Christopher Schultz wrote:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
David,
I'm already on my 3rd attempt at optimization for this section. The
first round was having the db do _all_ the work, submitting a
complex query (a view, actually) and returning a resultset with all
the data I need. The query took forever, and the java code to produce
the report was really fast.
I would bet on the speed of the (properly indexed and tuned) database
against Java any day, honestly. These things were built to perform these
The major slowdown in the db is related to a weak (not horrible, but not
great either) initial db design, plus some design requirements that make
writing clean queries nearly impossible. There are some things that
java (or most any programming language) is better than databases at, and
the comparing (not just summing up) data from sequential data records is
one of them (unless you are using a specialized data warehouse database,
which we are not).
types of requests. No offense to your abilities, but have you had a DBA
check out your tables and indexes to see if they are properly supporting
the queries you are attempting to execute? It's easy to write a
long-running query, but just a few tweaks can reduce the query time by
orders of magnitude.
My primary job is database design and management; the java side is where
I'm weak, and I've spent a LOT of time on these queries, including
pushing through some design changes that have helped a lot, but not enough.
One issue I ran into earlier today with plain HashMaps was that I
couldn't figure out how to search for a specific piece of data, which
requires matching on a site number, a date and a shift, and for some
data another date.
Right: you need to hash the 3 data items together. Typically, you would
write a class that bundles the 3 data together and provides a hashCode
method that incorporates them and an equals method that checks them.
Then, use that as the key to your map. Is there any reason why you can't
use the data objects themselves in this way?
map.put(myDataObj, myDataObj)?
I thought of that, but there is a lot more data in the object than just
the stuff I need to search on. Once I identify the particular object
instance I need (which only needs those 3 fields), there are about a
dozen other fields I need to retrieve from it to process each record.
I have a workable solution right now, though the code is rather more
complex than I would like to keep my navigation of all those TreeMaps
in sync.
When in doubt, abstract: write yourself a helper class that bundles all
this odd logic together. Then, your business code will be a lot simpler
to read, and you can separately unit test your
SpecializedTreeSetHashMapLinkedList class (or whatever). ;)
Yeah, that's what I'm working on right now. It's coming together,
though somewhat slowly since I'm still getting my head around the
different kinds of maps and lists, and what each of them is good for.
Thanks!
D
---------------------------------------------------------------------
To start a new topic, e-mail: users@tomcat.apache.org
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
