Suggestion : Replace case/switch-based tests (for the type of a token) by
Gee maps-based (or multimaps-based) tests in Vala scanners e.g.
(valaparser.vala).
Goal:
- Render the lookup O(1) instead of O(N).
- Make vala-based scanners, parsers, tokenizers faster.
Example:
HashSet<char> stopCharSet = new HashSet<char>();
string stopChars = " \r\n\t.;,:?!+=|()*[]\\|/<>&^%$#@`~\"'";
public scanner (string fname) {
f = FileStream.open(fname, "r");
if (f==null) {
Posix.stdout.printf("File missing !\n");
exit(1);
}
for(var i= 0; i<stopChars.length; i++) {
stopCharSet.add(stopChars[i]);
}
}
...
//usage:
//if (isspace(c)) {
if (c in stopCharSet) {
/* The following test is O(1), not O(N), it does not slow down
linearly with respect to the size of "stopChars" */
if (wordBuffer.len>0) {
//Posix.stdout.printf("word = %s\n", wordBuffer.str);
wordList.add(wordBuffer.str);
if (erased == false) {
wordBuffer.erase(0,-1);
wordBuffer_length = 0;
erased = true;
}
}
}
....
Serge.
_______________________________________________
vala-list mailing list
[email protected]
http://mail.gnome.org/mailman/listinfo/vala-list
