Slashdot Mirror
MySQL Gets Perl Stored Procedures
ryarger writes "Woo Hoo! After a seeming eternity of wait, there is finally an implementation of stored procedures for MySQL. It uses Perl as the stored proc language, too!" Also note that this piece of work was done by OSDNs own Krow. Very cool work I must say.
Why didn't they use a Transact-SQL compatible stored proc syntax? This would ease migrating and also enable people who prototype DBs on MySQL to move it to either Sybase or MS-SQL with a minimal of fuss.
I'm not saying Transact-SQL is great or anything but it'd be nice if it was a bit more compatible with other systems.
My boss (Windows NT admin) and I were just discussing MySQL. We're running a number of small databases with Oracle on NT (with a University License), but we started talking about MySQL when I mentioned Slashdot was powered by it. Our web server and my workstation are Linux in NT land, and I try to plug Linux wherever I can. My boss is even learning Perl so he can code for our web server.
He liked MySQL until he heard that it couldn't do two things: stored procedures and subselects. He said "I don't see how it could be useful without those things." All of the database apps he's ever written use those.
It's great to see stored procedures being implemented. It would be even better if/when subselects are implemented. I could make a stronger case for moving some things over.
Any chance of it happening?
What I find extremely funny about this all is that Microsoft is doing the same thing in SQL Server 9.0, by putting the CLR in the SQL Server database. This way you can write your procedures for SQL Server in many languages, including Perl. OUCH. It causes more overhead than you realize to have an interpreter for more than SQL in the database.
:).
:).
I'm a DBA. I have seen the last few versions of Oracle with their Java Stored Procedure and SQLJ support, which is pretty bad. Oracle can't even get their PL/SQL running right between queries and views and stored procedures (the engine has not changed for PL/SQL since 7.3 in 8i, and 9i does not change it that much. Yes, they run 2 engines, one for SQL and one for PL/SQL. It makes Oracle perf tuning a complete nightmare). Yet they find it necessary to shoehorn a complete JVM in. No, no one really uses it, because it doesn't provide advantages.
It only makes the code completely unmaintainable since it's nothing more than code that calls the internal JDBC driver rather than an external one.
PL/SQL, T-SQL, and the other stored procedure languages at least are written in a superset of the main DML/DDL language. This allows you to use the same language optimizer, which reduces code size, and allows for code consistency across the entire project. In other words, all the queries, including parts of stored procedures, get the same optimization treatment.
Having ONE optimizer means that you can make it run really well, and share query plans and cached information. Pretty cool
The other important reason you have stored procedures is because if they are written in the main language, you can leverage the optimizer for query plans and caching of frequently-used or prototyped statements. That's part of the other reason for stored procedures. You can share queries and query prototypes with views and user queries, and have optmization that is better than what writing a procedure in X language can do.
Now we've got Microsoft coming in with their CLR, and mySQL using Perl. This is going to lead to even more unmaintainable code, because you're going to have people coding business logic that can be optimized in the DDL/DML language used in a higher-level language that cannot be.
Talk about a performance problem
From a language and optimization perspective, you always try and use a derivative of the main DML/DDL language of the database, so that you can use the same optimizer for making the statements run faster and perform well.
Anyone can write internal hooks to have a code interpretation engine in a SQL database. Oracle's been doing it for years, and so has Sybase. No one I know uses it because it doesn't provide the real advantages of stored subprograms in a database, which is to store frequently-used and prototyped query statements and aggregations in such a way so that they can be optimally retrieved versus just executed. When you add additional languages, you lose that. Oracle's Java Stored Procedures are nothing more than Java code that calls a different JDBC driver. I don't even want to think of what ADO.NET is going to do in SQL Server 9.
While this seems like a good idea, remember that it's been out for a few years in two other products, and is coming out for another. It's not as big a deal as real SQL stored procedures, because it's not as optimal as they are due to their loose coupling (which describes it perfectly IMHO), and can't share in the same optimization techniques as user SQL queries.
In other words, this isn't something to be too happy about, since it's something that people already have and don't use.
Actually, this is very useful. If you want to do say 6000 inserts using a comma delimited string or something along those lines, to open a connection and call a specific stored procedure is EXTREMELY slow compared to parsing the string internal to the stored procedure (recent test I did when designing some software showed an insert called 6000x took 16 seconds compared to 1.5 seconds parsing the string in the db). Now when you use a language like java or perl to do that internal to the db, it will drop that time even more as SQL design wise has string manipulation features but isn't meant to be doing it. I would much rather have a Java parser that I call inside my sp and have it take half a second versus 1.5 seconds to have PL/SQL parse my strings. If you actually did any development involving database interaction, this would be pretty obvious.
Looks like it might be a while. Better just get PostgreSQL in the meantime.
