</lurk> Hello,
I am currently in the early stages of writing a (parrot targeting) compiler for an interpreted language I have been working on/with for the past 4 years.
Background:
The language (Basil) is basically a functional (ish) language with a C-like syntax that has an implicit concatenation (or list join for efficiency) between each statement. It kinda looks like slightly funny perl with tags in.
Thus a large tree of strings is efficiently filtered and a big chunk of text is thrown out at the end. It is used for creating very flexible web templates.
The original Basil was created for a company I worked for who wrote everything in C despite doing only web-based work. Perl and PHP were banned. So I wrote a scripting layer that would be handed a symbol table and return a huge chunk of formatted HTML. I guess these days most folks would use XSLT for this.
However, I still find XSLT a closed-book and would love to see Basil running on Parrot, being called from a Perl Web Application to spit out formatted HTML. Especially as you could link the two lots of code as one and use the usual calling convention to call the Basil as if it were a standard Perl sub (or Python, etc). Plus Basil has of course grown over the years to be an object-oriented, stand-alone CGI environment that is very good at one thing - making web pages - and I find it hard to work without it.
Problem:
1) To create an efficient implementation of Basil I'll need a big PerlArray that is formed by expressions that create PerlStrings or more PerlArrays. At the end the tree of PerlStrings and PerlArrays is iterated over and concatenated together and returned (or just printed out). The PerlArray is just a way of avoiding even more concats plus it allows you to build list stuctures to iterate over.. By using the standard Perl ones I'm hoping to be easy-to-call plus I get the added bonuses of stretchy arrays, int-to-string converstion etc.
So my main question is this: Are PerlArrays and PerlStrings an efficient way to go with this? - how much copying goes on within a string concatenation? Iterator? etc ( ie does using COW prevent a deep copy every time I make a concatenation).
I tried to make sense of PerlString concatenation and they seem to use string_concat which in turn has two mem_sys_memcopys doing the actual concatenation.
2) If concat isn't the way forward then push probably is (I could just hand back a tree and iterator (as a closure err do I mean continuation?) to the calling function). I take it there are no hidden caveats in using push on a PerlArray?
3) I take it a proper iterator is far more efficient than a loop with a shift in it? eg:-
.sub _printall .param PerlArray xs
.local PerlString s
.local int t
loop:
unless xs, endloop
shift s, xs
typeof t, s
if t == .PerlArray goto recurse
print s
branch loop
recurse:
_printall(s)
branch loopendloop:
.pcc_begin_return .pcc_end_return .end
4) why doesn't '.' work on a PerlString in IMC? :-)
5) I guess the way-forward for the future is a custom class of very-lazy string-lists that can be concatenated, split, iterated, searched, sorted etc by always doing the least amount of copying they can get away with. But I'll have to get my head round Basil OOP first (and to be honest, PerlArrays & PerlStrings have 90% of what I need if they too are very-lazy). Would you agree?
I may of course be just plain getting things wrong (I still can't get iterators to work properly).. just trying to get my head round it all.. so feel free to humiliate me (or indeed ignore me)..
Anyway - I'm having fun. Thanks for all your hard work so far It's most appreciated.
Any flames, advice & encouragement would be gratefully accepted.. Cheers, Sambeau [EMAIL PROTECTED] <lurk>
