On Thursday, 19 April 2012 at 12:55:50 UTC, Roman D. Boiko wrote:
On Thursday, 19 April 2012 at 12:31:24 UTC, Tobias Pankrath
wrote:
T foo(T)(T x) {
return x.something();
}
int something(int x) {
return 1;
}
float something(float x) {
return 1.0;
}
Now... go and rename the first function named "something".
What do you do with "x.something()" inside the template...
rename it or not?
I would go ReSharper way: present user a tree with all
occurrences of related symbols and checkboxes to include or
exclude each. Or, in this particular case, display a list of
affected types and checkboxes for each. This isn't VERY user
friendly, though, but very flexible and reasonable. Give
warnings if necessary.
Semantic analysis phase would provide information whether
symbol is actually related, or only has the same name (but
different semantics and scope).
Respective convenience API would simply provide overloads and
reasonable defaults for similar cases.
My point is that there is nothing in language itself preventing
such functionality, if we have semantic analysis. The problem is
related to user intentions, and decision should be made by the
user (but the tool should provide good defaults and enough
information).