I have been slowly migrating old code to use Perl modules at work. I found that I was fixing the same bugs over and over in different scripts. I found it was better to fix the bug once, and then have the different scripts use those modules.

Here are my requirements:

• No OO - the other programmers don't know object oriented Perl or the associated syntax
• Don't duplicate the same function across modules

My modules are traditional - just listing functions and exporting them for use in scripts. The modules seem to make sense - the SnapMirror functions are in the SnapMirror.pm module, the company-specific validation functions are in Company.pm, the functions relating to backups are in Backup.pm, and so on. The modules probably export more than they should, but I don't believe that to be the root of my problems.

Company.pm - functions specific to my company

• get_hostname() - given a FQDN, return the hostname. It's a function because we had planned to put more error checking in here, like only removing the DNS domain of the actually site the filer is in, and emitting an error if the DNS domain doesn't match the site.
• get_site_from_filer() - given a filer, return the site it's associated with.
• is_clearcase() - is the qtree in question a ClearCase qtree? Decodes qtree name mangling and/or uses a lookup to see if the qtree holds ClearCase data.
• is_valid_site() - is the site specified a valid site?

SnapMirrors.pm - functions specific to NetApp SnapMirrors

• get_snapmirror_source() - given a destination qtree, get the source qtree. Calls is_valid_site().
• is_snapmirror_destination() - is the given qtree the destination of a SnapMirror? Calls is_valid_site() and get_site_from_filer().

Basically, my code broke when I tried to make an enhancement. We wanted the is_clearcase() function to not only detect qtrees containing ClearCase data, but also to detect qtrees that are SnapMirror destinations for ClearCase data. This made is_clearcase() call is_snapmirror_destination() and get_snapmirror_source().

Searching around on the rest of the site, I see comments stating that circular dependencies are bad, but I don't have a good way of resolving this one.

I don't want to omit the error checking. It has helped us find errors in other scripts before. I also don't want to put more work on the other programmers. It's possible that every time that is_clearcase() is called, they could check for the is_snapmirror_destination() case, but it seems like this functionality belongs in the is_clearcase() function itself.

It looks like the code would work if is_clearcase() was in the SnapMirrors.pm module, but the problem is that the other coders wouldn't know to use SnapMirrors.pm when they wanted the is_clearcase() function - it doesn't seem to belong to that module.

I looked in the camel book, and they mentioned as an aside that one would use require instead of use "if you have two modules that each need a function from the other. (And we're not sure that's a good reason.)" They don't mention anything else about it.

There are other posts on this site about using require, import, and BEGIN blocks. They argue about design, but don't really provide helpful information, outside of "don't make circular dependencies."

Finally, I was messing around with a module that would use all the other ones, and then manually export functions that pointed to the correct functions. It works, but it also seems inelegant. It also seems unmaintainable.

I figure it would work if I put everything in one monolithic module, but that doesn't seem to be an elegant solution.

I seem to have coded/designed myself into a corner. Any help would be appreciated.

Thanks in advance,

Chris