The principal reason in this case is that I try to minimize the number of non-core modules that my code uses, since each non-core module introduces one more potential obstacle, however small it may be, for someone else using my code. I.e., I'd never use a non-core module just to import a simple one-line function. More generally, I weigh the tradeoff between coding something myself and introducing a non-core dependency. The bigger and more complex the candidate module (i.e. the more potentially difficult its installation by some other user), the more I am willing to code myself.
But I don't want to overstate this point! When I speak in this context about "coding something myself" I am usually referring to a small bit of functionality that I may be able to implement in only a few lines of code. Of course, this also assumes that I feel confident in coding something myself; there are many problem areas that I don't understand well enough and am glad to use modules written by others, even if my needs could have been met with only a few lines by an expert.
One other reason why I may code something myself is if I find the documentation for a module so poor that I suspect that the apparent time saving of using the module instead of coding it myself will be lost trying to make up for the bad documentation (e.g. by wading through the source).
A third reason is that sometimes the available CPAN alternative is not good for some reason (usually speed).
But again, just to be clear, even with all these considerations, I end up using plenty of modules, both core and non-core.
the lowliest monk
In reply to Re^3: Calculating statistical median
by tlm
in thread Calculating statistical median
by Lhamo_rin
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |