devanagari ligation wip

src/net/torvald/terrarumsansbitmap/gdx/TerrarumSansBitmap.kt

@@ -1035,17 +1035,30 @@ class TerrarumSansBitmap(
 
     // basically an Unicode NFD with some additional flavours
     private fun CodepointSequence.normalise(): CodepointSequence {
-        val dis = this.utf16to32()
         val seq = CodepointSequence()
         val seq2 = CodepointSequence()
 
         val yankedCharacters  = Stack<Pair<Int, CodePoint>>() // Stack of <Position, CodePoint>; codepoint use -1 if not applicable
 
+        fun emptyOutYanked() {
+            while (!yankedCharacters.empty()) {
+                val poppedChar = yankedCharacters.pop()
+                if (poppedChar.second == DEVANAGARI_RA)
+                    seq.add(DEVANAGARI_RA_SUPER)
+                else
+                    seq.add(yankedCharacters.pop().second)
+            }
+        }
+
         var i = 0
+        this.utf16to32().let { dis ->
         while (i < dis.size) {
-            val c = dis[i]
+            val cPrev2 = dis.getOrElse(i-2) { -1 }
             val cPrev = dis.getOrElse(i-1) { -1 }
+            val c = dis[i]
             val cNext = dis.getOrElse(i+1) { -1 }
+            val cNext2 = dis.getOrElse(i+2) { -1 }
+            // can't use regular sliding window as the 'i' value is changed way too often
 
             // LET THE NORMALISATION BEGIN //
 
@@ -1137,11 +1150,26 @@ class TerrarumSansBitmap(
             // END of tamil subsystem implementation
 
             // BEGIN of devanagari string replacer
-            else if (c == DEVANAGARI_VIRAMA) {
+            // Unicode Devanagari Rendering Rule R6-R8
-                yankedCharacters.push(i-1 to cPrev)
+            // (this must precede the ligaturing-machine coded on the 2nd pass, otherwise the rules below will cause undesirable effects)
+            else if (devanagariConsonants.contains(c) && cNext == DEVANAGARI_VIRAMA && cNext2 == DEVANAGARI_RA) {
+                seq.addAll(toRaAppended(c))
+                i += 2
             }
-            else if (c == DEVANAGARI_RA) {
+            // Unicode Devanagari Rendering Rule R5
+            else if (c == DEVANAGARI_RRA && cNext == DEVANAGARI_VIRAMA || c == DEVANAGARI_RA && cNext == DEVANAGARI_VIRAMA && cNext2 == ZWJ) {
+                seq.add(DEVANAGARI_EYELASH_RA)
+                i += 1
+            }
+            // Unicode Devanagari Rendering Rule R2-R4
+            else if (c == DEVANAGARI_RA && cNext == DEVANAGARI_VIRAMA && cNext2 != DEVANAGARI_RA) {
                 yankedCharacters.push(i to c)
+                i += 1
+            }
+            // Unicode Devanagari Rendering Rule R2-R4
+            else if (!isDevanagari(c) && !yankedCharacters.empty()) {
+                emptyOutYanked()
+                seq.add(c)
             }
             // WIP
             // END of devanagari string replacer
@@ -1164,7 +1192,11 @@ class TerrarumSansBitmap(
 
             i++
         }
+        emptyOutYanked()
+        }
 
+
+        // second scan
         // swap position of {letter, diacritics that comes before the letter}
         i = 1
         while (i <= seq.lastIndex) {
@@ -1175,6 +1207,27 @@ class TerrarumSansBitmap(
                 seq[i] = t
             }
 
+            val cPrev2 = seq.getOrElse(i-2) { -1 }
+            val cPrev = seq.getOrElse(i-1) { -1 }
+            val c = seq[i]
+
+            // BEGIN of Devanagari String Replacer 2 (lookbehind type)
+            // Devanagari Ligations (Lookbehind)
+            if (devanagariConsonants.contains(cPrev2) && cPrev == DEVANAGARI_VIRAMA && devanagariConsonants.contains(c)) {
+                i -= 2
+
+                repeat(3) { seq.removeAt(i) }
+
+                val ligature = ligateIndicConsonants(cPrev2, c)
+                ligature.forEachIndexed { index, char ->
+                    seq.add(i + index, char)
+                }
+
+                i += ligature.size
+            }
+            // END of Devanagari String Replacer 2
+
+
             i++
         }
 
@@ -1630,13 +1683,130 @@ class TerrarumSansBitmap(
             'j'.toInt() to 0x237
         )
 
