bug#24603: [PATCHv5 10/11] Implement casing rules for Lithuanian (bug#24603)

[Michal Nazarewicz]

In Lithuanian, tittle above lower case i and j are retained even if
there are other diacritics above present.  For that to work, an explicit
combining dot above must be added after i and j or otherwise the
rendering engine will remove the tittle.

* src/casefiddle.c (struct casing_context, prepare_casing_context): Add
SPECIAL_LT, SPECIAL_LT_DEL_DOT_ABOVE and SPECIAL_LT_INS_DOT_ABOVE
special flag valus for handling Lithuanian.  Set the flag to SPECIAL_LT
if buffer is in Lithuanian.
(maybe_case_lithuanian): New function which implements Lithuanian rules.
(case_characters): Make use of maybe_case_lithuanian.

* test/src/casefiddle-tests.el (casefiddle-tests-casing): Add test cases
for Lithuanian rules.
---
 src/casefiddle.c             | 178 ++++++++++++++++++++++++++++++++++++++++---
 test/src/casefiddle-tests.el |  27 ++++++-
 2 files changed, 195 insertions(+), 10 deletions(-)

diff --git a/src/casefiddle.c b/src/casefiddle.c
index 4785ebaddc4..a33bac7d21e 100644
--- a/src/casefiddle.c
+++ b/src/casefiddle.c
@@ -77,7 +77,15 @@ struct casing_context {
     SPECIAL_NL_UPCASE_J,
 
     /* Handle Azerbaijani and Turkish dotted and dotless i. */
-    SPECIAL_TR
+    SPECIAL_TR,
+
+    /* Apply Lithuanian rules for i’s and j’s tittle. */
+    SPECIAL_LT,
+    /* As above plus look out for combining dot above to delete. */
+    SPECIAL_LT_DEL_DOT_ABOVE,
+    /* As above plus look out for diacritics combining above because
+       we may need to inject dot above before them. */
+    SPECIAL_LT_INS_DOT_ABOVE,
   } special;
 };
 
@@ -116,6 +124,9 @@ prepare_casing_context (struct casing_context *ctx,
     case ('t' << 8) | 'r':  /* Turkish */
     case ('a' << 8) | 'z':  /* Azerbaijani */
       ctx->special = SPECIAL_TR;
+      break;
+    case ('l' << 8) | 't':  /* Lithuania */
+      ctx->special = SPECIAL_LT;
     }
 }
 
@@ -362,6 +373,142 @@ maybe_case_turkic (struct casing_str_buf *buf, struct 
casing_context *ctx,
     return ch == cased ? RES_NO_CHANGE : RES_CHANGED;
 }
 
+/* Lithuanian retains tittle in lower case i and j when there are more
+   accents above those letters. */
+
+#define CAPITAL_I_WITH_GRAVE  0x0CC
+#define CAPITAL_I_WITH_ACUTE  0x0CD
+#define CAPITAL_I_WITH_TILDE  0x128
+#define CAPITAL_I_WITH_OGONEK 0x12E
+#define SMALL_I_WITH_OGONEK   0x12F
+#define COMBINING_GRAVE_ABOVE 0x300
+#define COMBINING_ACUTE_ABOVE 0x301
+#define COMBINING_TILDE_ABOVE 0x303
+#define COMBINING_OGONEK      0x328
+
+/* Save in BUF result of casing character CH if Lithuanian casing rules apply.
+
+   If not-NULL, NEXT points to the next character in the cased string.  If 
NULL,
+   it is assumed current character is the last one being cased.  This is used 
to
+   apply some rules which depend on proceeding state.
+
+   FLAG is a normalised flag (as returned by normalise_flag function).
+
+   Return -2 (RES_NOT_TOUCHED) if Lithuanian rules did not apply, no changes
+   were made and other casing rules should be tried.  Otherwise, meaning of
+   return values is the same as in case_characters function. */
+static int
+maybe_case_lithuanian (struct casing_str_buf *buf, struct casing_context *ctx,
+                      enum case_action flag, int ch)
+{
+  switch (ctx->special) {
+  case SPECIAL_LT:
+    break;
+
+  case SPECIAL_LT_DEL_DOT_ABOVE:
+    /* When upper-casing i or j, a combining dot above that follows it must be
+       removed.  This is true even if there’s a combining ogonek in between.
+       But, if there’s another character combining above in between, combining
+       dot needs to stay (since the dot will be rendered above the other
+       diacritic). */
+    switch (ch) {
+    case COMBINING_DOT_ABOVE:
+      buf->len_chars = buf->len_bytes = 0;
+      ctx->special = SPECIAL_LT;
+      return RES_CHANGED;
+    case COMBINING_GRAVE_ABOVE:
+    case COMBINING_ACUTE_ABOVE:
+    case COMBINING_TILDE_ABOVE:
+      ctx->special = SPECIAL_LT;
+      return RES_NOT_TOUCHED;
+    case COMBINING_OGONEK:
+      return RES_NOT_TOUCHED;
+    default:
+      ctx->special = SPECIAL_LT;
+    }
+    break;
+
+  case SPECIAL_LT_INS_DOT_ABOVE:
+    /* When lower-casing I or J, if the letter has any accents above,
+       a combining dot above must be added before them.  If we are here, it
+       means that we have lower cased I or J and we’re now on the lookout for
+       accents combining above. */
+    switch (ch) {
+    case COMBINING_GRAVE_ABOVE:
+    case COMBINING_ACUTE_ABOVE:
+    case COMBINING_TILDE_ABOVE:
+      buf->len_chars = 2;
+      buf->len_bytes = CHAR_STRING (COMBINING_DOT_ABOVE, buf->data);
+      buf->len_bytes += CHAR_STRING (ch, buf->data + buf->len_bytes);
+      ctx->special = SPECIAL_LT;
+      return RES_CHANGED;
+    case COMBINING_OGONEK:
+      return RES_NOT_TOUCHED;
+    default:
+      ctx->special = SPECIAL_LT;
+    }
+    break;
+
+  default:
+    return RES_NOT_TOUCHED;
+  }
+
+  switch (flag) {
+  case CASE_UP:
+  case CASE_CAPITALIZE:
+    if (ch == 'i' || ch == 'j')
+      {
+       buf->data[0] = ch ^ ('i' ^ 'I');
+       buf->len_bytes = 1;
+      }
+    else if (ch == SMALL_I_WITH_OGONEK)
+      buf->len_bytes = CHAR_STRING (CAPITAL_I_WITH_OGONEK, buf->data);
+    else
+      break;
+    buf->len_chars = 1;
+    /* Change the state so we’re on the lookout for combining dot above. */
+    ctx->special = SPECIAL_LT_DEL_DOT_ABOVE;
+    return RES_CHANGED;
+
+  case CASE_DOWN:
+    /* Turning I or J to lower case requires combining dot above to be included
+       IF there are any other characters combining above present.  This is so
+       that the tittle is preserved. */
+    switch (ch) {
+    case CAPITAL_I_WITH_GRAVE:
+      ch = 0x80;  /* U+300, "\xCC\x80", combining grave accent */
+      goto has_accent;
+    case CAPITAL_I_WITH_ACUTE:
+      ch = 0x81;  /* U+301, "\xCC \x81", combining acute accent */
+      goto has_accent;
+    case CAPITAL_I_WITH_TILDE:
+      ch = 0x83;  /* U+303, "\xCC\x83", combining tilde */
+    has_accent:
+      memcpy (buf->data, "i\xCC\x87\xCC", 4);
+      buf->data[4] = ch;
+      buf->len_chars = 3;
+      buf->len_bytes = 5;
+      return RES_CHANGED;
+
+    case 'I':
+    case 'J':
+      buf->data[0] = ch ^ ('i' ^ 'I');
+      buf->len_bytes = 1;
+      if (false)
+    case CAPITAL_I_WITH_OGONEK:
+       buf->len_bytes = CHAR_STRING (SMALL_I_WITH_OGONEK, buf->data);
+      buf->len_chars = 1;
+      /* Change the state so we’re on the lookout for diacritics combining
+        above.  If one is found, we need to add combining dot above. */
+      ctx->special = SPECIAL_LT_INS_DOT_ABOVE;
+      return RES_CHANGED;
+    }
+    break;
+  }
+
+  return RES_NOT_TOUCHED;
+}
+
 /* Save in BUF result of casing character CH.
 
