blob: a98a31f40388894036a4122ef802465231b90149 [file] [log] [blame]
Matteo Scandoloa4285862020-12-01 18:10:10 -08001// Code generated by running "go generate" in golang.org/x/text. DO NOT EDIT.
2
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.
6
7// +build go1.10
8
9// Package idna implements IDNA2008 using the compatibility processing
10// defined by UTS (Unicode Technical Standard) #46, which defines a standard to
11// deal with the transition from IDNA2003.
12//
13// IDNA2008 (Internationalized Domain Names for Applications), is defined in RFC
14// 5890, RFC 5891, RFC 5892, RFC 5893 and RFC 5894.
15// UTS #46 is defined in https://www.unicode.org/reports/tr46.
16// See https://unicode.org/cldr/utility/idna.jsp for a visualization of the
17// differences between these two standards.
18package idna // import "golang.org/x/net/idna"
19
20import (
21 "fmt"
22 "strings"
23 "unicode/utf8"
24
25 "golang.org/x/text/secure/bidirule"
26 "golang.org/x/text/unicode/bidi"
27 "golang.org/x/text/unicode/norm"
28)
29
30// NOTE: Unlike common practice in Go APIs, the functions will return a
31// sanitized domain name in case of errors. Browsers sometimes use a partially
32// evaluated string as lookup.
33// TODO: the current error handling is, in my opinion, the least opinionated.
34// Other strategies are also viable, though:
35// Option 1) Return an empty string in case of error, but allow the user to
36// specify explicitly which errors to ignore.
37// Option 2) Return the partially evaluated string if it is itself a valid
38// string, otherwise return the empty string in case of error.
39// Option 3) Option 1 and 2.
40// Option 4) Always return an empty string for now and implement Option 1 as
41// needed, and document that the return string may not be empty in case of
42// error in the future.
43// I think Option 1 is best, but it is quite opinionated.
44
45// ToASCII is a wrapper for Punycode.ToASCII.
46func ToASCII(s string) (string, error) {
47 return Punycode.process(s, true)
48}
49
50// ToUnicode is a wrapper for Punycode.ToUnicode.
51func ToUnicode(s string) (string, error) {
52 return Punycode.process(s, false)
53}
54
55// An Option configures a Profile at creation time.
56type Option func(*options)
57
58// Transitional sets a Profile to use the Transitional mapping as defined in UTS
59// #46. This will cause, for example, "ß" to be mapped to "ss". Using the
60// transitional mapping provides a compromise between IDNA2003 and IDNA2008
61// compatibility. It is used by most browsers when resolving domain names. This
62// option is only meaningful if combined with MapForLookup.
63func Transitional(transitional bool) Option {
64 return func(o *options) { o.transitional = true }
65}
66
67// VerifyDNSLength sets whether a Profile should fail if any of the IDN parts
68// are longer than allowed by the RFC.
69func VerifyDNSLength(verify bool) Option {
70 return func(o *options) { o.verifyDNSLength = verify }
71}
72
73// RemoveLeadingDots removes leading label separators. Leading runes that map to
74// dots, such as U+3002 IDEOGRAPHIC FULL STOP, are removed as well.
75//
76// This is the behavior suggested by the UTS #46 and is adopted by some
77// browsers.
78func RemoveLeadingDots(remove bool) Option {
79 return func(o *options) { o.removeLeadingDots = remove }
80}
81
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.
85func ValidateLabels(enable bool) Option {
86 return func(o *options) {
87 // Don't override existing mappings, but set one that at least checks
88 // normalization if it is not set.
89 if o.mapping == nil && enable {
90 o.mapping = normalize
91 }
92 o.trie = trie
93 o.validateLabels = enable
94 o.fromPuny = validateFromPunycode
95 }
96}
97
98// StrictDomainName limits the set of permissible ASCII characters to those
99// allowed in domain names as defined in RFC 1034 (A-Z, a-z, 0-9 and the
100// hyphen). This is set by default for MapForLookup and ValidateForRegistration.
