Bug 11403 – functions in std.algo can't be used as pred
Status
RESOLVED
Resolution
FIXED
Severity
normal
Priority
P2
Component
phobos
Product
D
Version
D2
Platform
All
OS
All
Creation time
2013-10-31T12:55:00Z
Last change time
2014-01-06T10:30:34Z
Keywords
rejects-valid
Assigned to
nobody
Creator
monarchdodra
Comments
Comment #0 by monarchdodra — 2013-10-31T12:55:31Z
There are a ton of functions in std.algorithm that can be customized with a predicate. At ton of them can also be used as predicates themselves. Here is a cool example:
assert(equal!equal([[1, 2], [3, 4]], iota(1, 5).chunks(2)));
The problem with this piece of code is that:
The inner "equal" only works because it itself doesn't have a predicate.
The outer "equal" works because it takes arguments.
However, neither of this works:
alias fun = equal!equal; //Nope
equal!(equal!equal))(3Dmatrix1, 3Dmatrix2); //Nope
It is *not* possible to declare an algorithm function with a predicate, unless you also specify the argument types. Long story short, they are not eagerly specialize-able.
"equal" (I think) is the most obvious offender, but so are any/all:
import std.ascii : isWhite;
assert( all!(any!isWhite)(["a a", "b b"])); //Nope
assert(!any!(all!isWhite)(["a a", "b b"])); //Nope
Any function that takes a pred (+an argument) and returns a bool is faulty. Here is a (probably incomplete) list:
-equal
-canFind
-any
-all
-cmp
I don't know if we can *really* consider this "resolved", but the most important functions are now pred-compatible.
For any other improvements, I think it might be worth looking into a more generic solution, maybe. I'll just close it, and we'll deal with any other issue on a case-by-case basis.