Hi,
I've got an issue I cannot seem to find an appropriate solution for.
I've build a scanner and a parser for parsing some kind of custom
language.
Though, I can have a kind of expression like:
1 + 2.1 = 3
I am parsing expressions as terms of single statements inheriting from
a lex_unit class:
class lex_unit
{
enum lex_type_e { t_integer, t_double, t_plus, t_equal }
virtual int getIntValue() = 0;
virtual double getDoubleValue() = 0;
lex_unit* prev, next;
};
class lex_unit_int : public lex_unit
{
int getIntValue() { ...}
}
class lex_unit_double ..
... yadayada
In the above sample, my classes after parsing would look like this:
lex_unit_int (1), lex_unit(t_plus), lex_unit_double(2.1),
lex_unit(t_equal), lex_unit_int(3)
Sorry for the mass of pseudo code, I'd just like to make things
clear :)
Now, my concern is simply the overhead of classes, constructors and
destructors. This stuff gets called so often that it's a real
bottleneck (yes, I have checked it!). My usual approach is to
implement some pooling but in my case that'd only work with POD types,
not with classes inheriting each other.
So my question is basically, is there a better way to do it and/or can
I do some pooling or anything else?
Btw -- I need some kind of connected lex_units so that I can jump
through each one with prev/next and check them..
thanks!
Alex