101//
102// This option is useful, for instance, for browsers that allow characters
103// outside this range, for example a '_' (U+005F LOW LINE). See
104// http://www.rfc-editor.org/std/std3.txt for more details This option
105// corresponds to the UseSTD3ASCIIRules option in UTS #46.
106func StrictDomainName(use bool) Option {
107 return func(o *options) {
108 o.trie = trie
109 o.useSTD3Rules = use
110 o.fromPuny = validateFromPunycode
111 }
112}
113
114// NOTE: the following options pull in tables. The tables should not be linked
115// in as long as the options are not used.
116
117// BidiRule enables the Bidi rule as defined in RFC 5893. Any application
118// that relies on proper validation of labels should include this rule.
119func BidiRule() Option {
120 return func(o *options) { o.bidirule = bidirule.ValidString }
121}
122
123// ValidateForRegistration sets validation options to verify that a given IDN is
124// properly formatted for registration as defined by Section 4 of RFC 5891.
125func ValidateForRegistration() Option {
126 return func(o *options) {
127 o.mapping = validateRegistration
128 StrictDomainName(true)(o)
129 ValidateLabels(true)(o)
130 VerifyDNSLength(true)(o)
131 BidiRule()(o)
132 }
133}
134
135// MapForLookup sets validation and mapping options such that a given IDN is
136// transformed for domain name lookup according to the requirements set out in
137// Section 5 of RFC 5891. The mappings follow the recommendations of RFC 5894,
138// RFC 5895 and UTS 46. It does not add the Bidi Rule. Use the BidiRule option
139// to add this check.
140//
141// The mappings include normalization and mapping case, width and other
142// compatibility mappings.
143func MapForLookup() Option {
144 return func(o *options) {
145 o.mapping = validateAndMap
146 StrictDomainName(true)(o)
147 ValidateLabels(true)(o)
148 }
149}
150
151type options struct {
152 transitional bool
153 useSTD3Rules bool
154 validateLabels bool
155 verifyDNSLength bool
156 removeLeadingDots bool
157
158 trie *idnaTrie
159
160 // fromPuny calls validation rules when converting A-labels to U-labels.
161 fromPuny func(p *Profile, s string) error
162
163 // mapping implements a validation and mapping step as defined in RFC 5895
164 // or UTS 46, tailored to, for example, domain registration or lookup.
165 mapping func(p *Profile, s string) (mapped string, isBidi bool, err error)
166
167 // bidirule, if specified, checks whether s conforms to the Bidi Rule
168 // defined in RFC 5893.
169 bidirule func(s string) bool
170}
171
172// A Profile defines the configuration of an IDNA mapper.
173type Profile struct {
174 options
175}
176
177func apply(o *options, opts []Option) {
178 for _, f := range opts {
179 f(o)
180 }
181}
182
183// New creates a new Profile.
184//
185// With no options, the returned Profile is the most permissive and equals the
186// Punycode Profile. Options can be passed to further restrict the Profile. The
187// MapForLookup and ValidateForRegistration options set a collection of options,
188// for lookup and registration purposes respectively, which can be tailored by
189// adding more fine-grained options, where later options override earlier
190// options.
191func New(o ...Option) *Profile {
192 p := &Profile{}
193 apply(&p.options, o)
194 return p
195}
196
197// ToASCII converts a domain or domain label to its ASCII form. For example,
198// ToASCII("bücher.example.com") is "xn--bcher-kva.example.com", and
199// ToASCII("golang") is "golang". If an error is encountered it will return
200// an error and a (partially) processed result.
201func (p *Profile) ToASCII(s string) (string, error) {
202 return p.process(s, true)
203}
204
205// ToUnicode converts a domain or domain label to its Unicode form. For example,
206// ToUnicode("xn--bcher-kva.example.com") is "bücher.example.com", and
207// ToUnicode("golang") is "golang". If an error is encountered it will return
208// an error and a (partially) processed result.
209func (p *Profile) ToUnicode(s string) (string, error) {
210 pp := *p
211 pp.transitional = false
212 return pp.process(s, false)
213}
214
215// String reports a string with a description of the profile for debugging
216// purposes. The string format may change with different versions.
