MySQL vs PostgreSQL
wikivs.com
wikivs.com
oh yeah, go postgres!!! ;)
I do like SQLite, just not for anything where you cant control the writes per second, or need it to be high.
The driver you use to connect to the database is also very important. If I'm not mistaken the PHP driver was pretty terrible a while back. It may be better now. It didn't handle the locks well which sometimes resulted in deadlocked queries with only a handful of users.
Personally, I wouldn't run any multiuser application doing a significant number of writes on SQLite except during early development. I think where it really shines is when you need a lightweight embedded SQL database without the hassle of having to install something extra. Think Amarok or something similar.
I ran Trac on SQLite for a while. Setup was a breeze and performance was reasonable with 3 or 4 users, but a switch to PostgreSQL was a night and day improvement.
If you're set on using an embedded database you might want to check out Firebird or H2. I haven't used Firebird myself, but I have had excellent results with H2. You have the flexibility to use it a few different ways - embedded vs. server mode and in-memory vs. disk storage - and it is very compatible with MySQL syntax so it's useful as a drop-in replacement.
Also, you can consider sandwiching several updates between BEGIN and END. I've discovered this dramatically improves performance. You just query "BEGIN" to start, execute your updates, then query "END" to process them.
It's a flimsy reason tbh but it's all I had to go on. Having said all that I do really love PostgreSQL now after 4 years of using it (via SQLAlchemy).
As ORMs become more prevalent, things like query language start to become less relevant, but so do relational databases. I really don't care how my data is stored, as long as it's safe and performs well.
Would someone mind quickly posting the benefits of that architecture as opposed to focusing on one 'storage engine' and making it really good?
Would you rather have 1 type of car that the government makes that works pretty reliably and is a middle-ground car, or would you rather have a selection of cars to chose from; sports, truck, sedan, minivan, etc. ?
Everyone has different needs. I use MySQL extensively for it's memory-based storage engine, MyISAM for quick & dirty non-escential data i/o, and InnoDB for when data needs to be managed securely for processes.
I would use PG for transaction-based data processing in a high-volume situation with multi-processing clusters. Otherwise MySQL works fantastic for all my needs.
Uh, the definition of flexibility is lack of consistency between the options. Otherwise, it's a false choice.
Well, I pose you this question:
Would you rather have a car that runs into a wall and explodes and kills everyone inside or no car at all? The car is like using car analogies on Slashdot and not having a car is like not using car analogies on YC.
In conclusion and in summary, go back to Slashdot. KTHXBAI.
For example, if you needed to store key => value pairs, how would you do that? If you use an array (array[0] = ('key', value')), you would have to iterate over the entire array to find the key you're looking for (Order n), but if you use a hash table, you can do key look-ups in constant time (Order 1).
MySQL's different storage engines provide similar opportunities to use different ways of storing data for different ends. MyISAM has amazingly fast read speeds, but is really poor under write conditions. Sounds like a crappy storage engine, right? Well, what if you have a table of zip codes and their geo-coordinates? How often are you going to be writing new data to that table? Every couple months in a batch update? So, if you put your postal codes in a MyISAM table you can take advantage of high read rates and ignore the fact that MyISAM is terrible for writes since you don't really write to the table.
Oh... This is not Slashdot. Sorry.