in reply to elsif chain vs. dispatch

Even though looking up an element of a hash tends to be an O(1) operation, its algorithmic complexity only describes the scaling factor of the algorithm as n increases. If you rearranged the if/else branches in decreasing order of probability according to your expected corpus, it would be faster than the hash lookup until you reached (wild guess) a dozen or so entries in the hash, or if your working corpus deviated significantly from your expected corpus.

A good benchmark with working example data would reveal more, but remember that:

Replies are listed 'Best First'.
Re^2: elsif chain vs. dispatch
by sflitman (Hermit) on Apr 26, 2009 at 23:50 UTC
    For me, it is that the codebase is becoming unwieldy, so I wanted to switch to the dispatch style, and I was (honest) going to be putting it outside a sub so it only gets compiled once!

    Have you ever considered using the symbol table as the hash, would this be any good:

    $::{"handle_fieldcode_$code"}->($args); # for $code = 'A'..'Z'
    ??
    SSF

      The slow bit is the function call, not the hash lookup which you'd also have to do with the symtab anyway. With a hash, you have better control over what inputs are allowed and how they are dispatched.

      With symtab:

      my $handler = do { no strict 'refs'; \&{"handle_fieldcode_$code"} }; ...handle expections somehow?... die if !$handler; $handler->(@args);

      With hash;

      my $handler = $lookup{$code}; die if !$handler; $handler->(@args);

      with a one-time setup of

      my %lookup = map { no warnings 'refs'; $_ => \&{"handle_fieldcode_$_"} } 'A'..'Z'; ...modify %lookup for expections...