217func (p *Profile) String() string {
218 s := ""
219 if p.transitional {
220 s = "Transitional"
221 } else {
222 s = "NonTransitional"
223 }
224 if p.useSTD3Rules {
225 s += ":UseSTD3Rules"
226 }
227 if p.validateLabels {
228 s += ":ValidateLabels"
229 }
230 if p.verifyDNSLength {
231 s += ":VerifyDNSLength"
232 }
233 return s
234}
235
236var (
237 // Punycode is a Profile that does raw punycode processing with a minimum
238 // of validation.
239 Punycode *Profile = punycode
240
241 // Lookup is the recommended profile for looking up domain names, according
242 // to Section 5 of RFC 5891. The exact configuration of this profile may
243 // change over time.
244 Lookup *Profile = lookup
245
246 // Display is the recommended profile for displaying domain names.
247 // The configuration of this profile may change over time.
248 Display *Profile = display
249
250 // Registration is the recommended profile for checking whether a given
251 // IDN is valid for registration, according to Section 4 of RFC 5891.
252 Registration *Profile = registration
253
254 punycode = &Profile{}
255 lookup = &Profile{options{
256 transitional: true,
257 useSTD3Rules: true,
258 validateLabels: true,
259 trie: trie,
260 fromPuny: validateFromPunycode,
261 mapping: validateAndMap,
262 bidirule: bidirule.ValidString,
263 }}
264 display = &Profile{options{
265 useSTD3Rules: true,
266 validateLabels: true,
267 trie: trie,
268 fromPuny: validateFromPunycode,
269 mapping: validateAndMap,
270 bidirule: bidirule.ValidString,
271 }}
272 registration = &Profile{options{
273 useSTD3Rules: true,
274 validateLabels: true,
275 verifyDNSLength: true,
276 trie: trie,
277 fromPuny: validateFromPunycode,
278 mapping: validateRegistration,
279 bidirule: bidirule.ValidString,
280 }}
281
282 // TODO: profiles
283 // Register: recommended for approving domain names: don't do any mappings
284 // but rather reject on invalid input. Bundle or block deviation characters.
285)
286
287type labelError struct{ label, code_ string }
288
289func (e labelError) code() string { return e.code_ }
290func (e labelError) Error() string {
291 return fmt.Sprintf("idna: invalid label %q", e.label)
292}
293
294type runeError rune
295
296func (e runeError) code() string { return "P1" }
297func (e runeError) Error() string {
298 return fmt.Sprintf("idna: disallowed rune %U", e)
299}
300
301// process implements the algorithm described in section 4 of UTS #46,
302// see https://www.unicode.org/reports/tr46.
303func (p *Profile) process(s string, toASCII bool) (string, error) {
304 var err error
305 var isBidi bool
306 if p.mapping != nil {
307 s, isBidi, err = p.mapping(p, s)
308 }
309 // Remove leading empty labels.
310 if p.removeLeadingDots {
311 for ; len(s) > 0 && s[0] == '.'; s = s[1:] {
312 }
313 }
314 // TODO: allow for a quick check of the tables data.
315 // It seems like we should only create this error on ToASCII, but the
316 // UTS 46 conformance tests suggests we should always check this.
317 if err == nil && p.verifyDNSLength && s == "" {
318 err = &labelError{s, "A4"}
319 }
320 labels := labelIter{orig: s}
321 for ; !labels.done(); labels.next() {
322 label := labels.label()
323 if label == "" {
324 // Empty labels are not okay. The label iterator skips the last
325 // label if it is empty.
326 if err == nil && p.verifyDNSLength {
327 err = &labelError{s, "A4"}
328 }
329 continue
330 }
331 if strings.HasPrefix(label, acePrefix) {
332 u, err2 := decode(label[len(acePrefix):])
333 if err2 != nil {
334 if err == nil {
335 err = err2
336 }
337 // Spec says keep the old label.
