• bearophile (25/25) Mar 24 2011 As Phobos develops more, there are some commonly used Phobos functions t...

bearophile <bearophileHUGS lycos.com> writes:

As Phobos develops more, there are some commonly used Phobos functions that
aren't pure yet:


import std.algorithm, std.typecons, std.conv, std.array, std.range;
pure auto foo(int[] a) {
    sort(a);
    auto a2 = array(iota(a.length)); 
    return tuple(text(a), a2);
}
void main() {}


But a function like text() may call toString(), that's not always pure, so D
needs the conditionally purity to implement a conditionally pure text() (or
text code needs to be duplicated with a pure version, but probably this is not
an option).


Random number generation too is useful in pure functions, see:
http://d.puremagic.com/issues/show_bug.cgi?id=5249

// strongly pure
pure double[] foo(const int n, const double firstSeed=42) {
    double seed = firstSeed;
    auto res = new double[n];
    foreach (ref r; res) {
        auto seed_rnd = nextRandom(seed, 1.0);
        r = seed_rnd.rnd;
        seed = seed_rnd.seed;
    }
    return res;
}


When such parts of Phobos will become (conditionally) pure, it will become much
simpler to write D small programs as a small impure I/O part that calls mostly
pure functions/methods :-) Simon Peyton-Jones says this is one of the most
important qualities of Haskell, and from what I'm seeing, it's not a bad thing.

Bye,
bearophile