|Perl: the Markov chain saw|
Memoizeby ariels (Curate)
|on Nov 30, 2000 at 16:06 UTC||Need Help??|
Item Description: Make your functions faster by trading space for time
DescriptionMemoization speeds up calculation of a function by storing its previously-computed values. MJD's excellent Memoize module memoizes (almost) any function you ask it to. If you use Memoize;, applying this optimisation can be as simple as saying memoize 'slow_function';.
Additional methods exist, allowing such things as persistant storage of memoized values.
Why should I use it?Memoization is not a technique for everyday use. Only pure functions (functions which are executed only for their return values, not for any side effects) can be memoized. It's the sort of module you should download and learn how (and when) to use.
Eventually you'll write a slow routine that gets called in a slow loop. If only a small number of different argument combinations get passed to the routine, it's an excellent candidate for memoization.
The documentation contains several examples, some of which can serve as ideas for usage. The section on persistant storage of the cache is particularly noteworthy.
Why should I read it?The documentation, code and a Perl Journal article make excellent associated reading, sure to improve your Perl. Both explain in detail what memoization is, and how (and when!) to use it. In particular, the article explains how the module performs its magic, by giving a very short (but complete!) implementation of the principle routine, memoize.
Ways to use the module are mentioned, with examples:
You should also read the documentation (or the article), paying attention to the routine's limitation regarding functions taking multiple arguments which may contain the contents of $; (see "Why shouldn't I use it", below).
Finally, the author of this review is mentioned in the module's documentation. This gives at least one person an additional reason to read the documentation.
Why shouldn't I use it?Most likely, because your problem doesn't need it.
If your problem does need memoization, you might still need to help the module a bit.
Probably the single biggest drawback of memoization is that it doesn't work for functions with side effects. This is nothing to do with Memoize, and everything to do with the technique. Unfortunately, there is no way the module can check this for you. Blindly applying memoize to every slow function in your program will not work, and will cause your program to fail without warnings! Make sure to understand what you are doing, before you do it.
2001-03-15 Edit by Corion : Removed spurious links