Yeah, great. Let's visit this alternative reality for a minute.
Word goes out that a 'young guy with a handgun' is wandering around campus killing people. In response, all these students break out their handguns and start popping off at anyone matching that description.
Hilarity ensues.
This idea has been around for a while; at least since the late 1980's. The motivation for these kinds of DBMS features is that there are lots of programming situations where SQL's types and expressions aren't powerful enough, and that the language doesn't have a lot of modularity. User-defined functions are supposed to overcome this limitation.
This kind of feature brings MySQL closer to being an "Object-Relational" database.
From a theory point of view it goes a long way towards implementing the Relational Model's idea of a 'domain' (not just INTEGER, VARCHAR or whatever, but PART_NUM, PERSON_NAME etc). This is supposed to improve the integrity of the data in the database.
From a practical systems point of view it can have a big performance impact. If you're opening cursors and then looping over some Java code on the client to identify only those result rows that you're interested in, then you're paying a pretty big 'system tax' to transfer the data from the DBMS through the connection and into the external program's address space. Pushing the code (which will have to run anyway) into the DBMS eliminates the transfer overhead.
The point of the original Postgres was to figure out how you incorporate these features into a query processing framework. Most modern DBMS products have the feature; some of them do a better job implementing it than others.
Slashdot Mirror
Yeah, great. Let's visit this alternative reality for a minute. Word goes out that a 'young guy with a handgun' is wandering around campus killing people. In response, all these students break out their handguns and start popping off at anyone matching that description. Hilarity ensues.
This idea has been around for a while; at least since the late 1980's. The motivation for these kinds of DBMS features is that there are lots of programming situations where SQL's types and expressions aren't powerful enough, and that the language doesn't have a lot of modularity. User-defined functions are supposed to overcome this limitation.
This kind of feature brings MySQL closer to being an "Object-Relational" database.
From a theory point of view it goes a long way towards implementing the Relational Model's idea of a 'domain' (not just INTEGER, VARCHAR or whatever, but PART_NUM, PERSON_NAME etc). This is supposed to improve the integrity of the data in the database.
From a practical systems point of view it can have a big performance impact. If you're opening cursors and then looping over some Java code on the client to identify only those result rows that you're interested in, then you're paying a pretty big 'system tax' to transfer the data from the DBMS through the connection and into the external program's address space. Pushing the code (which will have to run anyway) into the DBMS eliminates the transfer overhead.
The point of the original Postgres was to figure out how you incorporate these features into a query processing framework. Most modern DBMS products have the feature; some of them do a better job implementing it than others.