Global or not global

Hi all,

I want to ask you for your opinion to the following general problem:
- As anybody would agree with: global variables are bad.
- That means: Feed as much information to a function as you need to do the job. That also is a building block to make the function well testable.
- In a program which needs a database more or less all over the codebase you have to have a database handle.
- Asuming the above statements are right you would have to provide the database handle to a function which needs that database handle. That's not too bad at the moment.
Now you're using the several functions

sub a {
    my ($dbhandle) = @_;
    ##does soemthing with dbhandle
}
sub b {
    ##doesn't need dbhandle
}
sub c {
    my ($dbhandle) = @_;
    ##does soemthing with dbhandle
}
[download]

Function a uses b and b has to use c. That means you have to insert dbhandle in the function signature of b even if b doesn't need it by itself.

So, as more or less the whole application needs the database it happens that more or less all functions need a dbhandle. That means that the signature of more or less any function does have the dbhandle parameter.

Is this the right way to do this? Or would it be better to implement the functions the way that they can "grab" the dbhandle from somewhere (singleton = global variable)? When you add other things to the example above, e.g. a logging object, than the function signatures are growing with elements which you can find in more or less any function. (Logging is IMHO a much better example of an item you need everywhere, but logging is not a requirement to solve the initial business requirement. Log::Log4perl solves this elegantly by providing a singleton. So you can add logging whereever you want simply by calling get_logger).

Advices and ideas welcome.

