Squirrel

Hello, I really like Squirrel but recently I discovered the need for a little addition to the language: Compile time string hashing. I wanted to help out so I went ahead and implemented the feature.

The patch can be found here, I created it using Beyond Compare so I'm not sure if it will work with GNU patch - but the modifications are quite restricted and to manual apply it wouldn't be too hard.

When the patch is applied a new string prefix operator '#' is added such that during lexing the string following is passed to a user provided hashing function. The lexer then creates an integer token instead of a string using the value you've generated.

You can now write Squirrel code like this and get.

   a <- #"WALK_ANIMATION"
   print ( a );

> a = -347238472348  // or whatever you return


The interface is very simple, you write a hash function with the signature

 (SQInteger)HASHFUNCTION( const SQChar* string );

 and bind it to the compiler using:

  sq_setcompilerhashfunction( SQVM, HASHFUNCTION );


I've use this feature, as you might have guessed, to speed up runtime animation lookups. I hope this helps other people too.

Regards,
David Walters

maybe using a string name in place of an operator # for the string hashing function....  the operator # often has different meanings in other script languages like Perl so Squirrel may want to reserve it for future use...

squirrel strings already carry a precomputed hash value, wouldn't be the same to add a function like sq_gethash(v,idx) ?

Alberto

My first thought was 'oops' - I didn't realise this data was already stored for string literals. But as I think about it you do get a few small advantages from hashing in the lexer.

* Reduced memory overhead (and possibly reduced dynamic allocations) for compiled code.
* Obfuscation of strings in compiled code, which would help for scripted passwords.
* Flexible and consistent hashing function, by specifying your own function you can use an algorithm that suits your application. ( Although if I were to improve this patch I would set the default compiler hash function to be the standard Squirrel one, having no default and an error case was a mistake. )

I'm not sure if this patch is a candidate for inclusion in your language, but I hope it demonstrates a useful new feature that may be of benefit to someone.

As atai points out the '#' prefix probably isn't the perfect choice for decoration due to conflicting usage in other languages ( it was easier to code detection for than say (hash)"FOO" ). Also I understand the overhead involved in updating the documentation and the apparent bloat which might go against your project aims.

David