Very cool, if true

Very true! I do like things to be nicely scoped.

From:

Can you at least do the C'ish

var blah;
if ((var = getBlah()) != null)
{
  //Do stuff with blah
}

From:

Hm, some sort of macro, hashdefine or metaprogramming sytem is what's needed to do this right.

The closest you could get in C# would be an extension method using some generics and lambdas:


public static T DoIfNotNull<T>(this T value, Action<T> action) where T: class
{
    if (value != null)
    {
        action(value);
    }

    // return the value for further chaining
    return value;
}

Then you could code:

GetBlah().DoIfNotNull(blah => DoStuffWith(blah));

of course there are variations on this like


public static U DoIfNotNull<T, U>(this T value, Func<T, U> func) where T : class
{
    U result = default(U);

    if (value != null)
    {
        result = func(value);
    }

    return result;
}

.. so then your action can change the type and you can still chain methods.

Edited Date: 2011-06-20 12:56 pm (UTC)

From:

Yup - that first one would be good. A bit unwieldy if the code being called is more than a couple of lines, but still a good step in the right direction.

From:

Aha! I can! I hadn't tried that before, and it works perfectly.

From:

In general yes, since (blah = getBlah()) returns the value assigned to blah.
You'd have to declare blah without using "var" though, unless there's an assignment on the same line the type inference has nothing to work on.

so this works


string blah;
if ((blah = getBlah()) != null)
{
  //Do stuff with blah
}

private string getBlah()
{ ... }

From:

If it's more than a few lines, you can put it in it's own method. I don't like long lambdas. A long lambda for me is > 2 statements.


// code is actually shorter - method is auto cast to action        
getBlah().DoIfNotNull(SomeMethod);

... 

private void SomeMethod(string blah)
{
 ...
}

From:

steer.livejournal.com

Related but not quite what you're after is Java 7's "null-safe" handling:

http://metoojava.wordpress.com/2010/11/15/java-7-awesome-features/

It's an interesting language feature I've not seen elsewhere that I hope spreads.

From:

You'd have to declare blah without using "var" though

Ah yes, I'd forgotten that C# has type inference; I'd assumed that Andy's 'var' was pseudocode, heh.

From:

it's very limited type inference. C# just looks for var foo = SomeExpression; and replaces "var" with the type of SomeExpression

The stuff that OCaml (and F#) do with type inference is just amazing.

From:

That _is_ nice. I'd love to see that in C#.

From:

What does F# do? I've not played with it at all.

From:

It does type inference across the whole method, so in let f a b = a + b + 100, the type a, b and the return type of f are all infered as int.

It does the equivalent of generics automatically, so let makeTuple a b = (a, b) is the same as in C# Tuple<A, B> MakeTupple<A, B>(A a, B b) { ... }

Edited Date: 2011-06-20 02:13 pm (UTC)

From:

skington.livejournal.com

Perl lets you do that, but not quite as much as you'd want.

You can say

if (my $blah = get_blah()) {
...
}

but not

if (my $blah = get_blah() and $blah->can('do_stuff')) {
$blah->do_stuff();
}

because $blah doesn't exist when the second part of the conditional is run. You need to say

if (my $blah = get_blah()) {
if ($blah->can('do_stuff')) {
$blah->do_stuff();
}
}

instead.

From:

Also, Haskell. <3

From:

I have a vague recollection that some language I used to know had a similar feature. It might be Verisity 'E', but I don't remember it having exactly this. It did have some very neat functional list operators anyway, and I think it might have been possible to easily construct an idiom that did that, but i can't remember enough of the details to work out if it's possible.

It also strongly reminds me of Icon's goal-directed evaluation, but since Icon doesn't have any OO features and very very limited structures, I'm pretty sure it doesn't have anything quite like this; but if it *did*, it would support this sort of thing straight out of the box.

(I recommend giving Icon a once-over for curiosity's sake, by the way; it's an interesting little language, primarily because of the goal-oriented evaluation and generation. )

From:

cartesiandaemon.livejournal.com

Hm. I was thinking that C and C++ can do that, but didn't follow the thought through to the conclusion. I can't remmeber if C or C++ allow variable declaration in any statements other than for loops?

However, I don't think I _want_ to write what Andy wanted to write. I think the assignment and test together are two ugly. I would accept a syntax like:

if (var blah=getBlah()) // if implicit test for false/null makes sense

or:

if (var blah=getBlah(); blah!=null)

or possibly some other convenient shorthand.

In C++ it might be possible to make an awful preprocessor hack, something like

#define TRY(var, value) for(auto var = value;var!=null;var=null)

TRY(blah, GetBlah()) {
 // do stuff
}

You mgiht be able to improve the semantics, I'm not sure. I doubt it's worth using, though. You may be able to do something fancier with templates, but it's getting it to act like an "if" statement that's hard.

From:

momentsmusicaux.livejournal.com

But the second version is so ugly! You can't as easily see what's going.

From:

I think what I want is the variable to be declared within the scope of the if{}.

From:

This is, of course true.

There ought to be a good way of saying "Is there an X? If so, do stuff with it?", and I don't think I've found that just yet.

From:

momentsmusicaux.livejournal.com

I like this in perl:

$saveme = $boring or $interesting;

$boring gets skipped if it gets evaluated as false in some way or other.

Apparently PHP now has this as:

$saveme = $boring ?: $interesting;

which... urgh.

From:

Hmm. I think that "or" already has a fairly understood meaning, and using it to coalesce nulls would be confusing to people.

From:

Yes, in C++ you can say:

if (blahType blah = GetBlah()) { ... }

From: