
 
 import (
 	"bufio"
+	"bytes"
+	"fmt"
 	"io"
+	"log"
 	"net/textproto"
 	"os"
+	"strconv"
+	"strings"
 	"time"
 )
 
+var debug = debugT(false)
+
+type debugT bool
+
+func (d debugT) Printf(format string, args ...interface{}) {
+	if d {
+		log.Printf(format, args...)
+	}
+}
+
 // A Message represents a parsed mail message.
 type Message struct {
 	Header Header
@@ -93,3 +118,277 @@ func (h Header) Date() (*time.Time, os.Error) {
 	}
 	return parseDate(hdr)
 }
+
+// AddressList parses the named header field as a list of addresses.
+func (h Header) AddressList(key string) ([]*Address, os.Error) {
+	hdr := h.Get(key)
+	if hdr == "" {
+		return nil, ErrHeaderNotPresent
+	}
+	return newAddrParser(hdr).parseAddressList()
+}
+
+// Address represents a single mail address.
+// An address such as "Barry Gibbs <bg@example.com>" is represented
+// as Address{Name: "Barry Gibbs", Address: "bg@example.com"}.
+type Address struct {
+	Name    string // Proper name; may be empty.
+	Address string // user@domain
+}
+
+func (a *Address) String() string {
+	s := "<" + a.Address + ">"
+	if a.Name == "" {
+		return s
+	}
+	return "\"" + strconv.Quote(a.Name) + "\" " + s
+}
+
+type addrParser []byte
+
+func newAddrParser(s string) *addrParser {
+	p := addrParser([]byte(s))
+	return &p
+}
+
+func (p *addrParser) parseAddressList() ([]*Address, os.Error) {
+	var list []*Address
+	for {
+		p.skipSpace()
+		addr, err := p.parseAddress()
+		if err != nil {
+			return nil, err
+		}
+		list = append(list, addr)
+
+		p.skipSpace()
+		if p.empty() {
+			break
+		}
+		if !p.consume(',') {
+			return nil, os.ErrorString("mail: expected comma")
+		}
+	}
+	return list, nil
+}
+
+// parseAddress parses a single RFC 5322 address at the start of p.
+func (p *addrParser) parseAddress() (addr *Address, err os.Error) {
+	debug.Printf("parseAddress: %q", *p)
+	p.skipSpace()
+	if p.empty() {
+		return nil, os.ErrorString("mail: no address")
+	}
+
+	// address = name-addr / addr-spec
+	// TODO(dsymonds): Support parsing group address.
+
+	// addr-spec has a more restricted grammar than name-addr,
+	// so try parsing it first, and fallback to name-addr.
+	// TODO(dsymonds): Is this really correct?
+	spec, err := p.consumeAddrSpec()
+	if err == nil {
+		return &Address{
+			Address: spec,
+		}, err
+	}
+	debug.Printf("parseAddress: not an addr-spec: %v", err)
+	debug.Printf("parseAddress: state is now %q", *p)
+
+	// display-name
+	var displayName string
+	if p.peek() != '<' {
+		displayName, err = p.consumePhrase()
+		if err != nil {
+			return nil, err
+		}
+	}
+	debug.Printf("parseAddress: displayName=%q", displayName)
+
+	// angle-addr = "<" addr-spec ">"
+	p.skipSpace()
+	if !p.consume('<') {
+		return nil, os.ErrorString("mail: no angle-addr")
+	}
+	spec, err = p.consumeAddrSpec()
+	if err != nil {
+		return nil, err
+	}
+	if !p.consume('>') {
+		return nil, os.ErrorString("mail: unclosed angle-addr")
+	}
+	debug.Printf("parseAddress: spec=%q", spec)
+
+	return &Address{
+		Name:    displayName,
+		Address: spec,
+	}, nil
+}
+
+// consumeAddrSpec parses a single RFC 5322 addr-spec at the start of p.
+func (p *addrParser) consumeAddrSpec() (spec string, err os.Error) {
+	debug.Printf("consumeAddrSpec: %q", *p)
+
+	orig := *p
+	defer func() {
+		if err != nil {
+			*p = orig
+		}
+	}()
+
+	// local-part = dot-atom / quoted-string
+	var localPart string
+	p.skipSpace()
+	if p.empty() {
+		return "", os.ErrorString("mail: no addr-spec")
+	}
+	if p.peek() == '"' {
+		// quoted-string
+		debug.Printf("consumeAddrSpec: parsing quoted-string")
+		localPart, err = p.consumeQuotedString()
+	} else {
+		// dot-atom
+		debug.Printf("consumeAddrSpec: parsing dot-atom")
+		localPart, err = p.consumeAtom(true)
+	}
+	if err != nil {
+		debug.Printf("consumeAddrSpec: failed: %v", err)
+		return "", err
+	}
+
+	if !p.consume('@') {
+		return "", os.ErrorString("mail: missing @ in addr-spec")
+	}
+
+	// domain = dot-atom / domain-literal
+	var domain string
+	p.skipSpace()
+	if p.empty() {
+		return "", os.ErrorString("mail: no domain in addr-spec")
+	}
+	// TODO(dsymonds): Handle domain-literal
+	domain, err = p.consumeAtom(true)
+	if err != nil {
+		return "", err
+	}
+
+	return localPart + "@" + domain, nil
+}
+
+// consumePhrase parses the RFC 5322 phrase at the start of p.
+func (p *addrParser) consumePhrase() (phrase string, err os.Error) {
+	debug.Printf("consumePhrase: [%s]", *p)
+	// phrase = 1*word
+	var words []string
+	for {
+		// word = atom / quoted-string
+		var word string
+		p.skipSpace()
+		if p.empty() {
+			return "", os.ErrorString("mail: missing phrase")
+		}
+		if p.peek() == '"' {
+			// quoted-string
+			word, err = p.consumeQuotedString()
+		} else {
+			// atom
+			word, err = p.consumeAtom(false)
+		}
+		if err != nil {
+			break
+		}
+		debug.Printf("consumePhrase: consumed %q", word)
+		words = append(words, word)
+	}
+	// Ignore any error if we got at least one word.
+	if err != nil && len(words) == 0 {
+		debug.Printf("consumePhrase: hit err: %v", err)
+		return "", os.ErrorString("mail: missing word in phrase")
+	}
+	return strings.Join(words, " "), nil
+}
+
+// consumeQuotedString parses the quoted string at the start of p.
+func (p *addrParser) consumeQuotedString() (qs string, err os.Error) {
+	// Assume first byte is '"'.
+	i := 1
+	qsb := make([]byte, 0, 10)
+Loop:
+	for {
+		if i >= p.len() {
+			return "", os.ErrorString("mail: unclosed quoted-string")
+		}
+		switch c := (*p)[i]; {
+		case c == '"':
+			break Loop
+		case c == '\\':
+			if i+1 == p.len() {
+				return "", os.ErrorString("mail: unclosed quoted-string")
+			}
+			qsb = append(qsb, (*p)[i+1])
+			i += 2
+		case '!' <= c && c <= '~', c == ' ' || c == '\t':
+			// qtext (printable US-ASCII excluding " and \), or
+			// FWS (almost; we're ignoring CRLF)
+			qsb = append(qsb, c)
+			i++
+		default:
+			return "", fmt.Errorf("mail: bad character in quoted-string: %q", c)
+		}
+	}
+	*p = (*p)[i+1:]
+	return string(qsb), nil
+}
+
+// consumeAtom parses an RFC 5322 atom at the start of p.
+// If dot is true, consumeAtom parses an RFC 5322 dot-atom instead.
+func (p *addrParser) consumeAtom(dot bool) (atom string, err os.Error) {
+	if !isAtext(p.peek(), false) {
+		return "", os.ErrorString("mail: invalid string")
+	}
+	i := 1
+	for ; i < p.len() && isAtext((*p)[i], dot); i++ {
+	}
+	// TODO(dsymonds): Remove the []byte() conversion here when 6g doesn't need it.
+	atom, *p = string([]byte((*p)[:i])), (*p)[i:]
+	return atom, nil
+}
+
+func (p *addrParser) consume(c byte) bool {
+	if p.empty() || p.peek() != c {
+		return false
+	}
+	*p = (*p)[1:]
+	return true
+}
+
+// skipSpace skips the leading space and tab characters.
+func (p *addrParser) skipSpace() {
+	*p = bytes.TrimLeft(*p, " \t")
+}
+
+func (p *addrParser) peek() byte {
+	return (*p)[0]
+}
+
+func (p *addrParser) empty() bool {
+	return p.len() == 0
+}
+
+func (p *addrParser) len() int {
+	return len(*p)
+}
+
+var atextChars = []byte("ABCDEFGHIJKLMNOPQRSTUVWXYZ" +
+	"abcdefghijklmnopqrstuvwxyz" +
+	"0123456789" +
+	"!#$%&'*+-/=?^_`{|}~")
+
+// isAtext returns true if c is an RFC 5322 atext character.
+// If dot is true, period is included.
+func isAtext(c byte, dot bool) bool {
+	if dot && c == '.' {
+		return true
+	}
+	return bytes.IndexByte(atextChars, c) >= 0
+}

