hugo/parser/pageparser/pagelexer_shortcode.go

365 lines
9.2 KiB
Go
Raw Normal View History

// Copyright 2018 The Hugo Authors. All rights reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package pageparser
type lexerShortcodeState struct {
currLeftDelimItem ItemType
currRightDelimItem ItemType
isInline bool
currShortcodeName string // is only set when a shortcode is in opened state
closingState int // > 0 = on its way to be closed
elementStepNum int // step number in element
paramElements int // number of elements (name + value = 2) found first
openShortcodes map[string]bool // set of shortcodes in open state
}
// Shortcode syntax
var (
leftDelimSc = []byte("{{")
leftDelimScNoMarkup = []byte("{{<")
rightDelimScNoMarkup = []byte(">}}")
leftDelimScWithMarkup = []byte("{{%")
rightDelimScWithMarkup = []byte("%}}")
leftComment = []byte("/*") // comments in this context us used to to mark shortcodes as "not really a shortcode"
rightComment = []byte("*/")
)
func (l *pageLexer) isShortCodeStart() bool {
return l.hasPrefix(leftDelimScWithMarkup) || l.hasPrefix(leftDelimScNoMarkup)
}
func lexShortcodeLeftDelim(l *pageLexer) stateFunc {
l.pos += len(l.currentLeftShortcodeDelim())
if l.hasPrefix(leftComment) {
return lexShortcodeComment
}
l.emit(l.currentLeftShortcodeDelimItem())
l.elementStepNum = 0
l.paramElements = 0
return lexInsideShortcode
}
func lexShortcodeComment(l *pageLexer) stateFunc {
posRightComment := l.index(append(rightComment, l.currentRightShortcodeDelim()...))
if posRightComment <= 1 {
return l.errorf("comment must be closed")
}
// we emit all as text, except the comment markers
l.emit(tText)
l.pos += len(leftComment)
l.ignore()
l.pos += posRightComment - len(leftComment)
l.emit(tText)
l.pos += len(rightComment)
l.ignore()
l.pos += len(l.currentRightShortcodeDelim())
l.emit(tText)
return lexMainSection
}
func lexShortcodeRightDelim(l *pageLexer) stateFunc {
l.closingState = 0
l.pos += len(l.currentRightShortcodeDelim())
l.emit(l.currentRightShortcodeDelimItem())
return lexMainSection
}
// either:
// 1. param
// 2. "param" or "param\"
// 3. param="123" or param="123\"
// 4. param="Some \"escaped\" text"
// 5. `param`
// 6. param=`123`
func lexShortcodeParam(l *pageLexer, escapedQuoteStart bool) stateFunc {
first := true
nextEq := false
var r rune
for {
r = l.next()
if first {
if r == '"' || (r == '`' && !escapedQuoteStart) {
// a positional param with quotes
if l.paramElements == 2 {
return l.errorf("got quoted positional parameter. Cannot mix named and positional parameters")
}
l.paramElements = 1
l.backup()
if r == '"' {
return lexShortcodeQuotedParamVal(l, !escapedQuoteStart, tScParam)
}
return lexShortCodeParamRawStringVal(l, tScParam)
} else if r == '`' && escapedQuoteStart {
return l.errorf("unrecognized escape character")
}
first = false
} else if r == '=' {
// a named param
l.backup()
nextEq = true
break
}
if !isAlphaNumericOrHyphen(r) && r != '.' { // Floats have period
l.backup()
break
}
}
if l.paramElements == 0 {
l.paramElements++
if nextEq {
l.paramElements++
}
} else {
if nextEq && l.paramElements == 1 {
return l.errorf("got named parameter '%s'. Cannot mix named and positional parameters", l.current())
} else if !nextEq && l.paramElements == 2 {
return l.errorf("got positional parameter '%s'. Cannot mix named and positional parameters", l.current())
}
}
l.emit(tScParam)
return lexInsideShortcode
}
func lexShortcodeParamVal(l *pageLexer) stateFunc {
l.consumeToSpace()
l.emit(tScParamVal)
return lexInsideShortcode
}
func lexShortCodeParamRawStringVal(l *pageLexer, typ ItemType) stateFunc {
openBacktickFound := false
Loop:
for {
switch r := l.next(); {
case r == '`':
if openBacktickFound {
l.backup()
break Loop
} else {
openBacktickFound = true
l.ignore()
}
case r == eof:
return l.errorf("unterminated raw string in shortcode parameter-argument: '%s'", l.current())
}
}
l.emitString(typ)
l.next()
l.ignore()
return lexInsideShortcode
}
func lexShortcodeQuotedParamVal(l *pageLexer, escapedQuotedValuesAllowed bool, typ ItemType) stateFunc {
openQuoteFound := false
escapedInnerQuoteFound := false
escapedQuoteState := 0
Loop:
for {
switch r := l.next(); {
