#1535 Optimize DBCS `MovePositionOutsideChar()` and `NextPosition()`

[Zufu Liu]

Call LineStartPosition(pos) is expensive inside both function.

For MovePositionOutsideChar() function, posStartLine can be removed: when pos is at line start, its previous character can only be CR, LF or NUL, which is not lead byte, so following code is empty loop.

[Zufu Liu]

Patch for MovePositionOutsideChar():

@@ -833,15 +833,9 @@
                // Else invalid UTF-8 so return position of isolated trail byte
            }
        } else {
-           // Anchor DBCS calculations at start of line because start of line can
-           // not be a DBCS trail byte.
-           const Sci::Position posStartLine = LineStartPosition(pos);
-           if (pos == posStartLine)
-               return pos;
-
            // Step back until a non-lead-byte is found.
            Sci::Position posCheck = pos;
-           while ((posCheck > posStartLine) && IsDBCSLeadByteNoExcept(cb.CharAt(posCheck-1)))
+           while ((posCheck > 0) && IsDBCSLeadByteNoExcept(cb.CharAt(posCheck-1)))
                posCheck--;

            // Check from known start of character.

Not yet figure out how to eliminate posStartLine in NextPosition().

[Zufu Liu]

Patch also removes posStartLine for NextPosition():

@@ -916,14 +910,11 @@
                if (pos > cb.Length())
                    pos = cb.Length();
            } else {
-               // Anchor DBCS calculations at start of line because start of line can
-               // not be a DBCS trail byte.
-               const Sci::Position posStartLine = LineStartPosition(pos);
-               // See http://msdn.microsoft.com/en-us/library/cc194792%28v=MSDN.10%29.aspx
-               // http://msdn.microsoft.com/en-us/library/cc194790.aspx
-               if ((pos - 1) <= posStartLine) {
-                   return pos - 1;
-               } else if (IsDBCSLeadByteNoExcept(cb.CharAt(pos - 1))) {
+               // How to Go Backward in a DBCS String
+               // https://msdn.microsoft.com/en-us/library/cc194792.aspx
+               // DBCS-Enabled Programs vs. Non-DBCS-Enabled Programs
+               // https://msdn.microsoft.com/en-us/library/cc194790.aspx
+               if (IsDBCSLeadByteNoExcept(cb.CharAt(pos - 1))) {
                    // Should actually be trail byte
                    if (IsDBCSDualByteAt(pos - 2)) {
                        return pos - 2;
@@ -934,7 +925,7 @@
                } else {
                    // Otherwise, step back until a non-lead-byte is found.
                    Sci::Position posTemp = pos - 1;
-                   while (posStartLine <= --posTemp && IsDBCSLeadByteNoExcept(cb.CharAt(posTemp)))
+                   while (--posTemp >= 0 && IsDBCSLeadByteNoExcept(cb.CharAt(posTemp)))
                        ;
                    // Now posTemp+1 must point to the beginning of a character,
                    // so figure out whether we went back an even or an odd

even with this change, DBCS backward brace match using NextPosition() is still much slower than using MovePositionOutsideChar().

[Neil Hodgson]

Committed with [24545b] and [2c0dbb].