1
0
Fork 0
forked from forgejo/forgejo

[Vendor] update go-swagger v0.21.0 -> v0.25.0 (#12670)

* Update go-swagger

* vendor
This commit is contained in:
6543 2020-09-01 16:01:23 +02:00 committed by GitHub
parent 66843f2237
commit 3270e7a443
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
350 changed files with 26353 additions and 5552 deletions

View file

@ -26,7 +26,7 @@ type tomlLexer struct {
currentTokenStart int
currentTokenStop int
tokens []token
depth int
brackets []rune
line int
col int
endbufferLine int
@ -123,6 +123,8 @@ func (l *tomlLexer) lexVoid() tomlLexStateFn {
for {
next := l.peek()
switch next {
case '}': // after '{'
return l.lexRightCurlyBrace
case '[':
return l.lexTableKey
case '#':
@ -140,10 +142,6 @@ func (l *tomlLexer) lexVoid() tomlLexStateFn {
l.skip()
}
if l.depth > 0 {
return l.lexRvalue
}
if isKeyStartChar(next) {
return l.lexKey
}
@ -167,10 +165,8 @@ func (l *tomlLexer) lexRvalue() tomlLexStateFn {
case '=':
return l.lexEqual
case '[':
l.depth++
return l.lexLeftBracket
case ']':
l.depth--
return l.lexRightBracket
case '{':
return l.lexLeftCurlyBrace
@ -188,12 +184,10 @@ func (l *tomlLexer) lexRvalue() tomlLexStateFn {
fallthrough
case '\n':
l.skip()
if l.depth == 0 {
return l.lexVoid
if len(l.brackets) > 0 && l.brackets[len(l.brackets)-1] == '[' {
return l.lexRvalue
}
return l.lexRvalue
case '_':
return l.errorf("cannot start number with underscore")
return l.lexVoid
}
if l.follow("true") {
@ -223,9 +217,12 @@ func (l *tomlLexer) lexRvalue() tomlLexStateFn {
}
possibleDate := l.peekString(35)
dateMatch := dateRegexp.FindString(possibleDate)
if dateMatch != "" {
l.fastForward(len(dateMatch))
dateSubmatches := dateRegexp.FindStringSubmatch(possibleDate)
if dateSubmatches != nil && dateSubmatches[0] != "" {
l.fastForward(len(dateSubmatches[0]))
if dateSubmatches[2] == "" { // no timezone information => local date
return l.lexLocalDate
}
return l.lexDate
}
@ -233,10 +230,6 @@ func (l *tomlLexer) lexRvalue() tomlLexStateFn {
return l.lexNumber
}
if isAlphanumeric(next) {
return l.lexKey
}
return l.errorf("no value can start with %c", next)
}
@ -247,12 +240,17 @@ func (l *tomlLexer) lexRvalue() tomlLexStateFn {
func (l *tomlLexer) lexLeftCurlyBrace() tomlLexStateFn {
l.next()
l.emit(tokenLeftCurlyBrace)
return l.lexRvalue
l.brackets = append(l.brackets, '{')
return l.lexVoid
}
func (l *tomlLexer) lexRightCurlyBrace() tomlLexStateFn {
l.next()
l.emit(tokenRightCurlyBrace)
if len(l.brackets) == 0 || l.brackets[len(l.brackets)-1] != '{' {
return l.errorf("cannot have '}' here")
}
l.brackets = l.brackets[:len(l.brackets)-1]
return l.lexRvalue
}
@ -261,6 +259,11 @@ func (l *tomlLexer) lexDate() tomlLexStateFn {
return l.lexRvalue
}
func (l *tomlLexer) lexLocalDate() tomlLexStateFn {
l.emit(tokenLocalDate)
return l.lexRvalue
}
func (l *tomlLexer) lexTrue() tomlLexStateFn {
l.fastForward(4)
l.emit(tokenTrue)
@ -294,6 +297,9 @@ func (l *tomlLexer) lexEqual() tomlLexStateFn {
func (l *tomlLexer) lexComma() tomlLexStateFn {
l.next()
l.emit(tokenComma)
if len(l.brackets) > 0 && l.brackets[len(l.brackets)-1] == '{' {
return l.lexVoid
}
return l.lexRvalue
}
@ -324,7 +330,26 @@ func (l *tomlLexer) lexKey() tomlLexStateFn {
} else if r == '\n' {
return l.errorf("keys cannot contain new lines")
} else if isSpace(r) {
break
str := " "
// skip trailing whitespace
l.next()
for r = l.peek(); isSpace(r); r = l.peek() {
str += string(r)
l.next()
}
// break loop if not a dot
if r != '.' {
break
}
str += "."
// skip trailing whitespace after dot
l.next()
for r = l.peek(); isSpace(r); r = l.peek() {
str += string(r)
l.next()
}
growingString += str
continue
} else if r == '.' {
// skip
} else if !isValidBareChar(r) {
@ -353,6 +378,7 @@ func (l *tomlLexer) lexComment(previousState tomlLexStateFn) tomlLexStateFn {
func (l *tomlLexer) lexLeftBracket() tomlLexStateFn {
l.next()
l.emit(tokenLeftBracket)
l.brackets = append(l.brackets, '[')
return l.lexRvalue
}
@ -504,7 +530,7 @@ func (l *tomlLexer) lexStringAsString(terminator string, discardLeadingNewLine,
} else {
r := l.peek()
if 0x00 <= r && r <= 0x1F && !(acceptNewLines && (r == '\n' || r == '\r')) {
if 0x00 <= r && r <= 0x1F && r != '\t' && !(acceptNewLines && (r == '\n' || r == '\r')) {
return "", fmt.Errorf("unescaped control character %U", r)
}
l.next()
@ -535,7 +561,6 @@ func (l *tomlLexer) lexString() tomlLexStateFn {
}
str, err := l.lexStringAsString(terminator, discardLeadingNewLine, acceptNewLines)
if err != nil {
return l.errorf(err.Error())
}
@ -607,6 +632,10 @@ func (l *tomlLexer) lexInsideTableKey() tomlLexStateFn {
func (l *tomlLexer) lexRightBracket() tomlLexStateFn {
l.next()
l.emit(tokenRightBracket)
if len(l.brackets) == 0 || l.brackets[len(l.brackets)-1] != '[' {
return l.errorf("cannot have ']' here")
}
l.brackets = l.brackets[:len(l.brackets)-1]
return l.lexRvalue
}
@ -733,7 +762,27 @@ func (l *tomlLexer) run() {
}
func init() {
dateRegexp = regexp.MustCompile(`^\d{1,4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}(\.\d{1,9})?(Z|[+-]\d{2}:\d{2})`)
// Regexp for all date/time formats supported by TOML.
// Group 1: nano precision
// Group 2: timezone
//
// /!\ also matches the empty string
//
// Example matches:
//1979-05-27T07:32:00Z
//1979-05-27T00:32:00-07:00
//1979-05-27T00:32:00.999999-07:00
//1979-05-27 07:32:00Z
//1979-05-27 00:32:00-07:00
//1979-05-27 00:32:00.999999-07:00
//1979-05-27T07:32:00
//1979-05-27T00:32:00.999999
//1979-05-27 07:32:00
//1979-05-27 00:32:00.999999
//1979-05-27
//07:32:00
//00:32:00.999999
dateRegexp = regexp.MustCompile(`^(?:\d{1,4}-\d{2}-\d{2})?(?:[T ]?\d{2}:\d{2}:\d{2}(\.\d{1,9})?(Z|[+-]\d{2}:\d{2})?)?`)
}
// Entry point