Best regards
McA

Comment on Global or not global Download Code

Replies are listed 'Best First'.
Re: Global or not global by davido (Cardinal) on Jul 03, 2012 at 05:43 UTC
One option is to put your data manipulation in a class (or classes), and as it is instantiated, pass a database handle to the object. Then the constructor requires that you pass a handle to it. Store it as an object attribute. If you're using something like Moose you could encapsulate your database handle in a role and mix it into classes that require access. Of course "mixins" or roles aren't unique to Moose; you could achieve similar with good old-fashioned Perl objects too. Dave	[reply]
Re: Global or not global by moritz (Cardinal) on Jul 03, 2012 at 06:29 UTC
(singleton = global variable)? It is global state that is a problem, not just global variables. And yes, singletons are also global state. So you win nothing by using a singleton over a global variable. So you have several options: you can understand why global state is bad, and still decide to use it. Or you can go the OO route where the database handle is available. Perl 6 - the future is here, just unevenly distributed	[reply]
Re^2: Global or not global by tobyink (Canon) on Jul 03, 2012 at 10:14 UTC
Indeed - global variables are a symptom of the underlying design problem, they're not a problem in themselves. `perl -E'sub Monkey::do{say$_,for@_,do{($monkey=[caller(0)]->[3])=~s{::}{ }and$monkey}}"Monkey say"->Monkey::do'`	[reply]
Re^2: Global or not global by BillKSmith (Monsignor) on Jul 30, 2012 at 02:41 UTC
In this case, I reluctantly vote for the global solution. The required object code would probably include most of the program so object variables would be little different than globals. The increase in code complexity would cancel any small improvement in encapsulation.	[reply]
Re: Global or not global by BrowserUk (Patriarch) on Jul 03, 2012 at 05:48 UTC
Log::Log4perl solves this elegantly by providing a singleton. So you can add logging whereever you want simply by calling get_logger). How is calling `my $log = Log::Log4perl->get_logger("My::MegaPackage")` in every function or method: elegant? Surely they could have fitted another couple of "log"s in there somewhere. Maybe `my $log_logger = Log::Log4perl->get_log_logger( ... )` With the rise and rise of 'Social' network sites: 'Computers are making people easier to use everyday' Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error. "Science is about questioning the status quo. Questioning authority". In the absence of evidence, opinion is indistinguishable from prejudice. The start of some sanity?	[reply] [d/l] [select]
Re^2: Global or not global by McA (Priest) on Jul 03, 2012 at 06:54 UTC
Hi, I use it it this way `use Log::Log4perl qw(get_logger :nowarn); ... get_logger->info("Shit happens");` [download] IMHO when you have decided to use Log::Log4perl then it is in the installation prerequisites. To have a function name like 'get_logger' is unique enough for loading it in your current namespace. When Log::Log4perl is not initialized, the code doesn't hurt. So, whenever I need an additional log statement, I just insert `get_logger->error("Somewhat");` That what I meant with 'elegant'. Or call is 'handy'. What is your solution to those cross cutting concerns like logging? Best regards McA	[reply] [d/l] [select]
Re^3: Global or not global by BrowserUk (Patriarch) on Jul 03, 2012 at 07:23 UTC
What is your solution to those cross cutting concerns like logging? What is logging? Simply stated, it is printing to a filehandle. Pretty much everyone is comfortable using STDOUT and STDERR. And STDERR serves all my logging needs. But if you want your logging to go to a different place than STDERR, what's wrong with (say) STDLOG? Point it to a file or port as needed. Want to turn logging off, direct it to the null device. Want your logging to add boilerplate (timestamps, callstack, threadids etc.), tie *STDLOG. Once it is tied, routing to multiple destinations is easy. IMO, L::L4p is a "let's pretend we're java bunnies", grandiose overkill, behemoth of a ~~package~~ dynasty of packages, that makes writing a few bits of text to a file, laborious, complicated and expensive. With the rise and rise of 'Social' network sites: 'Computers are making people easier to use everyday' Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error. "Science is about questioning the status quo. Questioning authority". In the absence of evidence, opinion is indistinguishable from prejudice. The start of some sanity?	[reply]
Re^4: Global or not global by McA (Priest) on Jul 03, 2012 at 09:38 UTC
Re^5: Global or not global by BrowserUk (Patriarch) on Jul 03, 2012 at 14:21 UTC
Some notes below your chosen depth have not been shown here
Re: Global or not global by cavac (Prior) on Jul 03, 2012 at 11:22 UTC
As anybody would agree with: global variables are bad. You mean like STDERR, @ARGV, %SIG? *SCNR* Sorry for any bad spelling, broken formatting and missing code examples. During a slight disagreement with my bicycle (which i lost), i broke my left forearm near the elbow. I'm doing the best i can here...	[reply]
Re^2: Global or not global by McA (Priest) on Jul 03, 2012 at 19:50 UTC
I waited for that. :-) The interesting aspect of these variables is: They are just there. Everyone expects them to be there. Regularly they are part of the runtime environment everywhere. So IMO they could be interpreted as a service which is always accessible. How is this different to a global variable which you introduce as a part of your own "runtime environment"? Best regards McA	[reply]
Re: Global or not global by locked_user sundialsvc4 (Abbot) on Jul 03, 2012 at 15:19 UTC
Historically, I have dealt with the problem of “global DB handles” and other such things simply by defining a package named `Global`. This package contains functions which produce the necessary values, if necessary doing so on-demand. For example, `my $dbh = Global::DBHandle;`. (If you need to use more than one database, simply have a separate function to automagically produce each handle.) You can replace this functionality with stuff to, say, use a DB handle cache, without affecting anything else in the program. And that is what I consider to be a win: there is a globally-available yet concretely documented way to obtain the necessary values and handles, and each is a function that returns it. If anything “has to happen,” here is a central and obvious place to put that. It is crystal-clear what is being done and why. This has for a long time been my standard practice.
Re^2: Global or not global by xtpu2 (Acolyte) on Feb 18, 2014 at 10:07 UTC
Sorry for posting in an old thread, but I'm faced with the dilemma being described, and yours seems to be one of the more elegant solutions. Could you share the code you would use for something like `Global::DBHandle`?	[reply]
Re: Global or not global by Anonymous Monk on Jul 03, 2012 at 06:29 UTC
Function a uses b and b has to use c. That means you have to insert dbhandle in the function signature of b even if b doesn't need it by itself. I don't think it means that, I think it actually means dbhandle should be an attribute instead of an argument	[reply]