-        private val tamilLigatingConsonants = listOf('க','ங','ச','ஞ','ட','ண','த','ந','ன','ப','ம','ய','ர','ற','ல','ள','ழ','வ').map { it.toInt() }.toIntArray()
+        private val ZWJ = 0x200D
 
+        private val tamilLigatingConsonants = listOf('க','ங','ச','ஞ','ட','ண','த','ந','ன','ப','ம','ய','ர','ற','ல','ள','ழ','வ').map { it.toInt() }.toIntArray()
         private val TAMIL_KSSA = 0xF00ED
         private val TAMIL_SHRII = 0xF00EE
 
+        private val devanagariConsonants = ((0x0915..0x0939) + (0x0958..0x095F) + (0x0978..0x097F) + (0xF0140..0xF01FF)).toIntArray()
+
+        private val devanagariBaseConsonants = 0x0915..0x0939
+        private val devanagariBaseConsonantsWithNukta = 0x0958..0x095F
+        private val devanagariBaseConsonantsExtended = 0x0978..0x097F
+        private val devanagariPresentationFormsConsonants = 0xF0140..0xF01FF
+
         private val DEVANAGARI_VIRAMA = 0x94D
         private val DEVANAGARI_RA = 0x930
+        private val DEVANAGARI_RRA = 0x931
+        private val DEVANAGARI_RA_SUPER = 0xF0104
+        private val DEVANAGARI_EYELASH_RA = 0xF012A
+
+        private val DEVANAGARI_LIG_K_SS = 0xF0181
+        private val DEVANAGARI_LIG_J_NY = 0xF0184
+        private val DEVANAGARI_LIG_T_T = 0xF018B
+        private val DEVANAGARI_LIG_T_R = 0xF0154
+        private val DEVANAGARI_LIG_SH_R = 0xF0166
+        private val DEVANAGARI_HALFLIG_K_SS = 0xF012B
+        private val DEVANAGARI_HALFLIG_J_NY = 0xF012C
+        private val DEVANAGARI_HALFLIG_T_T = 0xF012D
+        private val DEVANAGARI_HALFLIG_T_R = 0xF012E
+        private val DEVANAGARI_HALFLIG_SH_R = 0xF012F
+
+        private val DEVANAGARI_SYLL_RU = 0xF0100
+        private val DEVANAGARI_SYLL_RUU = 0xF0101
+
+        private val DEVANAGARI_HALF_FORMS = 0xF0100 // starting point for Devanagari half forms
+        private val DEVANAGARI_LIG_X_R = 0xF0140 // starting point for Devanagari ligature CONSONANT+RA
+
+        private fun CodePoint.toHalfFormOrNull(): CodePoint? {
+            if (this in devanagariBaseConsonants) return (this - 0x0910 + DEVANAGARI_HALF_FORMS)
+            if (this in devanagariBaseConsonantsWithNukta) return (this - 0x0920 + DEVANAGARI_HALF_FORMS)
+            else if (this == DEVANAGARI_LIG_K_SS) return DEVANAGARI_HALFLIG_K_SS
+            else if (this == DEVANAGARI_LIG_J_NY) return DEVANAGARI_HALFLIG_J_NY
+            else if (this == DEVANAGARI_LIG_T_T) return  DEVANAGARI_HALFLIG_T_T
+            else if (this == DEVANAGARI_LIG_T_R) return  DEVANAGARI_HALFLIG_T_R
+            else if (this == DEVANAGARI_LIG_SH_R) return DEVANAGARI_HALFLIG_SH_R
+            // TODO half forms of X_R-ligatures
+            else return null
+        }
+
+        // TODO use proper version of Virama for respective scripts
+        private fun CodePoint.toHalfFormOrVirama(): List<CodePoint> = this.toHalfFormOrNull().let {
+                if (it == null) listOf(this, DEVANAGARI_VIRAMA) else listOf(it)
+            }
+
+        // TODO use proper version of Virama for respective scripts
+        private fun toRaAppended(c: CodePoint): List<CodePoint> {
+            if (c in devanagariBaseConsonants) return listOf(c - 0x0910 + DEVANAGARI_LIG_X_R)
+            else return listOf(c, DEVANAGARI_VIRAMA, DEVANAGARI_RA)
+        }
+
+        private fun ligateIndicConsonants(c1: CodePoint, c2: CodePoint): List<CodePoint> {
+            if (c2 == DEVANAGARI_RA) return toRaAppended(c1) // Devanagari @.RA
+            when (c1) {
+                0x0915 -> /* Devanagari KA */ when (c2) {
+                    0x0924 -> return listOf(0xF0180) // K.TA
+                    0x0937 -> return listOf(DEVANAGARI_LIG_K_SS) // K.SSA
