bug-gawk
[Top][All Lists]
Advanced
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] Feature: parameters declared in body

From: Ed Morton
Subject: Re: [PATCH] Feature: parameters declared in body
Date: Fri, 8 Apr 2022 08:19:27 -0500
User-agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:91.0) Gecko/20100101 Thunderbird/91.7.0 
On 4/8/2022 7:41 AM, david kerns wrote:

What's with the @?? (we already chose awk over Pearl because Pearl reads
like a swear word @#$%&...)
Why not just make a new keyword "local" that limits a variable's scope to
the block {} it's in?
Because that would potentially break existing awk scripts that could be using the word "local" as a variable or function name. Don't get me wrong, I'm definitely not advocating for introducing "@local" or "@param" into the language, just stating the obvious reason to avoid introducing a keyword like "local" alone unless there was some massive advantage to doing so that outweighed the potential breakage of existing scripts. 

    Ed.



thus:
$ cat l.awk
function f(p1)
{
   local i, j = 7;
   for (i = 1; i < p1; i++) {
     j += p1;
   }
   return j;
}
BEGIN   {
         for (local i = 3; i < 4; i++) {
           q += f(i);
         }
         print q;
}
         { local i = $1; }
         { print "i = ", i }
$ echo 42 | awk -f l1.awk
13
i =

(which I faked with)
$ cat l1.awk
function f(p1,     i,j)
{
   j = 7;
   for (i = 1; i < p1; i++) {
     j += p1;
   }
   return j;
}
BEGIN   {
         for (i = 3; i < 4; i++) {
           q += f(i);
         }
         print q;
}
         { i = $1; }
         { print "i = ", j }

  which, frankly, doesn't read that bad, with the exception of the last line



On Fri, Apr 8, 2022 at 4:38 AM Ed Morton<mortoneccc@comcast.net>  wrote:


On 4/8/2022 3:00 AM,arnold@skeeve.com  wrote:

Thanks for this submission.  I will think about it for a bit.

Arnold

Kaz Kylheku<kaz@kylheku.com>   wrote:


Hi,

I'm posting this as a proposal. If this is accepted, in some
shape, then of course there will be test cases and
documentation.

The idea is that there is a new kind of attribute syntax
whereby, inside a function, if we put the @param: prefix
before a variable, that variable is added as a parameter
to the surrounding function, which has the effect of making
it a local variable.

This effectively extends the Awk language with a nice way
of defining locals, that is easy to use from code generators
and macro preprocessors, and is based in the existing model
of Awk local variables being parameters.

A bit of a cautionary word: this relies on knowing the order
in which things are parsed.  If a language construct mentions
a certain variable several times, and one of those mentions
has a @parm: prefix, that mention had better be processed by
the parser before the others, otherwise those others will be
compiled in a scope in which the variable was not defined
as a parameter yet.

For instance if we write:

     for (@param:i = 0; i < 10; i++)

we are relying on the parser processing @param:i
before i < 10 or i+i. We can infer that from knowing
the phrase structure rule, and how it is processed by the
LALR(1) algorithm.


Cheers ...

P.S. Small demo showing how the variables are local:

$ ./gawk -d '
function add(n, m) {
     for (@param:i = n; i <= m; i++)
        @param:sum += i;
     return sum
}

BEGIN { print add(2, 5); }'
14
$ grep '^[a-z]' awkvars.out
$

Now remove @param:

$ ./gawk -d '
function add(n, m) {
     for (i = n; i <= m; i++)
       sum += i;
     return sum
}

BEGIN { print add(2, 5); }'
14
$ grep '^[a-z]' awkvars.out
i: 6
sum: 14


Argument passing works with the @param variables:

$ ./gawk -d '
function add() {
     return @param:a + @param:b
}

BEGIN { print add(2, 5); }
'
7

I almost feel this just wants to be @:a + @:b.




If there's going to be a construct to declare local variables as in:

     function add(n, m) {
         for (@param:i = n; i <= m; i++)
            @param:sum += i;
         return sum
     }

then obviously it should be named "@local" instead of "@param" but IMHO
it's not worth adding an extra keyword just for that given the mantra
you've mentioned before that awk avoids language bloat by only having
constructs to do things that are hard to do with other constructs.

Also, please don't introduce a construct that lets you declare a
variable as a parameter without it appearing in the function declaration
as in:

     function add() {
         return @param:a + @param:b
     }

as that makes the code noticeably harder to read - I want to see the
function parameters and the order they have to be provided where the
function is declared, not have to look through the body of the function
for "@param" statements and have to think about the order those
statements will be parsed to figure out what arguments the function
requires in what order.

      Ed.
reply via email to
[Prev in Thread] Current Thread [Next in Thread]