src/pkg/mail/message_test.go

@@ -127,3 +127,74 @@ func TestDateParsing(t *testing.T) {
 		}
 	}
 }
+
+func TestAddressParsing(t *testing.T) {
+	tests := []struct {
+		addrsStr string
+		exp      []*Address
+	}{
+		// Bare address
+		{
+			`jdoe@machine.example`,
+			[]*Address{&Address{
+				Address: "jdoe@machine.example",
+			}},
+		},
+		// RFC 5322, Appendix A.1.1
+		{
+			`John Doe <jdoe@machine.example>`,
+			[]*Address{&Address{
+				Name:    "John Doe",
+				Address: "jdoe@machine.example",
+			}},
+		},
+		// RFC 5322, Appendix A.1.2
+		{
+			`"Joe Q. Public" <john.q.public@example.com>`,
+			[]*Address{&Address{
+				Name:    "Joe Q. Public",
+				Address: "john.q.public@example.com",
+			}},
+		},
+		{
+			`Mary Smith <mary@x.test>, jdoe@example.org, Who? <one@y.test>`,
+			[]*Address{
+				&Address{
+					Name:    "Mary Smith",
+					Address: "mary@x.test",
+				},
+				&Address{
+					Address: "jdoe@example.org",
+				},
+				&Address{
+					Name:    "Who?",
+					Address: "one@y.test",
+				},
+			},
+		},
+		{
+			`<boss@nil.test>, "Giant; \"Big\" Box" <sysservices@example.net>`,
+			[]*Address{
+				&Address{
+					Address: "boss@nil.test",
+				},
+				&Address{
+					Name:    `Giant; "Big" Box`,
+					Address: "sysservices@example.net",
+				},
+			},
+		},
+		// RFC 5322, Appendix A.1.3
+		// TODO(dsymonds): Group addresses.
+	}
+	for _, test := range tests {
+		addrs, err := newAddrParser(test.addrsStr).parseAddressList()
+		if err != nil {
+			t.Errorf("Failed parsing %q: %v", test.addrsStr, err)
+			continue
+		}
+		if !reflect.DeepEqual(addrs, test.exp) {
+			t.Errorf("Parse of %q: got %+v, want %+v", test.addrsStr, addrs, test.exp)
+		}
+	}
+}