Mata constructor with arguments

[Niels Henrik Bruun]

Hi
When building classes in Mata it seems like one needs to instantiate the class before setting it's properties (see example coord below which works).
There is a constructor (new) but as the documentation says it can't take any arguments.

I would like to do it like in class coord2 below where instantiation and setting of properties is done in 1 line.

Is this possible?

cls
mata
mata clear
mata describe


class coord {
    real scalar    x, y
    void setup()
}

    void coord::setup(real scalar x, y)
    {
        this.x = x
        this.y = y
    }

    
xy = coord()
xy.setup(1, 2)

xy.x
xy.y


class coord2 {
    real scalar x, y
    void new()
}

    void coord2::new(real scalar x, y)
    {
        this.x = x
        this.y = y
    }

    
xy2 = coord2(1, 2)

xy2.x
xy2.y
end

[wbuchanan]

Niels Henrik Bruun the same issue persists in version 14. It would definitely make life easier if it were possible to pass arguments to constructors.

[Matthew J. Baker]

Niels Henrik Bruun and wbuchanan :

I agree that this would be nice. While it takes a little bit more work, one can proceed as Niels Henrik Bruun initially suggests, and then make a wrapper that does everything in one step. This is my coordmaker() function below.

mata
class coord {
    real scalar    x, y
    void setup()
}

    void coord::setup(real scalar x, y)
    {
        this.x = x
        this.y = y
    }

class coord coordmaker(x,y)
{
    class coord scalar xy
    xy=coord()
    xy.setup(x,y)
    return(xy)

}    
xy = coord()
xy.setup(1, 2)

xy.x
xy.y

/* One step */
yz=coordmaker(2,3)
yz.x
yz.y
end

Hope that helps!

Matt

[wbuchanan]

Matthew J. Baker I ended up having to do something similar. It's just a bit of a pain in terms of introducing additional points of failure. In my case in particular, it's a bit more of a pain since I'm trying to wrap an existing library in a different language that keeps the native API as intact as possible (and then will build out Stata interfaces to the Mata classes/methods afterwards).

[Niels Henrik Bruun]

Matthew J. Baker : Thank you for your help. I ended up deciding not to use the function approach.

[wbuchanan]

Matthew J. Baker have you had any success accessing/chaining methods/functions (http://www.statalist.org/forums/foru...ion-chaining)? I have a larger class defined with methods that generate other classes, which also have their methods. Structs didn't seem like a plausible option since there are several occasions where same named methods exist for different classes and use different arguments/logic. I'm kind of stumped at this point since it wouldn't be that difficult to do something similar in Java and/or other languages, but was trying to keep things in Mata since there seems to be less control when things come of the stack with the Java API.

[Matthew J. Baker]

wbuchanan --

Sorry I just saw your last message today as I have been unavoidably detained by other matters. I couldn't get the link you posted to work is it possible to repost it?

Best,

Matt

[wbuchanan]

Hi Matthew J. Baker,

The link was to a related post about method/function chaining in mata:

http://www.statalist.org/forums/foru...ction-chaining