Issue 10265 - RandomSample fails when passed an InputRange as input

[WebDrake]

http://d.puremagic.com/issues/show_bug.cgi?id=10265
Issue fixed by making presence of save() property conditional on input being a ForwardRange.

[WebDrake]

This probably needs an extra unittest for the case where randomSample is passed an InputRange. Any suggestions for something to use? file.byLine springs to mind, but would require specifying a file to open.

[jmdavis]

@WebDrake It's simple enough to create one that I'd just create a simple one for the test. Something like

struct InputRange
{
    @property bool empty() const { return _arr.empty; }
    @property int front() const { return _arr.front; }
    void popFront() { _arr.popFront(); }
    int[] _arr;
}

[WebDrake]

It's simple enough to create one that I'd just create a simple one for the test.

Seems your example may have uncovered another issue with RandomSample. If I use it, I get the following error message:

std/random.d(1649): Error: constructor std.random.RandomSample!(TestInputRange, void).RandomSample.this
field _input must be initialized in constructor, because it is nested struct

This is because all the initialization of variables is done in a separate initialize() function which is called by the constructors.

Is there any other way of fixing this other than moving the _input = input line to the constructors? The whole point of the initialize() function was to avoid that kind of duplication :-P

[WebDrake]

Without a solution for the nested struct issue mentioned in the previous post, I've moved the assignment of _input to constructors for now -- to be moved back if anyone can suggest a superior alternative.

[WebDrake]

Ping ... ? I think this is in an acceptable state now.