    If not-NULL, NEXT points to the next character in the cased string.  If 
NULL,
@@ -381,17 +528,30 @@ case_characters (struct casing_str_buf *buf, struct 
casing_context *ctx,
                 int ch, const unsigned char *next)
 {
   enum case_action flag = normalise_flag (ctx);
-  int ret;
+  int ret = RES_NOT_TOUCHED;
+
+  switch (ctx->special) {
+  case SPECIAL_NONE:
+    break;
+
+  case SPECIAL_TR:
+    ret = maybe_case_turkic (buf, ctx, flag, ch, next);
+    break;
+
+  default:
+  /* case SPECIAL_LT: */
+  /* case SPECIAL_LT_DEL_DOT_ABOVE: */
+  /* case SPECIAL_LT_INS_DOT_ABOVE: */
+    ret = maybe_case_lithuanian (buf, ctx, flag, ch);
+  }
 
-  ret = maybe_case_turkic (buf, ctx, flag, ch, next);
-  if (ret != RES_NOT_TOUCHED)
-    return ret;
+  if (ret == RES_NOT_TOUCHED)
+    ret = maybe_case_greek (buf, ctx, flag, ch, next);
 
-  ret = maybe_case_greek (buf, ctx, flag, ch, next);
-  if (ret != RES_NOT_TOUCHED)
-    return ret;
+  if (ret == RES_NOT_TOUCHED)
+    ret = case_character_impl (buf, ctx, flag, ch);
 
-  return case_character_impl (buf, ctx, flag, ch);
+  return ret;
 }
 
 static Lisp_Object
diff --git a/test/src/casefiddle-tests.el b/test/src/casefiddle-tests.el
index ce1bb18dd40..f7b0da41029 100644
--- a/test/src/casefiddle-tests.el
+++ b/test/src/casefiddle-tests.el
@@ -241,7 +241,32 @@ casefiddle-tests--test-casing
         ("I\u0307si\u0307s" "I\u0307Sİ\u0307S" "isi\u0307s"
          "I\u0307si\u0307s" "I\u0307si\u0307s" "tr")
         ("I\u0307sI\u0307s" "I\u0307SI\u0307S" "isis"
-         "I\u0307sis" "I\u0307sI\u0307s" "tr"))))))
+         "I\u0307sis" "I\u0307sI\u0307s" "tr")
+
+        ;; Test combining dot above in inserted when needed when lower
+        ;; casing I or J.
+        ("I\u0328\u0300"          ; I + ogonek + grave
+         "I\u0328\u0300" "i\u0328\u0307\u0300"
+         "I\u0328\u0300" "I\u0328\u0300" "lt")
+
+        ("J\u0328\u0300"          ; J + ogonek + grave
+         "J\u0328\u0300" "j\u0328\u0307\u0300"
+         "J\u0328\u0300" "J\u0328\u0300" "lt")
+
+        ("Į\u0300"          ; I-ogonek + grave
+         "Į\u0300" "į\u0307\u0300" "Į\u0300" "Į\u0300" "lt")
+
+        ("Ì Í Ĩ"
+         "Ì Í Ĩ" "i\u0307\u0300 i\u0307\u0301 i\u0307\u0303"
+         "Ì Í Ĩ" "Ì Í Ĩ" "lt")
+
+        ;; Test combining dot above in removed when upper casing i or j.
+        ("i\u0328\u0307"          ; i + ogonek + dot above
+         "I\u0328" "i\u0328\u0307" "I\u0328" "I\u0328" "lt")
+        ("j\u0328\u0307"          ; j + ogonek + dot above
+         "J\u0328" "j\u0328\u0307" "J\u0328" "J\u0328" "lt")
+        ("į\u0307"                ; i-ogonek + dot above
+         "Į" "į\u0307" "Į" "Į" "lt"))))))
 
 (ert-deftest casefiddle-tests-casing-byte8 ()
   (should-not
-- 
2.12.0.246.ga2ecc84866-goog