Comment #0 by bearophile_hugs — 2012-06-11T13:33:22Z
import std.stdio: writefln;
import std.math: sin, asin, cos, acos, cbrt, pow;
import std.functional: compose;
import std.typetuple: TypeTuple;
void main() {
const sin = (in real x) => sin(x);
const asin = (in real x) => asin(x);
const cos = (in real x) => cos(x);
const acos = (in real x) => acos(x);
const cube = (in real x) => x ^^ 3;
const cbrt = (in real x) => cbrt(x);
alias TypeTuple!(sin, cos, cube) dir;
alias TypeTuple!(asin, acos, cbrt) inv;
foreach (i, f; dir)
writefln("%6.3f", compose!(f, inv[i])(0.5));
}
Output, dmd 2.060alpha:
0.500
0.866
0.713
Expected output similar to:
0.500
0.500
0.500
Including the source code of std.functional.compose and simplifying it, it shows the problem still:
import std.functional: unaryFun;
template compose2(fun...) {
alias unaryFun!(fun[0]) fun0;
alias unaryFun!(fun[1]) fun1;
auto compose2(T)(T a) {
return fun0(fun1(a));
}
}
void main() {
import std.stdio, std.math, std.typetuple;
const sin = (in real x) => sin(x);
const asin = (in real x) => asin(x);
const cos = (in real x) => cos(x);
const acos = (in real x) => acos(x);
const cube = (in real x) => x ^^ 3;
const cbrt = (in real x) => cbrt(x);
alias TypeTuple!(sin, cos, cube) dir;
alias TypeTuple!(asin, acos, cbrt) inv;
foreach (i, f; dir)
writefln("%6.3f", compose2!(f, inv[i])(0.5));
}
Output, dmd 2.060alpha:
0.500
0.866
0.713
Removing the unaryFun!() the results become correct:
import std.functional: unaryFun;
template compose2(fun...) {
alias fun[0] fun0;
alias fun[1] fun1;
auto compose2(T)(T a) {
return fun0(fun1(a));
}
}
void main() {
import std.stdio, std.math, std.typetuple;
const sin = (in real x) => sin(x);
const asin = (in real x) => asin(x);
const cos = (in real x) => cos(x);
const acos = (in real x) => acos(x);
const cube = (in real x) => x ^^ 3;
const cbrt = (in real x) => cbrt(x);
alias TypeTuple!(sin, cos, cube) dir;
alias TypeTuple!(asin, acos, cbrt) inv;
foreach (i, f; dir)
writefln("%6.3f", compose2!(f, inv[i])(0.5));
}
Output, dmd 2.060alpha:
0.500
0.500
0.500
Comment #1 by peter.alexander.au — 2014-02-07T14:43:08Z
Not a problem with unaryFun. I have opened Issue 12103, which reduces this bug.
Comment #2 by bearophile_hugs — 2014-02-07T15:28:32Z
(In reply to comment #1)
> Not a problem with unaryFun. I have opened Issue 12103, which reduces this bug.
Once Issue 12103 what's left to do for this Issue 8224? Or in other words, why did you open a new issue? :-)
Comment #3 by bearophile_hugs — 2014-02-07T15:29:27Z
(In reply to comment #2)
> Once Issue 12103 what's left to do for this Issue 8224?
I meant to write:
Once Issue 12103 is fixed, what's left to do for this Issue 8224?
Comment #4 by peter.alexander.au — 2014-02-07T15:46:39Z
(In reply to comment #2)
> (In reply to comment #1)
> > Not a problem with unaryFun. I have opened Issue 12103, which reduces this bug.
>
> Once Issue 12103 what's left to do for this Issue 8224? Or in other words, why
> did you open a new issue? :-)
I don't like to drastically change bugs, and the title in this one is misleading for people looking for bugs to solve. Perhaps I should have just changed this one...
Comment #5 by bearophile_hugs — 2014-02-07T16:02:19Z
(In reply to comment #4)
> I don't like to drastically change bugs, and the title in this one is
> misleading for people looking for bugs to solve. Perhaps I should have just
> changed this one...
Yes, in such cases the right thing to do is to rename the issue, instead of opening another one :-) So do you want me to close this issue down?
Comment #6 by bugzilla — 2019-12-08T12:43:29Z
Looks like a problem in the TypeTuple foreach-loop. f is always sin.
Comment #7 by simen.kjaras — 2019-12-09T12:16:38Z
The problem is reducible to this:
unittest {
import std.stdio, std.typetuple;
const fun1 = () => 0;
const fun2 = () => 1;
foreach (i, f; AliasSeq!(fun1, fun2)) {
alias f2 = Alias!f;
assert(i == f(), "foreach alias works great");
assert(i == f2(), "template is incorrectly cached");
}
}
As we can see, the issue occurs when f is passed to a template. It seems DMD caches the template instantiation and doesn't update it for each iteration.
Comment #8 by bugzilla — 2019-12-09T12:21:03Z
(In reply to Simen Kjaeraas from comment #7)
> It seems DMD caches the template instantiation and doesn't update it for each iteration.
Hence, this is a DMD bug and not Phobos?
Comment #9 by simen.kjaras — 2019-12-09T13:44:57Z
Definitely. I didn't even consider the possibility that this could have been filed as a Phobos bug when I commented. :p
Comment #10 by robert.schadek — 2024-12-13T18:00:27Z