case r == '\\':
if l.peek() == '"' {
if openQuoteFound && !escapedQuotedValuesAllowed {
l.backup()
break Loop
} else if openQuoteFound {
2020-12-16 11:11:32 +00:00
// the coming quote is inside
escapedInnerQuoteFound = true
escapedQuoteState = 1
}
} else if l.peek() == '`' {
return l.errorf("unrecognized escape character")
}
case r == eof, r == '\n':
return l.errorf("unterminated quoted string in shortcode parameter-argument: '%s'", l.current())
case r == '"':
if escapedQuoteState == 0 {
if openQuoteFound {
l.backup()
break Loop
} else {
openQuoteFound = true
l.ignore()
}
} else {
escapedQuoteState = 0
}
}
}
if escapedInnerQuoteFound {
l.ignoreEscapesAndEmit(typ, true)
} else {
l.emitString(typ)
}
r := l.next()
if r == '\\' {
if l.peek() == '"' {
// ignore the escaped closing quote
l.ignore()
l.next()
l.ignore()
}
} else if r == '"' {
// ignore closing quote
l.ignore()
} else {
// handled by next state
l.backup()
}
return lexInsideShortcode
}
// Inline shortcodes has the form {{< myshortcode.inline >}}
var inlineIdentifier = []byte("inline ")
// scans an alphanumeric inside shortcode
func lexIdentifierInShortcode(l *pageLexer) stateFunc {
lookForEnd := false
Loop:
for {
switch r := l.next(); {
case isAlphaNumericOrHyphen(r):
// Allow forward slash inside names to make it possible to create namespaces.
case r == '/':
case r == '.':
l.isInline = l.hasPrefix(inlineIdentifier)
if !l.isInline {
return l.errorf("period in shortcode name only allowed for inline identifiers")
}
default:
l.backup()
word := string(l.input[l.start:l.pos])
if l.closingState > 0 && !l.openShortcodes[word] {
return l.errorf("closing tag for shortcode '%s' does not match start tag", word)
} else if l.closingState > 0 {
l.openShortcodes[word] = false
lookForEnd = true
}
l.closingState = 0
l.currShortcodeName = word
l.openShortcodes[word] = true
l.elementStepNum++
if l.isInline {
l.emit(tScNameInline)
} else {
l.emit(tScName)
}
break Loop
}
}
if lookForEnd {
return lexEndOfShortcode
}
return lexInsideShortcode
}
func lexEndOfShortcode(l *pageLexer) stateFunc {
l.isInline = false
if l.hasPrefix(l.currentRightShortcodeDelim()) {
return lexShortcodeRightDelim
}
switch r := l.next(); {
case isSpace(r):
l.ignore()
default:
return l.errorf("unclosed shortcode")
}
return lexEndOfShortcode
}
// scans the elements inside shortcode tags
func lexInsideShortcode(l *pageLexer) stateFunc {
if l.hasPrefix(l.currentRightShortcodeDelim()) {
return lexShortcodeRightDelim
}
switch r := l.next(); {
case r == eof:
// eol is allowed inside shortcodes; this may go to end of document before it fails
return l.errorf("unclosed shortcode action")
case isSpace(r), isEndOfLine(r):
l.ignore()
case r == '=':
l.consumeSpace()
l.ignore()
peek := l.peek()
if peek == '"' || peek == '\\' {
return lexShortcodeQuotedParamVal(l, peek != '\\', tScParamVal)
} else if peek == '`' {
return lexShortCodeParamRawStringVal(l, tScParamVal)
}
return lexShortcodeParamVal
case r == '/':
if l.currShortcodeName == "" {
return l.errorf("got closing shortcode, but none is open")
}
l.closingState++
l.isInline = false
l.emit(tScClose)
case r == '\\':
l.ignore()
if l.peek() == '"' || l.peek() == '`' {
return lexShortcodeParam(l, true)
}
case l.elementStepNum > 0 && (isAlphaNumericOrHyphen(r) || r == '"' || r == '`'): // positional params can have quotes
l.backup()
return lexShortcodeParam(l, false)
case isAlphaNumeric(r):
l.backup()
return lexIdentifierInShortcode
default:
return l.errorf("unrecognized character in shortcode action: %#U. Note: Parameters with non-alphanumeric args must be quoted", r)
}
return lexInsideShortcode
}
func (l *pageLexer) currentLeftShortcodeDelimItem() ItemType {
return l.currLeftDelimItem
}
func (l *pageLexer) currentRightShortcodeDelimItem() ItemType {
return l.currRightDelimItem
}
func (l *pageLexer) currentLeftShortcodeDelim() []byte {
if l.currLeftDelimItem == tLeftDelimScWithMarkup {
return leftDelimScWithMarkup
}
return leftDelimScNoMarkup
}
func (l *pageLexer) currentRightShortcodeDelim() []byte {
if l.currRightDelimItem == tRightDelimScWithMarkup {
return rightDelimScWithMarkup
}
return rightDelimScNoMarkup
}