1
// Code generated by running "go generate" in golang.org/x/text. DO NOT EDIT.
3
// Copyright 2016 The Go Authors. All rights reserved.
4
// Use of this source code is governed by a BSD-style
5
// license that can be found in the LICENSE file.
10
// Package idna implements IDNA2008 using the compatibility processing
11
// defined by UTS (Unicode Technical Standard) #46, which defines a standard to
12
// deal with the transition from IDNA2003.
14
// IDNA2008 (Internationalized Domain Names for Applications), is defined in RFC
15
// 5890, RFC 5891, RFC 5892, RFC 5893 and RFC 5894.
16
// UTS #46 is defined in https://www.unicode.org/reports/tr46.
17
// See https://unicode.org/cldr/utility/idna.jsp for a visualization of the
18
// differences between these two standards.
19
package idna // import "golang.org/x/net/idna"
26
"golang.org/x/text/secure/bidirule"
27
"golang.org/x/text/unicode/bidi"
28
"golang.org/x/text/unicode/norm"
31
// NOTE: Unlike common practice in Go APIs, the functions will return a
32
// sanitized domain name in case of errors. Browsers sometimes use a partially
33
// evaluated string as lookup.
34
// TODO: the current error handling is, in my opinion, the least opinionated.
35
// Other strategies are also viable, though:
36
// Option 1) Return an empty string in case of error, but allow the user to
37
// specify explicitly which errors to ignore.
38
// Option 2) Return the partially evaluated string if it is itself a valid
39
// string, otherwise return the empty string in case of error.
40
// Option 3) Option 1 and 2.
41
// Option 4) Always return an empty string for now and implement Option 1 as
42
// needed, and document that the return string may not be empty in case of
43
// error in the future.
44
// I think Option 1 is best, but it is quite opinionated.
46
// ToASCII is a wrapper for Punycode.ToASCII.
47
func ToASCII(s string) (string, error) {
48
return Punycode.process(s, true)
51
// ToUnicode is a wrapper for Punycode.ToUnicode.
52
func ToUnicode(s string) (string, error) {
53
return Punycode.process(s, false)
56
// An Option configures a Profile at creation time.
57
type Option func(*options)
59
// Transitional sets a Profile to use the Transitional mapping as defined in UTS
60
// #46. This will cause, for example, "ß" to be mapped to "ss". Using the
61
// transitional mapping provides a compromise between IDNA2003 and IDNA2008
62
// compatibility. It is used by some browsers when resolving domain names. This
63
// option is only meaningful if combined with MapForLookup.
64
func Transitional(transitional bool) Option {
65
return func(o *options) { o.transitional = transitional }
68
// VerifyDNSLength sets whether a Profile should fail if any of the IDN parts
69
// are longer than allowed by the RFC.
71
// This option corresponds to the VerifyDnsLength flag in UTS #46.
72
func VerifyDNSLength(verify bool) Option {
73
return func(o *options) { o.verifyDNSLength = verify }
76
// RemoveLeadingDots removes leading label separators. Leading runes that map to
77
// dots, such as U+3002 IDEOGRAPHIC FULL STOP, are removed as well.
78
func RemoveLeadingDots(remove bool) Option {
79
return func(o *options) { o.removeLeadingDots = remove }
82
// ValidateLabels sets whether to check the mandatory label validation criteria
83
// as defined in Section 5.4 of RFC 5891. This includes testing for correct use
84
// of hyphens ('-'), normalization, validity of runes, and the context rules.
85
// In particular, ValidateLabels also sets the CheckHyphens and CheckJoiners flags
87
func ValidateLabels(enable bool) Option {
88
return func(o *options) {
89
// Don't override existing mappings, but set one that at least checks
90
// normalization if it is not set.
91
if o.mapping == nil && enable {
95
o.checkJoiners = enable
96
o.checkHyphens = enable
98
o.fromPuny = validateFromPunycode
105
// CheckHyphens sets whether to check for correct use of hyphens ('-') in
106
// labels. Most web browsers do not have this option set, since labels such as
107
// "r3---sn-apo3qvuoxuxbt-j5pe" are in common use.
