Bug797501 - Currency symbols in Hebrew (RTL) language

When Gnucash is run in Hebrew which is a RTL language, on the accounts
page the tree view is displaying the required number as the following...

TreeView entry is   '1,500.00 ₪' or '-1,500.00 ₪'
TreeModel string is '₪ 1,500.00‬' or '₪ 1,500.00-‬'

This seems to be down to the GTK 'Unicode Bidirectional Text Algorithm'
which is changing the representation of the model string based on the
first strongly typed character, in this case the Israeli shekel sign.

To fix this, when creating the displayed monetary amount insert a BiDi
ltr isolate uni-character at the start of the string.

libgnucash/app-utils/gnc-ui-util.c

@@ -1850,11 +1850,13 @@ xaccSPrintAmount (char * bufp, gnc_numeric val, GNCPrintAmountInfo info)
     return (bufp - orig_bufp);
 }
 
+#define BUFLEN 1024
+
 const char *
 xaccPrintAmount (gnc_numeric val, GNCPrintAmountInfo info)
 {
     /* hack alert -- this is not thread safe ... */
-    static char buf[1024];
+    static char buf[BUFLEN];
 
     if (!xaccSPrintAmount (buf, val, info))
         buf[0] = '\0';
@@ -1863,6 +1865,55 @@ xaccPrintAmount (gnc_numeric val, GNCPrintAmountInfo info)
     return buf;
 }
 
+const char *
+gnc_print_amount_with_bidi_ltr_isolate (gnc_numeric val, GNCPrintAmountInfo info)
+{
+    /* hack alert -- this is not thread safe ... */
+    static char buf[BUFLEN];
+    static const char ltr_isolate[] = { 0xe2, 0x81, 0xa6 };
+    static const char ltr_pop_isolate[] = { 0xe2, 0x81, 0xa9 };
+    size_t offset = info.use_symbol ? 3 : 0;
+
+    memset (buf, 0, BUFLEN);
+    if (!xaccSPrintAmount (buf + offset, val, info))
+    {
+        buf[0] = '\0';
+        return buf;
+    };
+
+    if (!info.use_symbol)
+        return buf;
+
+    memcpy (buf, ltr_isolate, 3);
+
+    if (buf[BUFLEN - 4] == '\0')
+    {
+        size_t length = strlen (buf);
+        memcpy (buf + length, ltr_pop_isolate, 3);
+    }
+    else
+    {
+        buf[BUFLEN - 1] = '\0';
+        memcpy (buf + BUFLEN - 4, ltr_pop_isolate, 3);
+
+        PWARN("buffer length %d exceeded, string truncated was %s", BUFLEN, buf);
+    }
+    /* its OK to return buf, since we declared it static
+       and is immediately g_strdup'd */
+    return buf;
+}
+
+gchar *
+gnc_wrap_text_with_bidi_ltr_isolate (const gchar *text)
+{
+    static const char *ltr = "\u2066"; // ltr isolate
+    static const char *pop = "\u2069"; // pop directional formatting
+
+    if (!text)
+        return NULL;
+
+    return g_strconcat (ltr, text, pop, NULL);
+}
 
 /********************************************************************\
  ********************************************************************/