[TYPO3-dev] The extbase dilemma

Mathias Schreiber [wmdb] mathias.schreiber at wmdb.de
Thu May 19 16:20:24 CEST 2011


Am 19.05.11 15:18, schrieb Franz Koch:
> that was probably back in 2009 - but I think things changed a lot since
> then.

If you say so, I hope it has.
I can only say that from my perspective, FAL and this discussion you see 
that people feel offended when you point out design flaws.
And to be honest, I totally get why, I wouldn't wanna hear my baby is 
ugly, too.
But all that personal-feeling stuff aside, there are flaws. Period.
If I am not allowed to point them out, well...

> So the conclusion I read in your posts is to stop being a backport of
> FLOW3, keep the good ideas/concepts, drop the rest and start building
> yet another new framework attempt but this time with enterprise
> requirements in focus regardless of code quality when it comes to design
> flaws?

Well, you tend to the other extreme.
I start questioning my entire existance since I don't know how to make 
myself clearer (maybe Email as a medium sucks here :)):
Basic rules:

If you write code, consider EVERY method call expensive.

Consider every SQL query expensive.

Learn to multiply - a quick "one more query here can't hurt" can snap 
your neck tomorrow.

Design (read: not test) with both larger datasets and more complex 
models in mind, if you don't have them at hand or can't imagine them, 
ask for help.

Test early, test often is a paradigm that also applies to profiling.

Don't tunel vision, keep things outside of your current method or class 
in mind. How often will this thing be called? How expensive is it? Can I 
afford it? Can I speed it up.

Make educated decisions. Can I get my data quicker? Can I store things 
anywhere else than in my DB to acces them faster? Does my temp store 
fill up my RAM?

Leave things inside their domain (funny one). There is nothing faster 
for remote data access than a database. If you don't know how to get the 
full potential out of one, ask someone to help. Why? Those things are 
built for this.
If you are not 100% sure you can build a better ORM tool than X, Y or Z, 
let it be. Use what's there.

Btw... just did a funny test on MSSQL server. I can use one query to get 
my entire model (nested over 5 levels) as XML, then use 
t3lib_div::xml2array on the resultset and feed it to Fluid... sweet.
Did I mention it's fast as hell?

So if you were a lunatic you could try to code a stored procedure that 
would deliver your model as a serialized PHP object.
Whoo haaa.

ok, enough of coding fun, but I gues you get my point.
To short-quote Dmitry: Performance and profiling has to be done early!

cheers :)



-- 
Ernesto, Nov. 9th 2010:
"In the graphics generation routines of TYPO3 *anything* could cause a
side effect."




More information about the TYPO3-dev mailing list