109
// This option corresponds to the CheckHyphens flag in UTS #46.
110
func CheckHyphens(enable bool) Option {
111
return func(o *options) { o.checkHyphens = enable }
114
// CheckJoiners sets whether to check the ContextJ rules as defined in Appendix
115
// A of RFC 5892, concerning the use of joiner runes.
117
// This option corresponds to the CheckJoiners flag in UTS #46.
118
func CheckJoiners(enable bool) Option {
119
return func(o *options) {
121
o.checkJoiners = enable
125
// StrictDomainName limits the set of permissible ASCII characters to those
126
// allowed in domain names as defined in RFC 1034 (A-Z, a-z, 0-9 and the
127
// hyphen). This is set by default for MapForLookup and ValidateForRegistration,
128
// but is only useful if ValidateLabels is set.
130
// This option is useful, for instance, for browsers that allow characters
131
// outside this range, for example a '_' (U+005F LOW LINE). See
132
// http://www.rfc-editor.org/std/std3.txt for more details.
134
// This option corresponds to the UseSTD3ASCIIRules flag in UTS #46.
135
func StrictDomainName(use bool) Option {
136
return func(o *options) { o.useSTD3Rules = use }
139
// NOTE: the following options pull in tables. The tables should not be linked
140
// in as long as the options are not used.
142
// BidiRule enables the Bidi rule as defined in RFC 5893. Any application
143
// that relies on proper validation of labels should include this rule.
145
// This option corresponds to the CheckBidi flag in UTS #46.
146
func BidiRule() Option {
147
return func(o *options) { o.bidirule = bidirule.ValidString }
150
// ValidateForRegistration sets validation options to verify that a given IDN is
151
// properly formatted for registration as defined by Section 4 of RFC 5891.
152
func ValidateForRegistration() Option {
153
return func(o *options) {
154
o.mapping = validateRegistration
155
StrictDomainName(true)(o)
156
ValidateLabels(true)(o)
157
VerifyDNSLength(true)(o)
162
// MapForLookup sets validation and mapping options such that a given IDN is
163
// transformed for domain name lookup according to the requirements set out in
164
// Section 5 of RFC 5891. The mappings follow the recommendations of RFC 5894,
165
// RFC 5895 and UTS 46. It does not add the Bidi Rule. Use the BidiRule option
168
// The mappings include normalization and mapping case, width and other
169
// compatibility mappings.
170
func MapForLookup() Option {
171
return func(o *options) {
172
o.mapping = validateAndMap
173
StrictDomainName(true)(o)
174
ValidateLabels(true)(o)
184
removeLeadingDots bool
188
// fromPuny calls validation rules when converting A-labels to U-labels.
189
fromPuny func(p *Profile, s string) error
191
// mapping implements a validation and mapping step as defined in RFC 5895
192
// or UTS 46, tailored to, for example, domain registration or lookup.
193
mapping func(p *Profile, s string) (mapped string, isBidi bool, err error)
195
// bidirule, if specified, checks whether s conforms to the Bidi Rule
196
// defined in RFC 5893.
197
bidirule func(s string) bool
200
// A Profile defines the configuration of an IDNA mapper.
205
func apply(o *options, opts []Option) {
206
for _, f := range opts {
211
// New creates a new Profile.
213
// With no options, the returned Profile is the most permissive and equals the
214
// Punycode Profile. Options can be passed to further restrict the Profile. The
215
// MapForLookup and ValidateForRegistration options set a collection of options,
216
// for lookup and registration purposes respectively, which can be tailored by
217
// adding more fine-grained options, where later options override earlier
219
func New(o ...Option) *Profile {
225
// ToASCII converts a domain or domain label to its ASCII form. For example,
226
// ToASCII("bücher.example.com") is "xn--bcher-kva.example.com", and
227
// ToASCII("golang") is "golang". If an error is encountered it will return
228
// an error and a (partially) processed result.
