Kay Röpke wrote: > Hi! > > On Aug 4, 2008, at 12:27 PM, Daniel Fischer wrote: > >> With antlr, I'd assume? You should use it to replace drizzle's parser, >> so we can get query rewriting on the AST level ;-) > > > The secret plan :) > > In fact, I fully plan to do just that. For a long time now, I wanted to > have the opportunity to modularize the parser code so we can have > multiple ones: > One for raw speed, one for funky things and most importantly one which > is not SQL ;) > > The first step here would be to make a clean API out of the current > (YACC-)action-based calls to the items classes and then separate the > parser out into a module. > > It goes further than that, of course. With a little work the protocol > can be extended to request a certain parser for a request which would > greatly enhance the flexibility of drizzle (and MySQL of course, too). > > This is also the way I intend to bring back "prepared" statements, but > on the client: > Parsing happens on the client and all that gets sent is a flattened > tree. Much simpler to parse because there are no string operations > involved. The downside is that validating a parse tree can be tricky and > allowing it is possibly a security risk. Much easier to provoke crashes > by sending a faulty tree than a faulty string. We'll see.
Nice to see that come around, I think it is a great idea! > I've also wondered about the merit of using a bytecode format for > statements instead of a parse tree, but that's even further down the line. At one time, I was considering using FORTH as "bytecode": it has all that is needed to and can easily be extended. :) It is also so compact that it readily fits into the cache of a modern CPU, which does wonders for speed. The only drawback is that it is a stack-oriented machine, and I think that a register-oriented machine would prove to be significantly more efficient on CPUs with pipeline and cache. Also, one disadvantage is that FORTH is postfix, and it is easier to work with a prefix format when it comes to trees (hence also expressions). /Matz > > cheers, > -k -- Mats Kindahl Lead Software Developer Replication Team MySQL AB, www.mysql.com
begin:vcard fn:Mats Kindahl n:Kindahl;Mats org:Sun Microsystems adr;quoted-printable:;;Tegv=C3=A4gen 3;Storvreta;SE;74334;Sweden email;internet:[EMAIL PROTECTED] title:Lead Replication Software Developer x-mozilla-html:FALSE version:2.1 end:vcard
_______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
