1
0
Fork 0
forked from forgejo/forgejo

Use Go1.11 module (#5743)

* Migrate to go modules

* make vendor

* Update mvdan.cc/xurls

* make vendor

* Update code.gitea.io/git

* make fmt-check

* Update github.com/go-sql-driver/mysql

* make vendor
This commit is contained in:
Mura Li 2019-03-27 19:15:23 +08:00 committed by Lunny Xiao
parent d578b71d61
commit d77176912b
575 changed files with 63239 additions and 13963 deletions

View file

@ -6,6 +6,39 @@ package language
import "errors"
// A MatchOption configures a Matcher.
type MatchOption func(*matcher)
// PreferSameScript will, in the absence of a match, result in the first
// preferred tag with the same script as a supported tag to match this supported
// tag. The default is currently true, but this may change in the future.
func PreferSameScript(preferSame bool) MatchOption {
return func(m *matcher) { m.preferSameScript = preferSame }
}
// TODO(v1.0.0): consider making Matcher a concrete type, instead of interface.
// There doesn't seem to be too much need for multiple types.
// Making it a concrete type allows MatchStrings to be a method, which will
// improve its discoverability.
// MatchStrings parses and matches the given strings until one of them matches
// the language in the Matcher. A string may be an Accept-Language header as
// handled by ParseAcceptLanguage. The default language is returned if no
// other language matched.
func MatchStrings(m Matcher, lang ...string) (tag Tag, index int) {
for _, accept := range lang {
desired, _, err := ParseAcceptLanguage(accept)
if err != nil {
continue
}
if tag, index, conf := m.Match(desired...); conf != No {
return tag, index
}
}
tag, index, _ = m.Match()
return
}
// Matcher is the interface that wraps the Match method.
//
// Match returns the best match for any of the given tags, along with
@ -36,25 +69,45 @@ func Comprehends(speaker, alternative Tag) Confidence {
// matched tag in t, but is augmented with the Unicode extension ('u')of the
// corresponding preferred tag. This allows user locale options to be passed
// transparently.
func NewMatcher(t []Tag) Matcher {
return newMatcher(t)
func NewMatcher(t []Tag, options ...MatchOption) Matcher {
return newMatcher(t, options)
}
func (m *matcher) Match(want ...Tag) (t Tag, index int, c Confidence) {
match, w, c := m.getBest(want...)
if match == nil {
t = m.default_.tag
} else {
if match != nil {
t, index = match.tag, match.index
} else {
// TODO: this should be an option
t = m.default_.tag
if m.preferSameScript {
outer:
for _, w := range want {
script, _ := w.Script()
if script.scriptID == 0 {
// Don't do anything if there is no script, such as with
// private subtags.
continue
}
for i, h := range m.supported {
if script.scriptID == h.maxScript {
t, index = h.tag, i
break outer
}
}
}
}
// TODO: select first language tag based on script.
}
if w.region != 0 && t.region != 0 && t.region.contains(w.region) {
t, _ = Raw.Compose(t, Region{w.region})
}
// Copy options from the user-provided tag into the result tag. This is hard
// to do after the fact, so we do it here.
// TODO: consider also adding in variants that are compatible with the
// matched language.
// TODO: Add back region if it is non-ambiguous? Or create another tag to
// preserve the region?
if u, ok := w.Extension('u'); ok {
t, _ = Raw.Compose(t, u)
// TODO: add in alternative variants to -u-va-.
// TODO: add preferred region to -u-rg-.
if e := w.Extensions(); len(e) > 0 {
t, _ = Raw.Compose(t, e)
}
return t, index, c
}
@ -91,7 +144,7 @@ var ErrMissingLikelyTagsData = errors.New("missing likely tags data")
// addLikelySubtags sets subtags to their most likely value, given the locale.
// In most cases this means setting fields for unknown values, but in some
// cases it may alter a value. It returns a ErrMissingLikelyTagsData error
// cases it may alter a value. It returns an ErrMissingLikelyTagsData error
// if the given locale cannot be expanded.