338 continue
339 }
340 isBidi = isBidi || bidirule.DirectionString(u) != bidi.LeftToRight
341 labels.set(u)
342 if err == nil && p.validateLabels {
343 err = p.fromPuny(p, u)
344 }
345 if err == nil {
346 // This should be called on NonTransitional, according to the
347 // spec, but that currently does not have any effect. Use the
348 // original profile to preserve options.
349 err = p.validateLabel(u)
350 }
351 } else if err == nil {
352 err = p.validateLabel(label)
353 }
354 }
355 if isBidi && p.bidirule != nil && err == nil {
356 for labels.reset(); !labels.done(); labels.next() {
357 if !p.bidirule(labels.label()) {
358 err = &labelError{s, "B"}
359 break
360 }
361 }
362 }
363 if toASCII {
364 for labels.reset(); !labels.done(); labels.next() {
365 label := labels.label()
366 if !ascii(label) {
367 a, err2 := encode(acePrefix, label)
368 if err == nil {
369 err = err2
370 }
371 label = a
372 labels.set(a)
373 }
374 n := len(label)
375 if p.verifyDNSLength && err == nil && (n == 0 || n > 63) {
376 err = &labelError{label, "A4"}
377 }
378 }
379 }
380 s = labels.result()
381 if toASCII && p.verifyDNSLength && err == nil {
382 // Compute the length of the domain name minus the root label and its dot.
383 n := len(s)
384 if n > 0 && s[n-1] == '.' {
385 n--
386 }
387 if len(s) < 1 || n > 253 {
388 err = &labelError{s, "A4"}
389 }
390 }
391 return s, err
392}
393
394func normalize(p *Profile, s string) (mapped string, isBidi bool, err error) {
395 // TODO: consider first doing a quick check to see if any of these checks
396 // need to be done. This will make it slower in the general case, but
397 // faster in the common case.
398 mapped = norm.NFC.String(s)
399 isBidi = bidirule.DirectionString(mapped) == bidi.RightToLeft
400 return mapped, isBidi, nil
401}
402
403func validateRegistration(p *Profile, s string) (idem string, bidi bool, err error) {
404 // TODO: filter need for normalization in loop below.
405 if !norm.NFC.IsNormalString(s) {
406 return s, false, &labelError{s, "V1"}
407 }
408 for i := 0; i < len(s); {
409 v, sz := trie.lookupString(s[i:])
410 if sz == 0 {
411 return s, bidi, runeError(utf8.RuneError)
412 }
413 bidi = bidi || info(v).isBidi(s[i:])
414 // Copy bytes not copied so far.
415 switch p.simplify(info(v).category()) {
416 // TODO: handle the NV8 defined in the Unicode idna data set to allow
417 // for strict conformance to IDNA2008.
418 case valid, deviation:
419 case disallowed, mapped, unknown, ignored:
420 r, _ := utf8.DecodeRuneInString(s[i:])
421 return s, bidi, runeError(r)
422 }
423 i += sz
424 }
425 return s, bidi, nil
426}
427
428func (c info) isBidi(s string) bool {
429 if !c.isMapped() {
430 return c&attributesMask == rtl
431 }
432 // TODO: also store bidi info for mapped data. This is possible, but a bit
433 // cumbersome and not for the common case.
434 p, _ := bidi.LookupString(s)
435 switch p.Class() {
436 case bidi.R, bidi.AL, bidi.AN:
437 return true
438 }
439 return false
440}
441
442func validateAndMap(p *Profile, s string) (vm string, bidi bool, err error) {
443 var (
444 b []byte
445 k int
446 )
447 // combinedInfoBits contains the or-ed bits of all runes. We use this
448 // to derive the mayNeedNorm bit later. This may trigger normalization
449 // overeagerly, but it will not do so in the common case. The end result
450 // is another 10% saving on BenchmarkProfile for the common case.
451 var combinedInfoBits info
452 for i := 0; i < len(s); {
453 v, sz := trie.lookupString(s[i:])
454 if sz == 0 {
455 b = append(b, s[k:i]...)
