From a59104191270b74d0a1312a6b83884ffb9334691 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?P=C5=99emysl=20Janouch?= <p.janouch@gmail.com>
Date: Mon, 26 Sep 2016 15:59:26 +0200
Subject: [PATCH] Highlight the longest matching prefix of entries

---
 src/sdtui.c    | 29 ++++++++++++++++++---
 src/stardict.c | 68 ++++++++++++++++++++++++++++++++++++++++++++++++++
 src/stardict.h |  3 +++
 3 files changed, 97 insertions(+), 3 deletions(-)

diff --git a/src/sdtui.c b/src/sdtui.c
index d41058b..f54d5f5 100644
--- a/src/sdtui.c
+++ b/src/sdtui.c
@@ -606,6 +606,9 @@ app_redraw_view (Application *self)
 
 	move (TOP_BAR_CUTOFF, 0);
 
+	gchar *input_utf8 = g_ucs4_to_utf8
+		((gunichar *) self->input->data, -1, NULL, NULL, NULL);
+
 	guint i, k = self->top_offset, shown = 0;
 	for (i = 0; i < self->entries->len; i++)
 	{
@@ -617,9 +620,27 @@ app_redraw_view (Application *self)
 			if (k + 1 == ve->definitions_length)  attrs |= A_UNDERLINE;
 			attrset (attrs);
 
-			guint left_width = app_get_left_column_width (self);
-			app_add_utf8_string (self, ve->word, 0, left_width);
-			addstr (" ");
+			RowBuffer buf;
+			row_buffer_init (&buf, self);
+
+			size_t common = stardict_longest_common_collation_prefix
+				(self->dict, ve->word, input_utf8);
+
+			gchar *prefix = g_strndup (ve->word, common);
+			row_buffer_append (&buf, prefix, A_BOLD);
+			g_free (prefix);
+
+			row_buffer_append (&buf, ve->word + common, 0);
+
+			gint left_width = app_get_left_column_width (self);
+			if (buf.total_width > left_width)
+				row_buffer_ellipsis (&buf, left_width, attrs);
+
+			row_buffer_flush (&buf);
+			for (int i = buf.total_width; i < left_width + 1; i++)
+				addch (' ');
+			row_buffer_free (&buf);
+
 			app_add_utf8_string (self,
 				ve->definitions[k], 0, COLS - left_width - 1);
 
@@ -631,6 +652,8 @@ app_redraw_view (Application *self)
 	}
 
 done:
+	free (input_utf8);
+
 	attrset (0);
 	clrtobot ();
 	refresh ();
diff --git a/src/stardict.c b/src/stardict.c
index 42c7548..9ce6059 100644
--- a/src/stardict.c
+++ b/src/stardict.c
@@ -29,6 +29,7 @@
 
 #include <unicode/ucol.h>
 #include <unicode/ustring.h>
+#include <unicode/ubrk.h>
 
 #include "stardict.h"
 #include "stardict-private.h"
@@ -934,6 +935,73 @@ stardict_dict_search (StardictDict *sd, const gchar *word, gboolean *success)
 	return stardict_iterator_new (sd, imin);
 }
 
+/// Return the longest sequence of bytes from @a s1 that form a common prefix
+/// with @a s2 wrt. collation rules for this dictionary.
+size_t
+stardict_longest_common_collation_prefix (StardictDict *sd,
+	const gchar *s1, const gchar *s2)
+{
+	UErrorCode error;
+	int32_t uc1_len = 0;
+	int32_t uc2_len = 0;
+
+	// It sets the error to overflow each time, even during pre-flight
+	error = U_ZERO_ERROR;
+	u_strFromUTF8 (NULL, 0, &uc1_len, s1, -1, &error);
+	error = U_ZERO_ERROR;
+	u_strFromUTF8 (NULL, 0, &uc2_len, s2, -1, &error);
+	error = U_ZERO_ERROR;
+
+	UChar uc1[uc1_len];
+	UChar uc2[uc2_len];
+	u_strFromUTF8 (uc1, uc1_len, NULL, s1, -1, &error);
+	u_strFromUTF8 (uc2, uc2_len, NULL, s2, -1, &error);
+
+	// Both inputs need to be valid UTF-8 because of all the iteration mess
+	if (U_FAILURE (error))
+		return 0;
+
+	// ucol_getSortKey() can't be used for these purposes, so the only
+	// reasonable thing remaining is iterating by full graphemes.  It doesn't
+	// work entirely correctly (e.g. Czech "ch" should be regarded as a single
+	// unit, and punctuation could be ignored).  It's just good enough.
+	//
+	// In theory we could set the strength to UCOL_PRIMARY and ignore accents
+	// but that's likely not what the user wants most of the time.
+	//
+	// Locale shouldn't matter much with graphemes, let's use the default.
+	UBreakIterator *it1 =
+		ubrk_open (UBRK_CHARACTER, NULL, uc1, uc1_len, &error);
+	UBreakIterator *it2 =
+		ubrk_open (UBRK_CHARACTER, NULL, uc2, uc2_len, &error);
+
+	int32_t longest = 0;
+	int32_t pos1, pos2;
+	while ((pos1 = ubrk_next (it1)) != UBRK_DONE
+		&& (pos2 = ubrk_next (it2)) != UBRK_DONE)
+	{
+		if (!ucol_strcoll (sd->priv->collator, uc1, pos1, uc2, pos2))
+			longest = pos1;
+	}
+	ubrk_close (it1);
+	ubrk_close (it2);
+
+	if (!longest)
+		return 0;
+
+	int32_t common_len = 0;
+	u_strToUTF8 (NULL, 0, &common_len, uc1, longest, &error);
+
+	// Since this heavily depends on UTF-16 <-> UTF-8 not modifying the chars
+	// (surrogate pairs interference?), let's add some paranoia here
+	char common[common_len];
+	error = U_ZERO_ERROR;
+	u_strToUTF8 (common, common_len, NULL, uc1, longest, &error);
+	g_return_val_if_fail (!memcmp (s1, common, common_len), 0);
+
+	return (size_t) common_len;
+}
+
 static void
 stardict_entry_field_free (StardictEntryField *sef)
 {
diff --git a/src/stardict.h b/src/stardict.h
index 7b8322a..25c7939 100644
--- a/src/stardict.h
+++ b/src/stardict.h
@@ -138,6 +138,9 @@ gchar **stardict_dict_get_synonyms (StardictDict *sd, const gchar *word);
 StardictIterator *stardict_dict_search
 	(StardictDict *sd, const gchar *word, gboolean *success);
 
+size_t stardict_longest_common_collation_prefix
+	(StardictDict *sd, const gchar *w1, const gchar *w2);
+
 // --- Dictionary iterators ----------------------------------------------------
 
 struct stardict_iterator