+                    else -> return c1.toHalfFormOrVirama() + c2
+                }
+                0x0919 -> /* Devanagari NGA */ when (c2) {
+                    0x0917 -> return listOf(0xF0182) // NG.G
+                    0x092E -> return listOf(0xF0183) // NG.M
+                    else -> return c1.toHalfFormOrVirama() + c2
+                }
+                0x091C -> /* Devanagari JA */ when (c2) {
+                    0x091E -> return listOf(DEVANAGARI_LIG_J_NY) // J.NY
+                    else -> return c1.toHalfFormOrVirama() + c2
+                }
+                0x091F -> /* Devanagari TTA */ when (c2) {
+                    0x091F -> return listOf(0xF0185) // TT.TT
+                    0x0920 -> return listOf(0xF0186) // TT.TTH
+                    else -> return c1.toHalfFormOrVirama() + c2
+                }
+                0x0920 -> /* Devanagari TTHA */ when (c2) {
+                    0x0920 -> return listOf(0xF0187) // TTH.TTH
+                    else -> return c1.toHalfFormOrVirama() + c2
+                }
+                0x0921 -> /* Devanagari DDA */ when (c2) {
+                    0x0921 -> return listOf(0xF0188) // DD.DD
+                    0x0922 -> return listOf(0xF0189) // DD.DDH
+                    else -> return c1.toHalfFormOrVirama() + c2
+                }
+                0x0922 -> /* Devanagari DDHA */ when (c2) {
+                    0x0922 -> return listOf(0xF018A) // DDH.DDH
+                    else -> return c1.toHalfFormOrVirama() + c2
+                }
+                0x0924 -> /* Devanagari TA */ when (c2) {
+                    0x0924 -> return listOf(DEVANAGARI_LIG_T_T) // T.T
+                    else -> return c1.toHalfFormOrVirama() + c2
+                }
+                0x0926 -> /* Devanagari DA */ when (c2) {
+                    0x0926 -> return listOf(0xF018C) // D.D
+                    0x0927 -> return listOf(0xF018D) // D.DH
+                    0x092C -> return listOf(0xF018E) // D.B
+                    0x092D -> return listOf(0xF018F) // D.BH
+                    0x092E -> return listOf(0xF0190) // D.M
+                    0x092F -> return listOf(0xF0191) // D.Y
+                    0x0935 -> return listOf(0xF0192) // D.V
+                    else -> return c1.toHalfFormOrVirama() + c2
+                }
+                0x0928 -> /* Devanagari NA */ when (c2) {
+                    0x0928 -> return listOf(0xF0193) // N.N
+                    else -> return c1.toHalfFormOrVirama() + c2
+                }
+                0x0939 -> /* Devanagari HA */ when (c2) {
+                    0x0923 -> return listOf(0xF0194) // H.NN
+                    0x0928 -> return listOf(0xF0195) // H.N
+                    0x092E -> return listOf(0xF0196) // H.M
+                    0x092F -> return listOf(0xF0197) // H.Y
+                    0x0932 -> return listOf(0xF0198) // H.L
+                    0x0935 -> return listOf(0xF0199) // H.v
+                    else -> return c1.toHalfFormOrVirama() + c2
+                }
+                else -> return c1.toHalfFormOrVirama() + c2 // TODO use proper version of Virama for respective scripts
+            }
+        }
 
 
         private fun Int.toHex() = "U+${this.toString(16).padStart(4, '0').toUpperCase()}"

testtext.txt

@@ -1,17 +1,3 @@
 
- A ดุ ตี ปู่ พี่ ป่ ม่ ปั มั พีุ ทิ่ท่ท่ิ ปิ่ป่ป่ิ ทิ้ ปิ้ มำด มําด
+फ़ॊण्ज़ीणू क्की क्कि र्क र्त्सा र्र्ल ऱ्ड क्राप् ठ्री त्र् त्र्मोत्र्यो ज्ञज्ञ्रत्त न्न
-
+ह्णह्नह्मह्यह्लह्वह्र णनमयलव ज़्लतान
-이는 일본의 요미가나(読み仮名)와 비슷한 용법이다.
-
-Sugarさとう설탕砂糖
-
-বাংলাদেশ
-
-தமிழ் லூ லு
-
-QWனோT
-னோ
-
-எழுத்து வடிவங்களுக்கு வார்த்தைகள் மற்றும் வாக்கியங்கள் போலவே தொனி, ஒலி பண்பு, தன்மை உண்டு.
-
-க்ஷ ஶ்ரீ க்ஷௌ