456 b = append(b, "\ufffd"...)
457 k = len(s)
458 if err == nil {
459 err = runeError(utf8.RuneError)
460 }
461 break
462 }
463 combinedInfoBits |= info(v)
464 bidi = bidi || info(v).isBidi(s[i:])
465 start := i
466 i += sz
467 // Copy bytes not copied so far.
468 switch p.simplify(info(v).category()) {
469 case valid:
470 continue
471 case disallowed:
472 if err == nil {
473 r, _ := utf8.DecodeRuneInString(s[start:])
474 err = runeError(r)
475 }
476 continue
477 case mapped, deviation:
478 b = append(b, s[k:start]...)
479 b = info(v).appendMapping(b, s[start:i])
480 case ignored:
481 b = append(b, s[k:start]...)
482 // drop the rune
483 case unknown:
484 b = append(b, s[k:start]...)
485 b = append(b, "\ufffd"...)
486 }
487 k = i
488 }
489 if k == 0 {
490 // No changes so far.
491 if combinedInfoBits&mayNeedNorm != 0 {
492 s = norm.NFC.String(s)
493 }
494 } else {
495 b = append(b, s[k:]...)
496 if norm.NFC.QuickSpan(b) != len(b) {
497 b = norm.NFC.Bytes(b)
498 }
499 // TODO: the punycode converters require strings as input.
500 s = string(b)
501 }
502 return s, bidi, err
503}
504
505// A labelIter allows iterating over domain name labels.
506type labelIter struct {
507 orig string
508 slice []string
509 curStart int
510 curEnd int
511 i int
512}
513
514func (l *labelIter) reset() {
515 l.curStart = 0
516 l.curEnd = 0
517 l.i = 0
518}
519
520func (l *labelIter) done() bool {
521 return l.curStart >= len(l.orig)
522}
523
524func (l *labelIter) result() string {
525 if l.slice != nil {
526 return strings.Join(l.slice, ".")
527 }
528 return l.orig
529}
530
531func (l *labelIter) label() string {
532 if l.slice != nil {
533 return l.slice[l.i]
534 }
535 p := strings.IndexByte(l.orig[l.curStart:], '.')
536 l.curEnd = l.curStart + p
537 if p == -1 {
538 l.curEnd = len(l.orig)
539 }
540 return l.orig[l.curStart:l.curEnd]
541}
542
543// next sets the value to the next label. It skips the last label if it is empty.
544func (l *labelIter) next() {
545 l.i++
546 if l.slice != nil {
547 if l.i >= len(l.slice) || l.i == len(l.slice)-1 && l.slice[l.i] == "" {
548 l.curStart = len(l.orig)
549 }
550 } else {
551 l.curStart = l.curEnd + 1
552 if l.curStart == len(l.orig)-1 && l.orig[l.curStart] == '.' {
553 l.curStart = len(l.orig)
554 }
555 }
556}
557
558func (l *labelIter) set(s string) {
559 if l.slice == nil {
560 l.slice = strings.Split(l.orig, ".")
561 }
562 l.slice[l.i] = s
563}
564
565// acePrefix is the ASCII Compatible Encoding prefix.
566const acePrefix = "xn--"
567
568func (p *Profile) simplify(cat category) category {
569 switch cat {
570 case disallowedSTD3Mapped:
571 if p.useSTD3Rules {
572 cat = disallowed
573 } else {
574 cat = mapped
575 }
576 case disallowedSTD3Valid:
577 if p.useSTD3Rules {
578 cat = disallowed
579 } else {
580 cat = valid
581 }
582 case deviation:
583 if !p.transitional {
584 cat = valid
585 }
586 case validNV8, validXV8:
587 // TODO: handle V2008
588 cat = valid
589 }
590 return cat
591}
592
593func validateFromPunycode(p *Profile, s string) error {
594 if !norm.NFC.IsNormalString(s) {
595 return &labelError{s, "V1"}
596 }
597 // TODO: detect whether string may have to be normalized in the following
598 // loop.
