[Zope] Help in deciding approach to Web App

[Fernando Martins]

On 12/04/2011 09:52 AM, Sareesh Sudhakaran wrote:
> Thanks Niels. Just to clarify:
> Does my particular instance fall under an OODBMS model or a RDBMS 
> model (with ORM if necessary)?
>
>
Data modelling is a bit of an art and probably you could tackle your 
problem with any approach. I think the important is for you to figure 
out which model suits more your personality. No kidding.

I would personally start with the RDBMS approach considering only the 
information you provide. Also, you can easily use zope in combination 
with a RDBMS. When you read the book, beware that zope has been changing 
from a "through the web" approach, to a typical file system based 
approach, which is a loss, but it seems to be what suits the needs of 
the zope developers.

The approach I use is:

zpt page -> (one) Python Script -> (some) ZSQL Methods -> MySQL

The zpt defines the layout of the pages, the ZSQL Methods retrieve the 
data needed and the Python script massages the data to make it suitable 
for the ZPT page. Or the other way around, from user input in a form to 
storage in the database.

The advantage of the relational approach is that it is a very well 
understood model and although different people will still come to 
different models there are sufficient objective guidelines out there 
(e.g., the normalization rules, and then criteria for when to 
denormalise) to help you along. Furthermore, there are lots of people in 
db-related forums that can help you.

Also, RDBMS provides you with a "standard" query language, SQL, which 
plenty of systems and tools can use. In general, RDBMS gives you the 
safest approach to keep your data and not loose it when you need to 
migrate either the front-end or the back-end. This language is very 
powerful and can avoid you a lot of low level programming.

However, plenty of people can not deal well with SQL because it follows 
a paradigm so different from the classic imperative programming. With 
SQL, you specify the "what", with the other languages you specify the 
"how". The advantage of the "what" is that you can do a lot of data 
processing in a few lines of code. The problem with the "what" is that 
because you don't know the "how" of it, you feel you don't have control 
and you are led to say the language is "obscure" or "unreadable".

However, even if you are not comfortable with the "what" (you have to 
try to know), you can still rely on an library like SQLAlchemy to keep 
you a bit in the comfort zone of the "how". So instead of learning SQL, 
you need to learn the API of a specific library. Your choice. I 
recommend the first by far.

The real main issue with Relational is that it is a highly structured 
data model. It allows you to keep high quality data but if you don't get 
it right soon enough in the development cycle, some later changes can 
have a huge impact in the application, requiring rewrites. Furthermore, 
it works the best when you have many objects with the same properties. 
If you have many entities all different from each other (the tools in 
your case, maybe), then maybe a OODBMS might be better. But here, there 
is no standard language, or standard whatever. Perosnally, I would avoid 
as much as possible to put data in something like ZODB (I use it merely 
to store and manage the application).

The problem with your specific case is that it does not seem to be a 
typical case of books and authors, which might be a risk for someone 
without experience. The issue "Tool A might have only three fixed rpms - 
100, 200 and 500, but Tool B might have rpms from 20 to 2000", is indeed 
tricky. I suspect in general the needs of your system will be too 
specific to be able to rely only on SQL queries. You would need to put a 
lot of intelligence in the data (really highly structured) and it might 
become unmanageable or not scalable.

I guess you will need to put a lot of intelligence in the Python Script. 
So, the ZSQL retrieves the relevant tool data and then makes the tool 
choice. The knowledge of the meaning of the attributes is maintained in 
programming.

I should say I am not totally sure the Python Script is the best 
approach here, in terms of management facilities. But Python is surely a 
very good language due to its readability. However, you might need to 
use External methods or a more typical file-system based Python approach.

Or maybe you actually need to create a Domain Specific Language to 
encode the intelligence needed for your tool selection process. If your 
python code becomes repetitive, with patterns showing up, then a DSL 
might be a good approach, but this might be too much engineering for you 
at this stage. It looks like you are in a typical CIM scenario and I 
remember handling a similar problem 20 years ago. I vaguely remember at 
that time to use custom graph structures in C and the the intelligence 
was coded in Lisp/Scheme. So, there is a big range of solutions to try 
out :)

If you have time, then the simple approach

zpt page -> (one) Python Script -> (some) ZSQL Methods -> MySQL database

might be a good starting point. You should not need to spend much time 
to implement a prototype using this approach. In the worse case scenario 
it helps you understand better your problem and what could be a better 
approach with little investment. Essentially you try the classical 
approach and if it does not work well, either you are doing something 
wrong, or you have a better understanding of your needs and where to go.

Good luck,
Fernando

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.zope.org/pipermail/zope/attachments/20111204/0ab00305/attachment.html>