func (t Tag) addLikelySubtags() (Tag, error) {
id, err := addTags(t)
@ -300,8 +353,9 @@ func minimizeTags(t Tag) (Tag, error) {
// 1) compute the match between the two tags.
// 2) if the match is better than the previous best match, replace it
// with the new match. (see next section)
// b) if the current best match is above a certain threshold, return this
// match without proceeding to the next tag in "desired". [See Note 1]
// b) if the current best match is Exact and pin is true the result will be
// frozen to the language found thusfar, although better matches may
// still be found for the same language.
// 3) If the best match so far is below a certain threshold, return "default".
//
// Ranking:
@ -350,9 +404,6 @@ func minimizeTags(t Tag) (Tag, error) {
// found wins.
//
// Notes:
// [1] Note that even if we may not have a perfect match, if a match is above a
// certain threshold, it is considered a better match than any other match
// to a tag later in the list of preferred language tags.
// [2] In practice, as matching of Exact is done in a separate phase from
// matching the other levels, we reuse the Exact level to mean MaxExact in
// the second phase. As a consequence, we only need the levels defined by
@ -388,16 +439,18 @@ func minimizeTags(t Tag) (Tag, error) {
// matcher keeps a set of supported language tags, indexed by language.
type matcher struct {
default_ *haveTag
index map[langID]*matchHeader
passSettings bool
default_ *haveTag
supported []*haveTag
index map[langID]*matchHeader
passSettings bool
preferSameScript bool
}
// matchHeader has the lists of tags for exact matches and matches based on
// maximized and canonicalized tags for a given language.
type matchHeader struct {
exact []*haveTag
max []*haveTag
haveTags []*haveTag
original bool
}
// haveTag holds a supported Tag and its maximized script and region. The maximized
@ -427,7 +480,7 @@ type haveTag struct {
func makeHaveTag(tag Tag, index int) (haveTag, langID) {
max := tag
if tag.lang != 0 {
if tag.lang != 0 || tag.region != 0 || tag.script != 0 {
max, _ = max.canonicalize(All)
max, _ = addTags(max)
max.remakeString()
@ -452,29 +505,27 @@ func altScript(l langID, s scriptID) scriptID {
// addIfNew adds a haveTag to the list of tags only if it is a unique tag.
// Tags that have the same maximized values are linked by index.
func (h *matchHeader) addIfNew(n haveTag, exact bool) {
h.original = h.original || exact
// Don't add new exact matches.
for _, v := range h.exact {
for _, v := range h.haveTags {
if v.tag.equalsRest(n.tag) {
return
}
}
if exact {
h.exact = append(h.exact, &n)
}
// Allow duplicate maximized tags, but create a linked list to allow quickly
// comparing the equivalents and bail out.
for i, v := range h.max {
for i, v := range h.haveTags {
if v.maxScript == n.maxScript &&
v.maxRegion == n.maxRegion &&
v.tag.variantOrPrivateTagStr() == n.tag.variantOrPrivateTagStr() {
for h.max[i].nextMax != 0 {
i = int(h.max[i].nextMax)
for h.haveTags[i].nextMax != 0 {
i = int(h.haveTags[i].nextMax)
}
h.max[i].nextMax = uint16(len(h.max))
h.haveTags[i].nextMax = uint16(len(h.haveTags))
break
}
}
h.max = append(h.max, &n)
h.haveTags = append(h.haveTags, &n)
}
// header returns the matchHeader for the given language. It creates one if
@ -501,9 +552,13 @@ func toConf(d uint8) Confidence {
// newMatcher builds an index for the given supported tags and returns it as
// a matcher. It also expands the index by considering various equivalence classes
// for a given tag.
func newMatcher(supported []Tag) *matcher {
func newMatcher(supported []Tag, options []MatchOption) *matcher {
m := &matcher{
index: make(map[langID]*matchHeader),
index: make(map[langID]*matchHeader),
preferSameScript: true,
}
for _, o := range options {
o(m)
}
if len(supported) == 0 {
m.default_ = &haveTag{}
@ -514,25 +569,28 @@ func newMatcher(supported []Tag) *matcher {
for i, tag := range supported {
pair, _ := makeHaveTag(tag, i)