229
func (p *Profile) ToASCII(s string) (string, error) {
230
return p.process(s, true)
233
// ToUnicode converts a domain or domain label to its Unicode form. For example,
234
// ToUnicode("xn--bcher-kva.example.com") is "bücher.example.com", and
235
// ToUnicode("golang") is "golang". If an error is encountered it will return
236
// an error and a (partially) processed result.
237
func (p *Profile) ToUnicode(s string) (string, error) {
239
pp.transitional = false
240
return pp.process(s, false)
243
// String reports a string with a description of the profile for debugging
244
// purposes. The string format may change with different versions.
245
func (p *Profile) String() string {
250
s = "NonTransitional"
261
if p.verifyDNSLength {
262
s += ":VerifyDNSLength"
268
// Punycode is a Profile that does raw punycode processing with a minimum
270
Punycode *Profile = punycode
272
// Lookup is the recommended profile for looking up domain names, according
273
// to Section 5 of RFC 5891. The exact configuration of this profile may
275
Lookup *Profile = lookup
277
// Display is the recommended profile for displaying domain names.
278
// The configuration of this profile may change over time.
279
Display *Profile = display
281
// Registration is the recommended profile for checking whether a given
282
// IDN is valid for registration, according to Section 4 of RFC 5891.
283
Registration *Profile = registration
285
punycode = &Profile{}
286
lookup = &Profile{options{
287
transitional: transitionalLookup,
292
fromPuny: validateFromPunycode,
293
mapping: validateAndMap,
294
bidirule: bidirule.ValidString,
296
display = &Profile{options{
301
fromPuny: validateFromPunycode,
302
mapping: validateAndMap,
303
bidirule: bidirule.ValidString,
305
registration = &Profile{options{
307
verifyDNSLength: true,
311
fromPuny: validateFromPunycode,
312
mapping: validateRegistration,
313
bidirule: bidirule.ValidString,
317
// Register: recommended for approving domain names: don't do any mappings
318
// but rather reject on invalid input. Bundle or block deviation characters.
321
type labelError struct{ label, code_ string }
323
func (e labelError) code() string { return e.code_ }
324
func (e labelError) Error() string {
325
return fmt.Sprintf("idna: invalid label %q", e.label)
330
func (e runeError) code() string { return "P1" }
331
func (e runeError) Error() string {
332
return fmt.Sprintf("idna: disallowed rune %U", e)
335
// process implements the algorithm described in section 4 of UTS #46,
336
// see https://www.unicode.org/reports/tr46.
337
func (p *Profile) process(s string, toASCII bool) (string, error) {
340
if p.mapping != nil {
341
s, isBidi, err = p.mapping(p, s)
343
// Remove leading empty labels.
344
if p.removeLeadingDots {
345
for ; len(s) > 0 && s[0] == '.'; s = s[1:] {
348
// TODO: allow for a quick check of the tables data.
349
// It seems like we should only create this error on ToASCII, but the
350
// UTS 46 conformance tests suggests we should always check this.
351
if err == nil && p.verifyDNSLength && s == "" {
352
err = &labelError{s, "A4"}
354
labels := labelIter{orig: s}
355
for ; !labels.done(); labels.next() {
356
label := labels.label()
358
// Empty labels are not okay. The label iterator skips the last
359
// label if it is empty.
360
if err == nil && p.verifyDNSLength {
361
err = &labelError{s, "A4"}
365
if strings.HasPrefix(label, acePrefix) {
366
u, err2 := decode(label[len(acePrefix):])
371
// Spec says keep the old label.
374
isBidi = isBidi || bidirule.DirectionString(u) != bidi.LeftToRight
376
if err == nil && p.fromPuny != nil {
377
err = p.fromPuny(p, u)
380
// This should be called on NonTransitional, according to the
381
// spec, but that currently does not have any effect. Use the
382
// original profile to preserve options.
