Support blockquotes in formatter

1 files changed, 186 insertions, 45 deletions

diff --git a/telegram/formatter/formatter.go b/telegram/formatter/formatter.go
index 740fa09..9403198 100644
--- a/telegram/formatter/formatter.go
+++ b/telegram/formatter/formatter.go
@@ -8,15 +8,29 @@ import (
 	"github.com/zelenin/go-tdlib/client"
 )
 
-// Insertion is a piece of text in given position
-type Insertion struct {
+type insertionType int
+const (
+	insertionOpening insertionType = iota
+	insertionClosing
+	insertionUnpaired
+)
+
+type MarkupModeType int
+const (
+	MarkupModeXEP0393 MarkupModeType = iota
+	MarkupModeMarkdown
+)
+
+// insertion is a piece of text in given position
+type insertion struct {
 	Offset int32
 	Runes  []rune
+	Type   insertionType
 }
 
-// InsertionStack contains the sequence of insertions
+// insertionStack contains the sequence of insertions
 // from the start or from the end
-type InsertionStack []*Insertion
+type insertionStack []*insertion
 
 var boldRunesMarkdown = []rune("**")
 var boldRunesXEP0393 = []rune("*")
@@ -26,11 +40,16 @@ var strikeRunesXEP0393 = []rune("~")
 var codeRunes = []rune("`")
 var preRuneStart = []rune("```\n")
 var preRuneEnd = []rune("\n```")
+var quoteRunes = []rune("> ")
+var newlineRunes = []rune("\n")
+var doubleNewlineRunes = []rune("\n\n")
+var newlineCode = rune(0x0000000a)
+var bmpCeil = rune(0x0000ffff)
 
 // rebalance pumps all the values until the given offset to current stack (growing
 // from start) from given stack (growing from end); should be called
 // before any insertions to the current stack at the given offset
-func (s InsertionStack) rebalance(s2 InsertionStack, offset int32) (InsertionStack, InsertionStack) {
+func (s insertionStack) rebalance(s2 insertionStack, offset int32) (insertionStack, insertionStack) {
 	for len(s2) > 0 && s2[len(s2)-1].Offset <= offset {
 		s = append(s, s2[len(s2)-1])
 		s2 = s2[:len(s2)-1]
@@ -41,10 +60,10 @@ func (s InsertionStack) rebalance(s2 InsertionStack, offset int32) (InsertionSta
 
 // NewIterator is a second order function that sequentially scans and returns
 // stack elements; starts returning nil when elements are ended
-func (s InsertionStack) NewIterator() func() *Insertion {
+func (s insertionStack) NewIterator() func() *insertion {
 	i := -1
 
-	return func() *Insertion {
+	return func() *insertion {
 		i++
 		if i < len(s) {
 			return s[i]
@@ -120,21 +139,10 @@ func MergeAdjacentEntities(entities []*client.TextEntity) []*client.TextEntity {
 }
 
 // ClaspDirectives to the following span as required by XEP-0393
-func ClaspDirectives(text string, entities []*client.TextEntity) []*client.TextEntity {
+func ClaspDirectives(doubledRunes []rune, entities []*client.TextEntity) []*client.TextEntity {
 	alignedEntities := make([]*client.TextEntity, len(entities))
 	copy(alignedEntities, entities)
 
-	// transform the source text into a form with uniform runes and code points,
-	// by duplicating the Basic Multilingual Plane
-	doubledRunes := make([]rune, 0, len(text)*2)
-
-	for _, cp := range text {
-		if cp > 0x0000ffff {
-			doubledRunes = append(doubledRunes, cp, cp)
-		} else {
-			doubledRunes = append(doubledRunes, cp)
-		}
-	}
 	for i, entity := range alignedEntities {
 		var dirty bool
 		endOffset := entity.Offset + entity.Length
@@ -167,18 +175,89 @@ func ClaspDirectives(text string, entities []*client.TextEntity) []*client.TextE
 	return alignedEntities
 }
 
