Basically, almost any tree structure or graph, will use less memory and have faster traversal times, if implemented in C using direct pointer access and manipulations, than when the same structures are built using Perl's references.

As fast as Perl's hashes and arrays are, you do pay some costs for their generality and ease of use. Indirecting through a reference costs more than indirecting through a raw address, and raw C-style arrays use considerably less memory than the equivalent Perl arrays.

However, for those costs, you get a whole slew of benefits:

• Automated memory management. (No alloc/free to contend with).
• Automated expansion/contraction. (No realloc/memcpy et al).
• Autovivication. (No need to predeclare the size of your arrays).
• Tranparent manipulation, coersion and mixing of strings, integers and reals.
• Automated pointer management. It's (almost) impossible to segfault Perl unlike C.
• Programmer productivity.

For anything other than the most demanding of applications that need extremely fast response times, or to manipulate huge amounts of memory, the gains far outweight the costs. And even when you do need to manipulate large volumes of data, a little extra effort can sometimes allow you to accomodate upto an order of magnitude more data in the same amount of ram, whilst still avoiding the need to leave the convenience of Perl and get down and dirty with C.

That said, a well thought out set of tree primitives written in XS, available in Perl would be a nice addition.

Edited by planetscape - added <ul></ul> tags