m.header(tag.lang).addIfNew(pair, true)
m.supported = append(m.supported, &pair)
}
m.default_ = m.header(supported[0].lang).exact[0]
m.default_ = m.header(supported[0].lang).haveTags[0]
// Keep these in two different loops to support the case that two equivalent
// languages are distinguished, such as iw and he.
for i, tag := range supported {
pair, max := makeHaveTag(tag, i)
if max != tag.lang {
m.header(max).addIfNew(pair, false)
m.header(max).addIfNew(pair, true)
}
}
// update is used to add indexes in the map for equivalent languages.
// If force is true, the update will also apply to derived entries. To
// avoid applying a "transitive closure", use false.
update := func(want, have uint16, conf Confidence, force bool) {
// update will only add entries to original indexes, thus not computing any
// transitive relations.
update := func(want, have uint16, conf Confidence) {
if hh := m.index[langID(have)]; hh != nil {
if !force && len(hh.exact) == 0 {
if !hh.original {
return
}
hw := m.header(langID(want))
for _, ht := range hh.max {
for _, ht := range hh.haveTags {
v := *ht
if conf < v.conf {
v.conf = conf
@ -541,7 +599,7 @@ func newMatcher(supported []Tag) *matcher {
if v.altScript != 0 {
v.altScript = altScript(langID(want), v.maxScript)
}
hw.addIfNew(v, conf == Exact && len(hh.exact) > 0)
hw.addIfNew(v, conf == Exact && hh.original)
}
}
}
@ -549,9 +607,9 @@ func newMatcher(supported []Tag) *matcher {
// Add entries for languages with mutual intelligibility as defined by CLDR's
// languageMatch data.
for _, ml := range matchLang {
update(ml.want, ml.have, toConf(ml.distance), false)
update(ml.want, ml.have, toConf(ml.distance))
if !ml.oneway {
update(ml.have, ml.want, toConf(ml.distance), false)
update(ml.have, ml.want, toConf(ml.distance))
}
}
@ -561,10 +619,6 @@ func newMatcher(supported []Tag) *matcher {
// (their canonicalization simply substitutes a different language code, but
// nothing else), the match confidence is Exact, otherwise it is High.
for i, lm := range langAliasMap {
if lm.from == _sh {
continue
}
// If deprecated codes match and there is no fiddling with the script or
// or region, we consider it an exact match.
conf := Exact
@ -572,9 +626,9 @@ func newMatcher(supported []Tag) *matcher {
if !isExactEquivalent(langID(lm.from)) {
conf = High
}
update(lm.to, lm.from, conf, true)
update(lm.to, lm.from, conf)
}
update(lm.from, lm.to, conf, true)
update(lm.from, lm.to, conf)
}
return m
}
@ -583,28 +637,29 @@ func newMatcher(supported []Tag) *matcher {
// account the order of preference of the given tags.
func (m *matcher) getBest(want ...Tag) (got *haveTag, orig Tag, c Confidence) {
best := bestMatch{}
for _, w := range want {
for i, w := range want {
var max Tag
// Check for exact match first.
h := m.index[w.lang]
if w.lang != 0 {
// Base language is defined.
if h == nil {
continue
}
for i := range h.exact {
have := h.exact[i]
if have.tag.equalsRest(w) {
return have, w, Exact
}
// Base language is defined.
max, _ = w.canonicalize(Legacy | Deprecated | Macro)
// A region that is added through canonicalization is stronger than
// a maximized region: set it in the original (e.g. mo -> ro-MD).
if w.region != max.region {
w.region = max.region
}
max, _ = w.canonicalize(Legacy | Deprecated)
// TODO: should we do the same for scripts?
// See test case: en, sr, nl ; sh ; sr
max, _ = addTags(max)
} else {
// Base language is not defined.
if h != nil {
for i := range h.exact {
have := h.exact[i]
for i := range h.haveTags {
have := h.haveTags[i]
if have.tag.equalsRest(w) {
return have, w, Exact
}
@ -620,16 +675,23 @@ func (m *matcher) getBest(want ...Tag) (got *haveTag, orig Tag, c Confidence) {
continue
}
}
pin := true
for _, t := range want[i+1:] {
if w.lang == t.lang {
pin = false
break
}
}
// Check for match based on maximized tag.
for i := range h.max {
have := h.max[i]
best.update(have, w, max.script, max.region)
for i := range h.haveTags {
have := h.haveTags[i]
