Hashing (rambling)


#1

I was wondering if you ever considered adding a hash function to the «value types» of juce?

My reasons for suggesting this is, as you may already have guessed, related to my PILS language and its data model, but I think it might very well be useful to C++ programmers in general.

One might want to analyse images using a hash table for the colours, or perhaps optimise complex text rendering tasks by some scheme that implies hashing of fonts, just to mention a few possibilities.

(The Colour class, obviously, is rather trivial to hash, but having a named hash function would still be nice when working with hash table templates.)

The String, Typeface and a few File related classes already have an int hashCode() function, so I suggest to define it for a larger set of classes, generally those with copy constructor and == and perhaps = operators, and no virtuals.

This wouldn’t need to be done by hand. I have PILS code that parses the juce headers, and I’ve written apps that do GUI-controlled operations on VB6 forms, so I think I could, without too much pain, write an app that analyses the headers, shows a checkable list of classes deemed to be candidates for hashability, and show the code changes in a preview as you walk the list (this is the sort of stuff I use the caret-scrolling TextEditor for). An implementation might also be generated, based on the fields of the class - but this might need a bit of hand-editing.

If you’re interested (or perhaps just curious to watch your beautiful code being molested by the evil jucePILS chimera), you could write a hint or two about how the hash functions should look, and I’ll give it a try. (You would have to run the app on Windows XP or similar, I haven’t done a linux jucePILS yet.)


#2

Yes, adding more hashes to classes is certainly a good idea. Would love to see what your parser code comes up with, though surely writing an algorithm to generate the right functions would take much longer than just writing them all by hand!