I read the announcement about Oracle and the Open Invention Network. When I get back to San Francisco next week I hope to hear more about it from my team who are working faithfully to make Linux better.
In the meantime here's a PHP performance tip from Krishna Mohan.
Oracle has the concept of a client character set that application data is in. Oracle will convert data from the character set of the stored data into the client character set when data is fetched and vice versa when data is inserted. There may be a conversion cost or data loss if the client character set doesn't match the database character set.
The database character set can be chosen when the database is created. You can find what character set a database uses with this query:
select value
from nls_database_parameters
where parameter = 'NLS_CHARACTERSET';
But you knew all that.
What Krishna pointed out was that specifying the client character set name as the optional fourth parameter to oci_pconnect() has a postive effect on performance. If you let the value default then PHP has to invoke code to determine what the client character set should be. This lookup may involve a potentially expensive environment check.
Test it yourself. Benchmark a few thousand oci_pconnect calls with and without the character set:
oci_pconnect("hr", "hr", "//localhost/XE", "AL32UTF8");oci_pconnect("hr", "hr", "//localhost/XE");I used AL32UTF8 which is the character set of the Oracle XE "Universal" database I have.
Testing with the database and PHP both on my little old machine I was getting figures like 0.6 vs 2.7 seconds for 50,000 pconnect calls in the one script. I saw some other results that showed only a three-times difference.
This tip is a snippet of information: use it wisely. The cost may not be anywhere near, say, the cost of writing a query poorly.