best.update(have, w, max.script, max.region, pin)
if best.conf == Exact {
for have.nextMax != 0 {
have = h.max[have.nextMax]
best.update(have, w, max.script, max.region)
have = h.haveTags[have.nextMax]
best.update(have, w, max.script, max.region, pin)
}
return best.have, best.want, High
return best.have, best.want, best.conf
}
}
}
@ -644,42 +706,68 @@ func (m *matcher) getBest(want ...Tag) (got *haveTag, orig Tag, c Confidence) {
// bestMatch accumulates the best match so far.
type bestMatch struct {
have *haveTag
want Tag
conf Confidence
have *haveTag
want Tag
conf Confidence
pinnedRegion regionID
pinLanguage bool
sameRegionGroup bool
// Cached results from applying tie-breaking rules.
origLang bool
origReg bool
regDist uint8
origScript bool
parentDist uint8 // 255 if have is not an ancestor of want tag.
origLang bool
origReg bool
paradigmReg bool
regGroupDist uint8
origScript bool
}
// update updates the existing best match if the new pair is considered to be a
// better match.
// To determine if the given pair is a better match, it first computes the rough
// confidence level. If this surpasses the current match, it will replace it and
// update the tie-breaker rule cache. If there is a tie, it proceeds with applying
// a series of tie-breaker rules. If there is no conclusive winner after applying
// the tie-breaker rules, it leaves the current match as the preferred match.
func (m *bestMatch) update(have *haveTag, tag Tag, maxScript scriptID, maxRegion regionID) {
// better match. To determine if the given pair is a better match, it first
// computes the rough confidence level. If this surpasses the current match, it
// will replace it and update the tie-breaker rule cache. If there is a tie, it
// proceeds with applying a series of tie-breaker rules. If there is no
// conclusive winner after applying the tie-breaker rules, it leaves the current
// match as the preferred match.
//
// If pin is true and have and tag are a strong match, it will henceforth only
// consider matches for this language. This corresponds to the nothing that most
// users have a strong preference for the first defined language. A user can
// still prefer a second language over a dialect of the preferred language by
// explicitly specifying dialects, e.g. "en, nl, en-GB". In this case pin should
// be false.
func (m *bestMatch) update(have *haveTag, tag Tag, maxScript scriptID, maxRegion regionID, pin bool) {
// Bail if the maximum attainable confidence is below that of the current best match.
c := have.conf
if c < m.conf {
return
}
if have.maxScript != maxScript {
// Don't change the language once we already have found an exact match.
if m.pinLanguage && tag.lang != m.want.lang {
return
}
// Pin the region group if we are comparing tags for the same language.
if tag.lang == m.want.lang && m.sameRegionGroup {
_, sameGroup := regionGroupDist(m.pinnedRegion, have.maxRegion, have.maxScript, m.want.lang)
if !sameGroup {
return
}
}
if c == Exact && have.maxScript == maxScript {
// If there is another language and then another entry of this language,
// don't pin anything, otherwise pin the language.
m.pinLanguage = pin
}
if have.tag.equalsRest(tag) {
} else if have.maxScript != maxScript {
// There is usually very little comprehension between different scripts.
// In a few cases there may still be Low comprehension. This possibility is
// pre-computed and stored in have.altScript.
// In a few cases there may still be Low comprehension. This possibility
// is pre-computed and stored in have.altScript.
if Low < m.conf || have.altScript != maxScript {
return
}
c = Low
} else if have.maxRegion != maxRegion {
// There is usually a small difference between languages across regions.
// We use the region distance (below) to disambiguate between equal matches.
if High < c {
// There is usually a small difference between languages across regions.
c = High
}
}
@ -715,10 +803,17 @@ func (m *bestMatch) update(have *haveTag, tag Tag, maxScript scriptID, maxRegion