383
err = p.validateLabel(u)
385
} else if err == nil {
386
err = p.validateLabel(label)
389
if isBidi && p.bidirule != nil && err == nil {
390
for labels.reset(); !labels.done(); labels.next() {
391
if !p.bidirule(labels.label()) {
392
err = &labelError{s, "B"}
398
for labels.reset(); !labels.done(); labels.next() {
399
label := labels.label()
401
a, err2 := encode(acePrefix, label)
409
if p.verifyDNSLength && err == nil && (n == 0 || n > 63) {
410
err = &labelError{label, "A4"}
415
if toASCII && p.verifyDNSLength && err == nil {
416
// Compute the length of the domain name minus the root label and its dot.
418
if n > 0 && s[n-1] == '.' {
421
if len(s) < 1 || n > 253 {
422
err = &labelError{s, "A4"}
428
func normalize(p *Profile, s string) (mapped string, isBidi bool, err error) {
429
// TODO: consider first doing a quick check to see if any of these checks
430
// need to be done. This will make it slower in the general case, but
431
// faster in the common case.
432
mapped = norm.NFC.String(s)
433
isBidi = bidirule.DirectionString(mapped) == bidi.RightToLeft
434
return mapped, isBidi, nil
437
func validateRegistration(p *Profile, s string) (idem string, bidi bool, err error) {
438
// TODO: filter need for normalization in loop below.
439
if !norm.NFC.IsNormalString(s) {
440
return s, false, &labelError{s, "V1"}
442
for i := 0; i < len(s); {
443
v, sz := trie.lookupString(s[i:])
445
return s, bidi, runeError(utf8.RuneError)
447
bidi = bidi || info(v).isBidi(s[i:])
448
// Copy bytes not copied so far.
449
switch p.simplify(info(v).category()) {
450
// TODO: handle the NV8 defined in the Unicode idna data set to allow
451
// for strict conformance to IDNA2008.
452
case valid, deviation:
453
case disallowed, mapped, unknown, ignored:
454
r, _ := utf8.DecodeRuneInString(s[i:])
455
return s, bidi, runeError(r)
462
func (c info) isBidi(s string) bool {
464
return c&attributesMask == rtl
466
// TODO: also store bidi info for mapped data. This is possible, but a bit
467
// cumbersome and not for the common case.
468
p, _ := bidi.LookupString(s)
470
case bidi.R, bidi.AL, bidi.AN:
476
func validateAndMap(p *Profile, s string) (vm string, bidi bool, err error) {
481
// combinedInfoBits contains the or-ed bits of all runes. We use this
482
// to derive the mayNeedNorm bit later. This may trigger normalization
483
// overeagerly, but it will not do so in the common case. The end result
484
// is another 10% saving on BenchmarkProfile for the common case.
485
var combinedInfoBits info
486
for i := 0; i < len(s); {
487
v, sz := trie.lookupString(s[i:])
489
b = append(b, s[k:i]...)
490
b = append(b, "\ufffd"...)
493
err = runeError(utf8.RuneError)
497
combinedInfoBits |= info(v)
498
bidi = bidi || info(v).isBidi(s[i:])
501
// Copy bytes not copied so far.
502
switch p.simplify(info(v).category()) {
507
r, _ := utf8.DecodeRuneInString(s[start:])
511
case mapped, deviation:
512
b = append(b, s[k:start]...)
513
b = info(v).appendMapping(b, s[start:i])
515
b = append(b, s[k:start]...)
518
b = append(b, s[k:start]...)
519
b = append(b, "\ufffd"...)
524
// No changes so far.
525
if combinedInfoBits&mayNeedNorm != 0 {
526
s = norm.NFC.String(s)
529
b = append(b, s[k:]...)
530
if norm.NFC.QuickSpan(b) != len(b) {
531
b = norm.NFC.Bytes(b)
533
// TODO: the punycode converters require strings as input.
539
// A labelIter allows iterating over domain name labels.
