From e18edd3e249d8cc0e2f113f62ba5b017d590ec17 Mon Sep 17 00:00:00 2001
From: Vlasta Vesely <vlastavesely@protonmail.ch>
Date: Wed, 21 Apr 2021 21:22:11 +0200
Subject: [PATCH] preserve the zero-width joiner character

---
 tests/transcript.c |  2 ++
 tests/translit.c   |  8 ++++++++
 transcription.c    | 15 +++++++++------
 transliteration.c  | 37 ++++++++++++++++++++++++++++++-------
 4 files changed, 49 insertions(+), 13 deletions(-)

diff --git a/tests/transcript.c b/tests/transcript.c
index c62370d..93b5b24 100644
--- a/tests/transcript.c
+++ b/tests/transcript.c
@@ -29,6 +29,8 @@ START_TEST(test_transcript_devanagari_to_czech)
 	test_transcript("अग्निमीळे", "agnimílé");
 
 	test_transcript("तान्यजत्राँ", "tánjadžatrám");
+
+	test_transcript("श‍ृ शृ", "šr šr"); /* the zero width joiner */
 }
 END_TEST
 
diff --git a/tests/translit.c b/tests/translit.c
index 35c9720..ba85560 100644
--- a/tests/translit.c
+++ b/tests/translit.c
@@ -73,6 +73,13 @@ START_TEST(test_translit_candrabindu)
 }
 END_TEST
 
+START_TEST(test_translit_zero_width_joiner)
+{
+	test_transliterate_devanagari_to_latin("श‍ृ शृ", "ś‍ṛ śṛ");
+	test_transliterate_latin_to_devanagari("ś‍ṛ śṛ", "श‍ृ शृ");
+}
+END_TEST
+
 START_TEST(test_translit_detect_hindi)
 {
 	char *hindi = NULL;
@@ -87,5 +94,6 @@ void register_translit_tests(TCase *test_case)
 	tcase_add_test(test_case, test_translit_latin_to_devanagari);
 	tcase_add_test(test_case, test_translit_lla_sylable);
 	tcase_add_test(test_case, test_translit_candrabindu);
+	tcase_add_test(test_case, test_translit_zero_width_joiner);
 	tcase_add_test(test_case, test_translit_detect_hindi);
 }
diff --git a/transcription.c b/transcription.c
index 104f9a9..68049f2 100644
--- a/transcription.c
+++ b/transcription.c
@@ -6,10 +6,11 @@
 #include "transliteration.h"
 #include "utf8.h"
 
-#define SCHWA_CHARACTER 'a'
-#define VIRAMA    0x094d
-#define NUKTA     0x093c
-#define CHUNKSIZE 1024
+#define SCHWA_CHARACTER   'a'
+#define ZERO_WIDTH_JOINER 0x200d
+#define VIRAMA            0x094d
+#define NUKTA             0x093c
+#define CHUNKSIZE         1024
 
 static struct translit_letter table[] = {
 
@@ -189,13 +190,16 @@ int transcript_devanagari_to_czech(const char *devanagari, char **ret)
 		len = utf8_char_length(c);
 		src += len;
 
-		nasal_consonants_filter(latin, &done, prev, c);
+		if (c == ZERO_WIDTH_JOINER)
+			continue;
 
 		if (c == NUKTA) {
 			*ret = NULL;
 			return EHINDI;
 		}
 
+		nasal_consonants_filter(latin, &done, prev, c);
+
 		letter = letter_by_code(c);
 		if (letter) {
 			switch (letter->type) {
@@ -221,7 +225,6 @@ int transcript_devanagari_to_czech(const char *devanagari, char **ret)
 
 		end_of_word_filter(latin, &done, prev, c);
 
-
 		if (c == 0)
 			break;
 		prev = c;
diff --git a/transliteration.c b/transliteration.c
index e550d4f..6dfc08d 100644
--- a/transliteration.c
+++ b/transliteration.c
@@ -5,10 +5,11 @@
 #include "transliteration.h"
 #include "utf8.h"
 
-#define SCHWA_CHARACTER 'a'
-#define VIRAMA    0x094d
-#define NUKTA     0x093c
-#define CHUNKSIZE 1024
+#define SCHWA_CHARACTER   'a'
+#define ZERO_WIDTH_JOINER 0x200d
+#define VIRAMA            0x094d
+#define NUKTA             0x093c
+#define CHUNKSIZE         1024
 
 static struct translit_letter table[] = {
 
@@ -149,7 +150,7 @@ int transliterate_devanagari_to_latin(const char *devanagari, char **ret)
 				*(latin + done++) = SCHWA_CHARACTER;
 				break;
 			case VOWEL_SIGN:
-				if (done) {
+				if (done && *(latin + done - 1) == 'a') {
 					/* delete the inherent schwa */
 					done--;
 				}
@@ -159,6 +160,16 @@ int transliterate_devanagari_to_latin(const char *devanagari, char **ret)
 				break;
 			}
 		} else {
+			if (c == ZERO_WIDTH_JOINER) {
+				/*
+				 * the zero width joiner joins consonants
+				 * so the inherent schwa has to be removed.
+				 */
+				if (done && *(latin + done - 1) == 'a') {
+					done--;
+				}
+			}
+
 			utf8_pack_char(latin + done, c);
 			done += len;
 		}
@@ -204,7 +215,7 @@ static struct translit_letter *vowel_sign_by_data(const char *data)
 int transliterate_latin_to_devanagari(const char *latin, char **ret)
 {
 	struct translit_letter *letter, *next;
-	unsigned int alloc = 0, done = 0, len;
+	unsigned int alloc = 0, done = 0, c, len;
 	const char *src = latin;
 	char *devanagari = NULL;
 
@@ -240,6 +251,14 @@ int transliterate_latin_to_devanagari(const char *latin, char **ret)
 			done += len;
 			src += strlen(letter->data);
 
+			/* zero width joiner */
+			c = utf8_unpack_char(src);
+			if (c == ZERO_WIDTH_JOINER) {
+				utf8_pack_char(devanagari + done, ZERO_WIDTH_JOINER);
+				src += 3;
+				done += 3;
+			}
+
 			if (letter->type == VOWEL || letter->type == CODA)
 				continue;
 encode_vowel_modifier:
@@ -259,7 +278,11 @@ encode_vowel_modifier:
 				}
 			}
 		} else {
-			devanagari[done++] = *src++;
+			c = utf8_unpack_char(src);
+			len = utf8_char_length(c);
+			utf8_pack_char(devanagari + done, c);
+			done += len;
+			src += len;
 		}
 	}