More Phobos testing

bearophile bearophileHUGS at lycos.com
Wed Jan 26 04:56:10 PST 2011


Now and then I try to use the std.algorithm/std.range parts of Phobos2 to see how they are going and developing.

Here I have found Python3 code to compute the Pascal triangle:
http://code.activestate.com/recipes/577542-pascals-triangle/

-------------------

Code adapted for Python2:

def pascal_triangle(n):
    r = [[1]]
    for _ in xrange(n - 1):
        r.append(map(sum, zip([0]+r[-1], r[-1]+[0])))
    return r

print pascal_triangle(5)

-------------------

A similar algorithm translated to D2:

import std.array, std.stdio, std.algorithm, std.range;

auto pascalTriangle(int n) {
    auto r = [[1]];
    foreach (_; 0 .. n-1)
        r ~= array(map!q{a[0] + a[1]}(zip([0]~r[$-1], r[$-1]~[0])));
    return r;
}

void main() {
    writeln(pascalTriangle(5));
}

-------------------

Some notes on the D2 version:

1) In Python I am used to think of higher order functions like map, filter, and zip as similar things. But in Phobos zip is in std.range while map is in std.algorithm. I think this different placement is bad.

-------------------

2) The Python version performs a sum() on tuples. But in D2 tuples, even ones that have an uniform type, aren't iterable, this doesn't work:


import std.stdio, std.typecons;
void main() {
    auto t1 = tuple(1, 2, 3, 4);
    foreach (x; t1)
        writeln(x);
}


Once tuples of uniform type are iterable, and sum() (http://d.puremagic.com/issues/show_bug.cgi?id=4725 ) is implemented, the middle line of code becomes simpler to read:

r ~= array(map!sum(zip([0]~r[$-1], r[$-1]~[0])));


I am aware that this works, but this is static foreach, so this is something different:
foreach (x; t.tupleof)

This is closer to what I'm asking for, but it's not good enough yet:
foreach (x; [t.tupleof])


I have added an enhancement request on this:
http://d.puremagic.com/issues/show_bug.cgi?id=5489

-------------------

3) I like API orthogonality because it allows you to write code like this, that creates a linked list instead of an array out of the lazy map iterable:
linkedList(map!(...)(...))

But in my Python3 and D2 code I write often enough:
array(map!(...)(...))

So I may suggest an amap() the returns an array, it helps remove some parenthesys clutter:
r ~= amap!q{a[0] + a[1]}(zip([0]~r[$-1], r[$-1]~[0]));

r ~= amap!sum(zip([0] ~ r[$-1], r[$-1] ~ [0]));

-------------------

4) In the Python2 version if I print using the pretty print module:

from pprint import pprint
pprint(pascal_triangle(5), width=20)

The triangle gets visualized like this:

[[1],
 [1, 1],
 [1, 2, 1],
 [1, 3, 3, 1],
 [1, 4, 6, 4, 1]]

A function like pprint() will be useful in std.string of Phobos2 too, to give a more readable printing. It's useful to print 2D arrays, arrays of dicts, etc, in a more readable way.

Bye,
bearophile


More information about the Digitalmars-d mailing list