540
type labelIter struct {
548
func (l *labelIter) reset() {
554
func (l *labelIter) done() bool {
555
return l.curStart >= len(l.orig)
558
func (l *labelIter) result() string {
560
return strings.Join(l.slice, ".")
565
func (l *labelIter) label() string {
569
p := strings.IndexByte(l.orig[l.curStart:], '.')
570
l.curEnd = l.curStart + p
572
l.curEnd = len(l.orig)
574
return l.orig[l.curStart:l.curEnd]
577
// next sets the value to the next label. It skips the last label if it is empty.
578
func (l *labelIter) next() {
581
if l.i >= len(l.slice) || l.i == len(l.slice)-1 && l.slice[l.i] == "" {
582
l.curStart = len(l.orig)
585
l.curStart = l.curEnd + 1
586
if l.curStart == len(l.orig)-1 && l.orig[l.curStart] == '.' {
587
l.curStart = len(l.orig)
592
func (l *labelIter) set(s string) {
594
l.slice = strings.Split(l.orig, ".")
599
// acePrefix is the ASCII Compatible Encoding prefix.
600
const acePrefix = "xn--"
602
func (p *Profile) simplify(cat category) category {
604
case disallowedSTD3Mapped:
610
case disallowedSTD3Valid:
620
case validNV8, validXV8:
621
// TODO: handle V2008
627
func validateFromPunycode(p *Profile, s string) error {
628
if !norm.NFC.IsNormalString(s) {
629
return &labelError{s, "V1"}
631
// TODO: detect whether string may have to be normalized in the following
633
for i := 0; i < len(s); {
634
v, sz := trie.lookupString(s[i:])
636
return runeError(utf8.RuneError)
638
if c := p.simplify(info(v).category()); c != valid && c != deviation {
639
return &labelError{s, "V6"}
654
stateStart joinState = iota
662
var joinStates = [][numJoinTypes]joinState{
664
joiningL: stateBefore,
665
joiningD: stateBefore,
668
joinVirama: stateVirama,
671
joiningL: stateBefore,
672
joiningD: stateBefore,
675
joiningL: stateBefore,
676
joiningD: stateBefore,
677
joiningT: stateBefore,
678
joinZWNJ: stateAfter,
680
joinVirama: stateBeforeVirama,
683
joiningL: stateBefore,
684
joiningD: stateBefore,
685
joiningT: stateBefore,
689
joiningD: stateBefore,
690
joiningT: stateAfter,
691
joiningR: stateStart,
694
joinVirama: stateAfter, // no-op as we can't accept joiners here
704
joinVirama: stateFAIL,
708
// validateLabel validates the criteria from Section 4.1. Item 1, 4, and 6 are
709
// already implicitly satisfied by the overall implementation.
710
func (p *Profile) validateLabel(s string) (err error) {
712
if p.verifyDNSLength {
713
return &labelError{s, "A4"}
718
if len(s) > 4 && s[2] == '-' && s[3] == '-' {
719
return &labelError{s, "V2"}
721
if s[0] == '-' || s[len(s)-1] == '-' {
722
return &labelError{s, "V3"}
728
trie := p.trie // p.checkJoiners is only set if trie is set.
729
// TODO: merge the use of this in the trie.
730
v, sz := trie.lookupString(s)
733
return &labelError{s, "V5"}
735
// Quickly return in the absence of zero-width (non) joiners.
736
if strings.Index(s, zwj) == -1 && strings.Index(s, zwnj) == -1 {
742
if s[i:i+sz] == zwj {
744
} else if s[i:i+sz] == zwnj {
747
st = joinStates[st][jt]
748
if x.isViramaModifier() {
749
st = joinStates[st][joinVirama]
751
if i += sz; i == len(s) {
754
v, sz = trie.lookupString(s[i:])
757
if st == stateFAIL || st == stateAfter {
758
return &labelError{s, "C"}
763
func ascii(s string) bool {
764
for i := 0; i < len(s); i++ {
765
if s[i] >= utf8.RuneSelf {