beaten = true
}
// Next we prefer smaller distances between regions, as defined by regionDist.
regDist := regionDist(have.maxRegion, maxRegion, tag.lang)
if !beaten && m.regDist != regDist {
if regDist > m.regDist {
regGroupDist, sameGroup := regionGroupDist(have.maxRegion, maxRegion, maxScript, tag.lang)
if !beaten && m.regGroupDist != regGroupDist {
if regGroupDist > m.regGroupDist {
return
}
beaten = true
}
paradigmReg := isParadigmLocale(tag.lang, have.maxRegion)
if !beaten && m.paradigmReg != paradigmReg {
if !paradigmReg {
return
}
beaten = true
@ -733,77 +828,52 @@ func (m *bestMatch) update(have *haveTag, tag Tag, maxScript scriptID, maxRegion
beaten = true
}
// Finally we prefer tags which have a closer parent relationship.
parentDist := parentDistance(have.tag.region, tag)
if !beaten && m.parentDist != parentDist {
if parentDist > m.parentDist {
return
}
beaten = true
}
// Update m to the newly found best match.
if beaten {
m.have = have
m.want = tag
m.conf = c
m.pinnedRegion = maxRegion
m.sameRegionGroup = sameGroup
m.origLang = origLang
m.origReg = origReg
m.paradigmReg = paradigmReg
m.origScript = origScript
m.regDist = regDist
m.parentDist = parentDist
m.regGroupDist = regGroupDist
}
}
// parentDistance returns the number of times Parent must be called before the
// regions match. It is assumed that it has already been checked that lang and
// script are identical. If haveRegion does not occur in the ancestor chain of
// tag, it returns 255.
func parentDistance(haveRegion regionID, tag Tag) uint8 {
p := tag.Parent()
d := uint8(1)
for haveRegion != p.region {
if p.region == 0 {
return 255
}
p = p.Parent()
d++
}
return d
}
// regionDist wraps regionDistance with some exceptions to the algorithmic distance.
func regionDist(a, b regionID, lang langID) uint8 {
if lang == _en {
// Two variants of non-US English are close to each other, regardless of distance.
if a != _US && b != _US {
return 2
func isParadigmLocale(lang langID, r regionID) bool {
for _, e := range paradigmLocales {
if langID(e[0]) == lang && (r == regionID(e[1]) || r == regionID(e[2])) {
return true
}
}
return uint8(regionDistance(a, b))
return false
}
// regionDistance computes the distance between two regions based on the
// distance in the graph of region containments as defined in CLDR. It iterates
// over increasingly inclusive sets of groups, represented as bit vectors, until
// the source bit vector has bits in common with the destination vector.
func regionDistance(a, b regionID) int {
if a == b {
return 0
// regionGroupDist computes the distance between two regions based on their
// CLDR grouping.
func regionGroupDist(a, b regionID, script scriptID, lang langID) (dist uint8, same bool) {
const defaultDistance = 4
aGroup := uint(regionToGroups[a]) << 1
bGroup := uint(regionToGroups[b]) << 1
for _, ri := range matchRegion {
if langID(ri.lang) == lang && (ri.script == 0 || scriptID(ri.script) == script) {
group := uint(1 << (ri.group &^ 0x80))
if 0x80&ri.group == 0 {
if aGroup&bGroup&group != 0 { // Both regions are in the group.
return ri.distance, ri.distance == defaultDistance
}
} else {
if (aGroup|bGroup)&group == 0 { // Both regions are not in the group.
return ri.distance, ri.distance == defaultDistance
}
}
}
}
p, q := regionInclusion[a], regionInclusion[b]
if p < nRegionGroups {
p, q = q, p
}
set := regionInclusionBits
if q < nRegionGroups && set[p]&(1<<q) != 0 {
return 1
}
d := 2
for goal := set[q]; set[p]&goal == 0; p = regionInclusionNext[p] {
d++
}
return d
return defaultDistance, true
}
func (t Tag) variants() string {
@ -850,4 +920,14 @@ func init() {
notEquivalent = append(notEquivalent, langID(lm.from))
}
}
// Maximize undefined regions of paradigm locales.
for i, v := range paradigmLocales {
max, _ := addTags(Tag{lang: langID(v[0])})
if v[1] == 0 {
paradigmLocales[i][1] = uint16(max.region)
}
if v[2] == 0 {
paradigmLocales[i][2] = uint16(max.region)
}
}
}