[Issue 3433] [tdpl] Comparing structs for equality is not member-by-member

[d-bugmail at puremagic.com]

http://d.puremagic.com/issues/show_bug.cgi?id=3433



--- Comment #7 from Don <clugdbug at yahoo.com.au> 2009-11-14 12:22:30 PST ---
(In reply to comment #6)
> (In reply to comment #5)
> > This has some tricky points. The first is that it's recursive. As well as
> > structs, it also applies to fixed-length arrays: if either contains a struct
> > with an opEquals, the entire struct must be compared member-by-member; and this
> > check must be performed recursively.
> 
> Correct - more precisely, transitive for direct fields. I think the comparison
> should simply call a template object.structCompare that we define as a library
> function.

I'm pretty sure the compiler can just translate it into (x.field1==y.field1 &&
x.field2==y.field2 && ...) if any field has an opEquals(). This is recursive:
so any struct fields which don't have opEquals() can simply use bitwise
comparison as now. The recursion doesn't apply to classes; if it doesn't have
opEquals(), it's bitwise regardless of what its members do. I guess this is
what you mean by direct fields.

> > The third difficult part relates to protection. Any of the struct members may
> > be private and defined in a different module.
> > 
> > If a field in the struct is a class, it probably applies to it as well.
> 
> I'm not sure how to address this.


I have now got this successfully detecting the cases where it needs to use
member-by-member comparison for the cases above. I'll add an error for the
union case. I don't yet have a solution to the protection issue. Obviously it
can done by suppressing any protection errors which occur, but I hope it can be
done without a compiler hack.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------