-func markupBraces(entity *client.TextEntity, lbrace, rbrace []rune) (*Insertion, *Insertion) {
-	return &Insertion{
+func markupBraces(entity *client.TextEntity, lbrace, rbrace []rune) []*insertion {
+	return []*insertion{
+		&insertion{
 			Offset: entity.Offset,
 			Runes:  lbrace,
-		}, &Insertion{
+			Type:   insertionOpening,
+		},
+		&insertion{
 			Offset: entity.Offset + entity.Length,
 			Runes:  rbrace,
+			Type:   insertionClosing,
+		},
+	}
+}
+
+func quotePrependNewlines(entity *client.TextEntity, doubledRunes []rune, markupMode MarkupModeType) []*insertion {
+	if len(doubledRunes) == 0 {
+		return []*insertion{}
+	}
+
+	startRunes := []rune("\n> ")
+	if entity.Offset == 0 || doubledRunes[entity.Offset-1] == newlineCode {
+		startRunes = quoteRunes
+	}
+	insertions := []*insertion{
+		&insertion{
+			Offset: entity.Offset,
+			Runes:  startRunes,
+			Type:   insertionUnpaired,
+		},
+	}
+
+	entityEnd := entity.Offset + entity.Length
+	entityEndInt := int(entityEnd)
+
+	var wasNewline bool
+	// last newline is omitted, there's no need to put quote mark after the quote
+	for i := entity.Offset; i < entityEnd-1; i++ {
+		isNewline := doubledRunes[i] == newlineCode
+		if (isNewline && markupMode == MarkupModeXEP0393) || (wasNewline && isNewline && markupMode == MarkupModeMarkdown) {
+			insertions = append(insertions, &insertion{
+				Offset: i+1,
+				Runes:  quoteRunes,
+				Type:   insertionUnpaired,
+			})
+		}
+
+		if isNewline {
+			wasNewline = true
+		} else {
+			wasNewline = false
 		}
+	}
+
+	var rbrace []rune
+	if len(doubledRunes) > entityEndInt {
+		if doubledRunes[entityEnd] == newlineCode {
+			if markupMode == MarkupModeMarkdown && len(doubledRunes) > entityEndInt+1 && doubledRunes[entityEndInt+1] != newlineCode {
+				rbrace = newlineRunes
+			}
+		} else {
+			if markupMode == MarkupModeMarkdown {
+				rbrace = doubleNewlineRunes
+			} else {
+				rbrace = newlineRunes
+			}
+		}
+	}
+	insertions = append(insertions, &insertion{
+		Offset: entityEnd,
+		Runes:  rbrace,
+		Type:   insertionClosing,
+	})
+
+	return insertions
 }
 
-// EntityToMarkdown generates the wrapping Markdown tags
-func EntityToMarkdown(entity *client.TextEntity) (*Insertion, *Insertion) {
+// entityToMarkdown generates the wrapping Markdown tags
+func entityToMarkdown(entity *client.TextEntity, doubledRunes []rune, markupMode MarkupModeType) []*insertion {
+	if entity == nil || entity.Type == nil {
+		return []*insertion{}
+	}
+
 	switch entity.Type.TextEntityTypeType() {
 	case client.TypeTextEntityTypeBold:
 		return markupBraces(entity, boldRunesMarkdown, boldRunesMarkdown)
@@ -193,18 +272,20 @@ func EntityToMarkdown(entity *client.TextEntity) (*Insertion, *Insertion) {
 	case client.TypeTextEntityTypePreCode:
 		preCode, _ := entity.Type.(*client.TextEntityTypePreCode)
 		return markupBraces(entity, []rune("\n```"+preCode.Language+"\n"), codeRunes)
+	case client.TypeTextEntityTypeBlockQuote:
+		return quotePrependNewlines(entity, doubledRunes, MarkupModeMarkdown)
 	case client.TypeTextEntityTypeTextUrl:
 		textURL, _ := entity.Type.(*client.TextEntityTypeTextUrl)
 		return markupBraces(entity, []rune("["), []rune("]("+textURL.Url+")"))
 	}
 
-	return nil, nil
+	return []*insertion{}
 }
 
-// EntityToXEP0393 generates the wrapping XEP-0393 tags
-func EntityToXEP0393(entity *client.TextEntity) (*Insertion, *Insertion) {
+// entityToXEP0393 generates the wrapping XEP-0393 tags
+func entityToXEP0393(entity *client.TextEntity, doubledRunes []rune, markupMode MarkupModeType) []*insertion {
 	if entity == nil || entity.Type == nil {
-		return nil, nil
+		return []*insertion{}
 	}
 
 	switch entity.Type.TextEntityTypeType() {
@@ -221,29 +302,55 @@ func EntityToXEP0393(entity *client.TextEntity) (*Insertion, *Insertion) {
 	case client.TypeTextEntityTypePreCode:
 		preCode, _ := entity.Type.(*client.TextEntityTypePreCode)
 		return markupBraces(entity, []rune("\n```"+preCode.Language+"\n"), codeRunes)
+	case client.TypeTextEntityTypeBlockQuote:
+		return quotePrependNewlines(entity, doubledRunes, MarkupModeXEP0393)
 	case client.TypeTextEntityTypeTextUrl:
 		textURL, _ := entity.Type.(*client.TextEntityTypeTextUrl)
 		// non-standard, Pidgin-specific
 		return markupBraces(entity, []rune{}, []rune(" <"+textURL.Url+">"))
 	}
 