I wouldn't bitch about performance too much - if that's all you want, just ditch SQL entirely and use an embedded DB like Berkeley. Truth is, features/performance is a sliding scale, not a binary option. For some applications, being able to use stored procedures in a few different languages might be very helpful. For other applications, an SQL parser itself is unacceptable overhead.
Your right to not believe: Americans United for Separation of Church and
You don't know the first damn thing about database programming, do you? The stored procedure code isn't re-parsed every time it's run. The execution plan for the query is cached and *that* is run. The performance hit would only be seen the first time the SP was run, when the recompile occurs. Having multiple SP languages would be a very good thing.
No, not really, it is that kind of attitude that got MySQL into the position it's in today, everyone acknowledges it's fast, but nobody has any respect for it as a real database.
No no no, damnit. We need to get past these shitty Procedural SQL hacks. T/SQL and PL/SQL are crap, Why do you think Oracle is integrating Java and Microsoft is integrating ActiveX into the database engine? Because trying to do high-level programming in SQL is complete shit. Why would MySQL want to integrate a legacy language like PL/SQL?
A vast majority of the performance gains to be had from stored procedure programming comes from two sources -- precompilation and elimination of network round trips.
Neither of these options are precluded by a correct abstract interface.
There's no doubt that such an abstract interface would hurt performance, but I would venture to say that you would give back much less than 1% of the stored procedure benefits by doing this. You can do a large number of JNI-Like calls in the ~8ms required to complete a network round trip. Add compilation and query optimization and you have a large number Vs. a very small number.
To be able to provide a migration path for both sers of MS SQL Server/Sybase, Oracle, and DB/2 seems compelling even though and such migration path is likely to be an 80/20 proposition.
Not a bad idea, in my view.
You are not a beautiful or unique snowflake -- but you could be if you got off your ass.
Yes, but it's a nightmare trying to perform an extended update without subselects. I usually end up just writing a small Perl proggy to do it with DBI.
Wooden armaments to battle your imaginary foes!
You know, most of the time someone using a subselect just lacks the imagination needed to do a join :)
While some people will always do brain dead things, there are definitely many queries that you simply cannot write unless you do subqueries.
Consider something as simple as finding all students whose IQ is above average.
Oh very true. Sometimes you can't get around using subselects (there are a couple of places in Slashdot that are exactly in this situation).
There are other things I would rather see though. For instance in Oracle you can represent Graph structures in the databases (quite cool). I would love to have this for comment storage. Right now it is quite a pain in the ass to generate the comment pages in threaded/nested mode.
You can't grep a dead tree.
Actually the code for the DB is very modularized. You just have to exchange one library and you can use a different databases. I have tinkered with this for PostgreSQL and Oracle but never for any other DB (while I have experience with Sybase and Informix, I have never used DB2).
I have had people from IBM approach me about doing a DB2 port, but no one has every offered any code and I have enough to do as is.
One of these days though I would love to setup slash on an IBM mainframe and actually benchmark it. I suspect it would run quite well (even with MySQL).
You can't grep a dead tree.
PERL????
Jesus, PERL????
You know, the strength of query languages is that you don't have to use (and in face, are usually punished for using) loops and cursors to make massive changes. Perl is the most loop oriented language on earth. And even if, underneath it all, the optimizer is turning your code into a loop anyway, it's goddamn doing it more efficiently than Perl ever would. This addition is NOT going to increase the likelihood of people migrating from sybase or other TSQL based databases to MySql...it's going to increase the number of hardliners who feel that MySQL is a pathetic ghost of "real" servers, and as such decrease the cadence of better open source solutions like PostGreSQL. MySQL and Perl...it's fast becoming a database for control freaks who don't believe in doing anything automatically, or allowing the machine to do our optimizations for us -- and that's what computers are all about, goddamnit!
It is nice that there's finally a way to perform object operations on a server without performing the logic in scripted code, and it's nice that MySQL is trying to make a grab for usefulness beyond its INSERT, SELECT, DELETE simplicity. But Perl is not a standard language in the DB world...it's asking for DBAs and programmers used to TSQL and looking for a cheaper, freer alternative to gain new custom knowledge that is complex and no better then the knowledge they already have! All those linux sysadmins to have a little database are going to be overjoyed...but for the rest of us, this is totally useless, just like the rest of MySQL's features.
Hey freaks: now you're ju
- UPDATE 'foo', map {
... } grep { ... } sort SELECT 'bar', ...
This is the heart of the power of Perl, and if the interface is built right, it could be a huge boon to database work.Of course, done wrong it could be slow, difficult to maintain and immediately obsolete.
" if they put hooks into a generic MySQL facility which allows *any* programming language to serve as a SP language in the server. Why can't I use Python?"
Postgresql does this. You can use python, perl, TCL, and PL/PGSQL.
They are debating loading up Java but there seems to be some resistance from the hackers.
War is necrophilia.