599 for i := 0; i < len(s); {
600 v, sz := trie.lookupString(s[i:])
601 if sz == 0 {
602 return runeError(utf8.RuneError)
603 }
604 if c := p.simplify(info(v).category()); c != valid && c != deviation {
605 return &labelError{s, "V6"}
606 }
607 i += sz
608 }
609 return nil
610}
611
612const (
613 zwnj = "\u200c"
614 zwj = "\u200d"
615)
616
617type joinState int8
618
619const (
620 stateStart joinState = iota
621 stateVirama
622 stateBefore
623 stateBeforeVirama
624 stateAfter
625 stateFAIL
626)
627
628var joinStates = [][numJoinTypes]joinState{
629 stateStart: {
630 joiningL: stateBefore,
631 joiningD: stateBefore,
632 joinZWNJ: stateFAIL,
633 joinZWJ: stateFAIL,
634 joinVirama: stateVirama,
635 },
636 stateVirama: {
637 joiningL: stateBefore,
638 joiningD: stateBefore,
639 },
640 stateBefore: {
641 joiningL: stateBefore,
642 joiningD: stateBefore,
643 joiningT: stateBefore,
644 joinZWNJ: stateAfter,
645 joinZWJ: stateFAIL,
646 joinVirama: stateBeforeVirama,
647 },
648 stateBeforeVirama: {
649 joiningL: stateBefore,
650 joiningD: stateBefore,
651 joiningT: stateBefore,
652 },
653 stateAfter: {
654 joiningL: stateFAIL,
655 joiningD: stateBefore,
656 joiningT: stateAfter,
657 joiningR: stateStart,
658 joinZWNJ: stateFAIL,
659 joinZWJ: stateFAIL,
660 joinVirama: stateAfter, // no-op as we can't accept joiners here
661 },
662 stateFAIL: {
663 0: stateFAIL,
664 joiningL: stateFAIL,
665 joiningD: stateFAIL,
666 joiningT: stateFAIL,
667 joiningR: stateFAIL,
668 joinZWNJ: stateFAIL,
669 joinZWJ: stateFAIL,
670 joinVirama: stateFAIL,
671 },
672}
673
674// validateLabel validates the criteria from Section 4.1. Item 1, 4, and 6 are
675// already implicitly satisfied by the overall implementation.
676func (p *Profile) validateLabel(s string) (err error) {
677 if s == "" {
678 if p.verifyDNSLength {
679 return &labelError{s, "A4"}
680 }
681 return nil
682 }
683 if !p.validateLabels {
684 return nil
685 }
686 trie := p.trie // p.validateLabels is only set if trie is set.
687 if len(s) > 4 && s[2] == '-' && s[3] == '-' {
688 return &labelError{s, "V2"}
689 }
690 if s[0] == '-' || s[len(s)-1] == '-' {
691 return &labelError{s, "V3"}
692 }
693 // TODO: merge the use of this in the trie.
694 v, sz := trie.lookupString(s)
695 x := info(v)
696 if x.isModifier() {
697 return &labelError{s, "V5"}
698 }
699 // Quickly return in the absence of zero-width (non) joiners.
700 if strings.Index(s, zwj) == -1 && strings.Index(s, zwnj) == -1 {
701 return nil
702 }
703 st := stateStart
704 for i := 0; ; {
705 jt := x.joinType()
706 if s[i:i+sz] == zwj {
707 jt = joinZWJ
708 } else if s[i:i+sz] == zwnj {
709 jt = joinZWNJ
710 }
711 st = joinStates[st][jt]
712 if x.isViramaModifier() {
713 st = joinStates[st][joinVirama]
714 }
715 if i += sz; i == len(s) {
716 break
717 }
718 v, sz = trie.lookupString(s[i:])
719 x = info(v)
720 }
721 if st == stateFAIL || st == stateAfter {
722 return &labelError{s, "C"}
723 }
724 return nil
725}
726
727func ascii(s string) bool {
728 for i := 0; i < len(s); i++ {
729 if s[i] >= utf8.RuneSelf {
730 return false
731 }
732 }
733 return true
734}