-	return nil, nil
+	return []*insertion{}
+}
+
+// transform the source text into a form with uniform runes and code points,
+// by duplicating anything beyond the Basic Multilingual Plane
+func textToDoubledRunes(text string) []rune {
+	doubledRunes := make([]rune, 0, len(text)*2)
+	for _, cp := range text {
+		if cp > bmpCeil {
+			doubledRunes = append(doubledRunes, cp, cp)
+		} else {
+			doubledRunes = append(doubledRunes, cp)
+		}
+	}
+
+	return doubledRunes
 }
 
 // Format traverses an already sorted list of entities and wraps the text in a markup
 func Format(
 	sourceText string,
 	entities []*client.TextEntity,
-	entityToMarkup func(*client.TextEntity) (*Insertion, *Insertion),
+	markupMode MarkupModeType,
 ) string {
 	if len(entities) == 0 {
 		return sourceText
 	}
 
-	mergedEntities := SortEntities(ClaspDirectives(sourceText, MergeAdjacentEntities(SortEntities(entities))))
+	var entityToMarkup func(*client.TextEntity, []rune, MarkupModeType) []*insertion
+	if markupMode == MarkupModeXEP0393 {
+		entityToMarkup = entityToXEP0393
+	} else {
+		entityToMarkup = entityToMarkdown
+	}
 
-	startStack := make(InsertionStack, 0, len(sourceText))
-	endStack := make(InsertionStack, 0, len(sourceText))
+	doubledRunes := textToDoubledRunes(sourceText)
+
+	mergedEntities := SortEntities(ClaspDirectives(doubledRunes, MergeAdjacentEntities(SortEntities(entities))))
+
+	startStack := make(insertionStack, 0, len(sourceText))
+	endStack := make(insertionStack, 0, len(sourceText))
 
 	// convert entities to a stack of brackets
 	var maxEndOffset int32
@@ -260,36 +367,70 @@ func Format(
 
 		startStack, endStack = startStack.rebalance(endStack, entity.Offset)
 
-		startInsertion, endInsertion := entityToMarkup(entity)
-		if startInsertion != nil {
-			startStack = append(startStack, startInsertion)
+		insertions := entityToMarkup(entity, doubledRunes, markupMode)
+		if len(insertions) > 1 {
+			startStack = append(startStack, insertions[0:len(insertions)-1]...)
 		}
-		if endInsertion != nil {
-			endStack = append(endStack, endInsertion)
+		if len(insertions) > 0 {
+			endStack = append(endStack, insertions[len(insertions)-1])
 		}
 	}
 	// flush the closing brackets that still remain in endStack
 	startStack, endStack = startStack.rebalance(endStack, maxEndOffset)
+	// sort unpaired insertions
+	sort.SliceStable(startStack, func(i int, j int) bool {
+		ins1 := startStack[i]
+		ins2 := startStack[j]
+		if ins1.Type == insertionUnpaired && ins2.Type == insertionUnpaired {
+			return ins1.Offset < ins2.Offset
+		}
+		if ins1.Type == insertionUnpaired {
+			if ins1.Offset == ins2.Offset {
+				if ins2.Type == insertionOpening { // > **
+					return true
+				} else if ins2.Type == insertionClosing { // **> 
+					return false
+				}
+			} else {
+				return ins1.Offset < ins2.Offset
+			}
+		}
+		if ins2.Type == insertionUnpaired {
+			if ins1.Offset == ins2.Offset {
+				if ins1.Type == insertionOpening { // > **
+					return false
+				} else if ins1.Type == insertionClosing { // **> 
+					return true
+				}
+			} else {
+				return ins1.Offset < ins2.Offset
+			}
+		}
+		return false
+	})
 
 	// merge brackets into text
 	markupRunes := make([]rune, 0, len(sourceText))
 
 	nextInsertion := startStack.NewIterator()
 	insertion := nextInsertion()
-	var runeI int32
+	var skipNext bool
 
-	for _, cp := range sourceText {
-		for insertion != nil && insertion.Offset <= runeI {
+	for i, cp := range doubledRunes {
+		if skipNext {
+			skipNext = false
+			continue
+		}
+
+		for insertion != nil && int(insertion.Offset) <= i {
 			markupRunes = append(markupRunes, insertion.Runes...)
 			insertion = nextInsertion()
 		}
 
 		markupRunes = append(markupRunes, cp)
 		// skip two UTF-16 code units (not points actually!) if needed
-		if cp > 0x0000ffff {
-			runeI += 2
-		} else {
-			runeI++
+		if cp > bmpCeil {
+			skipNext = true
 		}
 	}
 	for insertion != nil {