Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: [Mata] passing a function to mata (new question)


From   "Jann Ben" <ben.jann@soz.gess.ethz.ch>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: [Mata] passing a function to mata (new question)
Date   Fri, 26 Aug 2005 23:22:48 +0200

Many thanks for the clarifications, Bill. Using pointers is exactly what
I did in my program. I will be looking forward to the revised
findexternal() function. 

I will shut up now and stop asking complicated Mata questions. Why is it
that I always stumble over the 0.01% of problems.

ben


> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of 
> William Gould, Stata
> Sent: Friday, August 26, 2005 3:48 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: [Mata] passing a function to mata (new question)
> 
> 
> Do you want to read this?
> -------------------------
> 
> Mata is new and many of you have not tried it yet.  
> 
> I fear that if you read this question and answer, you will 
> never use Mata.
> You will come to the conclusion that Mata is suitable only 
> for advanced 
> programming problems and that using Mata requires knowledge 
> that you do 
> not have, nor much care to gain.
> 
> That would be unfortunate, because Mata really is easy to use 
> and is worth
> learning.  99.99% of all problems that will ever be 
> approached using Mata will
> not involve the topics discussed below.  Do not hold it 
> against Mata that it
> can deal with the 0.01% of problems, too.
> 
> All that said, what follows really is not difficult, it is 
> merely unlikely 
> that the majority of Statalisters would ever face the problem.
> 
> 
> Question and response
> ---------------------
> 
> The subject is passing funtions to functions in Mata, and Ben Jann
> <ben.jann@soz.gess.ethz.ch> imagines a situation where the 
> function he wants
> to pass is given by a string variable:
> 
>         function me1(string scalar he)
>         {
>                       ...
>                 if (he=="he1")      y = me2(x, &he1())
>                 else if (he=="he2") y = me2(x, &he2())
>                 else if (he=="he3") y = me2(x, &he3())
>                 ...
>                 return(...)        
>         }
> 
> Benn wonders if it is possible to code the above in such a 
> way that the value
> of string scalar he is used directly.  As Benn says,
> 
> > If Mata had macro expansion, I would type
> >
> >       function me1(string scalar he)
> >       {
> >              ...
> >               y = me2(x, &`he'())
> >               ...
> >               return(...)        
> >       }
> 
> 
> Comment 1, "If Mata had macro expansion..."
> -------------------------------------------
> 
> My recommendation is to stop thinking like that, although I know that 
> after years of ado programming, that is difficult.
> 
> What makes Mata faster than ado?  It is the fact that Mata 
> programs are
> compiled, whereas ado programs are interpreted.  When you 
> think in terms 
> of macro expansion, you are thinking interpretive.
> 
> Mata in fact does have macro expansion but you have to 
> distinguish between
> expansion at compile time and expansion at run time.  The 
> little bit of code
> Benn produced would expand `he' at compile-time, and that is 
> *NOT* what Benn
> wants.
> 
> Expansion at run-time requires writing lots of code.  Don't go there.
> 
> 
> Comment 2:  Can the string scalar be used directly?
> ---------------------------------------------------
> 
> No, not right now.  I have more to say on this, but let's go 
> with the "No"
> answer right now.
> 
> The most efficient thing to do is just what Benn said, 
> 
>         function me1(string scalar he)
>         {
>                       ...
>                 if (he=="he1")      y = me2(x, &he1())
>                 else if (he=="he2") y = me2(x, &he2())
>                 else if (he=="he3") y = me2(x, &he3())
>                 ...
>                 return(...)        
>         }
> 
> Now imagine you had 15 places in your code where you wanted 
> to execute 
> function `he'() (excuse the macro notation).  You could do this:
> 
> 
>         function me1(string scalar he)
>         {
>                 ...
>                 if (he=="he1")      p = &he1()
>                 else if (he=="he2") p = &he2()
>                 else if (he=="he3") p = &he3()
>                       ...
>                       ...
>                 y = me2(x, p)                (1st call)
>                 ...
>                 y = me2(x, p)                (2nd call)
>                 ...
>                 ...
>                 y = me2(x, p)                (15th call)
> }
> 
> In the above, p is a pointer.  If you felt like delcaring it, you
> could delcare it
> 
>                 pointer(function) p 
> 
> The point of the above is that something like &he1() is just 
> an address,
> and pointers hold addresses, and so you can use pointer variables to 
> hold addresses of functions just as you can use them to hold addresses
> of matrices, vectors, and scalars.
> 
> 
> Comment 3:  Can the string scalar be used directly?
> ---------------------------------------------------
> 
> "No," I just said.
> 
> Have you seen the function [M-5] findexternal()?  Quoting from the
> documentation,
> 
> 
>     findexternal(name) returns a pointer to the external global 
>     matrix, vector, or scalr whose name is specified by name; 
>     it returns NULL if the external global is not found.
> 
> 
> Thus, findexternal("x") returns the address of global x, if it exists.
> 
> findexternal("he1()") ought to return the address of function he1().
> It does not, but it ought to, and I feel so strongly about 
> that, I have 
> just recorded it as a bug that it does not.
> 
> So let's imagine the "bug" is fixed.  Then Benn could code, 
> 
> 
>         function me1(string scalar he)
>         {
>                       ...
>                 p = findexternal(he+"()")
>                 ...
>                 y = me2(x, p)
>                 ...
>         }
> 
> Presumably, Benn ought to worry about p being NULL.
> 
> -- Bill
> wgould@stata.com
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
> 

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index