nms-cgi-devel

[Nms-cgi-devel] Open question on variable/constants

[Olivier D. <dr...@sh...>]

Hi,

I would just like to submit a question to everyone to either answer or
debate.

Would it be a better programming practice to use subroutines instead of
variables as constants? (As my knowledge of Perl is relatively limited,
please let me know if this is not backward compatible to 5.004_04)

Let me explain why I believe that yes is the right answer. I've been
told several times and read it various times as well that a constants is
simply a function that takes no arguments and only returns a value.
Here's an example taken from formmail:

my $emulate_matts_code = 0;
my $secure = 1;
my $mailprog = '/usr/lib/sendmail -oi -t';
my @referers = qw(dave.org.uk 209.207.222.64 localhost);
my @allow_mail_to = qw(yo...@yo... som...@yo... localhost);

This should be enough to illustrate my example. Now this would be
translated into the following format under the premise:

sub EMULATE_MATTS_CODE	{ 0 }
sub SECURE		{ 1 }
sub MAILPROG		{ '/usr/lib/senmail -oi -t' }
sub REFERERS		{ qw(dave.org.uk 209.207.222.64 localhost) }
sub ALLOW_MAIL_TO	{ qw(yo...@yo... som...@yo... localhost) }

(ps, I might be wrong about the behavior of the last two... but you get
my point, at least I hope. I've been successfully writing programs using
this technique)

My main arguments in favor or this:
1. Subroutine values cannot be modified (these are constants right?)
2. Improved code readability ( &CONSTANT instead of $constant or
$CONSTANT) which really distinguishes between code variables and
constants.
3. Seperate configuration (through constants) and algorithms (that use
variables which makes it easier for newbies or people who wish to learn
Perl).

My main arguments against:
1. Most, if not all Perl programs use scalars, arrays and hashes
variables as constants.
2. Might break MWS (although I don't see why if we get the code working
right with it).
3. Extra work (in the case where it's done by me, cannot be argued as
extra work for others :o)

Whaddyatink?
-Olivier


-- 
+----------------------------------------------+
| Olivier Dragon        dr...@sh... |
| Software Engineering II, McMaster University |
+----------------------------------------------+

Re: [Nms-cgi-devel] Open question on variable/constants

[Christopher R. <red...@ma...>]

On Mon, 2002-01-28 at 16:10, Olivier Dragon wrote:
> My main arguments against:
> 1. Most, if not all Perl programs use scalars, arrays and hashes
> variables as constants.
> 2. Might break MWS (although I don't see why if we get the code working
> right with it).
> 3. Extra work (in the case where it's done by me, cannot be argued as
> extra work for others :o)

4. Doesn't interpolate directly.
ie compare:
print "Mail program is $mailprog.\n";
vs.
print "Mail program is ", MAILPROG, ".\n";
or even
print "Mail program is @{[MAILPROG]}.\n";

Re: [Nms-cgi-devel] Open question on variable/constants

[Jonathan S. <gel...@ge...>]

On Mon, 28 Jan 2002, Olivier Dragon wrote:
> I would just like to submit a question to everyone to either answer or
> debate.
>
> Would it be a better programming practice to use subroutines instead of
> variables as constants? (As my knowledge of Perl is relatively limited,
> please let me know if this is not backward compatible to 5.004_04)
>

It would be backwardly compatible as far as the Perl interpreter is
concerned but I have grave doubts as to whether using this idiom is going
to address our aims.  In the first place I really am not going to be the
one who goes through the code altering the places where these variables
are interpolated into quoted strings, which really would mean assigning
their values to new variables rather than some @{[EMULATE_MATTS_CODE()]}
hack ... and then there is the documentation and the support issues :)

/J\
-- 
Jonathan Stowe                      |
<http://www.gellyfish.com>          |      This space for rent
                                    |

RE: [Nms-cgi-devel] Open question on variable/constants

[Wizard <wi...@ne...>]

> Would it be a better programming practice to use subroutines instead of
> variables as constants?

In an ideal world, it might. However, there is much more overhead associated
with creating and calling a subroutine then using a simple scalar. When you
create a subroutine, Perl needs to create a new namespace, stack, etc., and
then when called, there is quite a bit of code that needs to be executed in
order to pass all of the information (even if there isn't any explicitly
passed) to the subroutine. Although your suggestion would work, I think it's
a bit of overkill for the sake of saying that it's "better programming
practice".

Grant M.

Re: [Nms-cgi-devel] Open question on variable/constants

[Olivier D. <dr...@sh...>]

On Mon, Jan 28, 2002 at 06:17:27PM -0800, Wizard wrote:
> In an ideal world, it might. However, there is much more overhead associated
> with creating and calling a subroutine then using a simple scalar. When you
> create a subroutine, Perl needs to create a new namespace, stack, etc., and
> then when called, there is quite a bit of code that needs to be executed in
> order to pass all of the information (even if there isn't any explicitly
> passed) to the subroutine.

Would the "use constant CONSTANTNAME => VALUE" be more efficient? As
efficient as scalars, arrays or hash?

> Although your suggestion would work, I think it's
> a bit of overkill for the sake of saying that it's "better programming
> practice".

Maybe I'm pushing it too far, but I thought the aim of this project was
to provide better quality (in delivering a service, in security and as
a coding model) In my opinion, allowing what should be constants to be
modified isn't a very secure practice. Would you do the same in C?

-Oli

-- 
+----------------------------------------------+
| Olivier Dragon        dr...@sh... |
| Software Engineering II, McMaster University |
+----------------------------------------------+

RE: [Nms-cgi-devel] Open question on variable/constants

[Wizard <wi...@ne...>]

> Would the "use constant CONSTANTNAME => VALUE" be more efficient? As
> efficient as scalars, arrays or hash?
No, all that 'use constant X' does is create an anonymous subroutine 'X'.
It's equivalent to:
*X = sub () { <value> };
which is basically what you are suggesting. 'use constant' also isn't
supported in 5.004x, which I believe is the target.

> Maybe I'm pushing it too far, but I thought the aim of this project was
> to provide better quality (in delivering a service, in security and as
> a coding model) In my opinion, allowing what should be constants to be
> modified isn't a very secure practice. Would you do the same in C?

Provided that the code is written well, it shouldn't be possible to modify
ANY variables that aren't exposed through CGI (and I should hope that only
the ones that are necessary are).

The idea that Perl doesn't really have any constants is an ongoing debate
(and the reason for the 'use constants' hack). I personally believe that
Perl manages just fine without them, and provided the developer uses due
caution in the code, it really is only a issue of resources. As far as C
goes, It's an entirely different language, and not really a practical
comparison (Perl say's "What's a macro?") ;-)

Grant M.