[Discuss] SQLAlchemy Has Annoyances

[kentborg at borg.org (Kent Borg)]

On 02/04/2015 08:00 PM, Richard Pieri wrote:
> On 2/4/2015 7:02 PM, Kent Borg wrote:
>> I was once at a Apple Developers' Conference--a *way* long time ago--and
>> Iw ent to a session on debugging. At the beginning they put a slide
>> saying that a well written program should be able to run for hours--even
>> days!--without crashing. Seemed both radical and obvious.
>
> Seems neither? I mean, I've seen old database systems with uptimes 
> measured in years, and one that was approaching a decade at the time 
> (and if it's still running then it's a few years away from a second 
> decade). A well-written program won't crash. Period.

Correct. That's my point.

So it is annoying that a library might have a black hole where one can 
throw new objects but never get rid of old ones.

> But regarding SQLAlchemy, that sounds like a Python-ism. You can't 
> delete classes in Python. 

It seems that all these libraries for doing Python to SQL databases are 
just different flavors or the same paradigm*: An "ORM" an "object 
relational model".  And that seems my disagreement: (I think) they are 
all mapping one class to one table. But classes are rather upfront 
things that are expected to be somewhat stable, yet I have stuck in my 
head the idea that tables should be able to come and go, with little 
notice, without restarting my program because resources ran out. (See 
earlier point about "well-written programs".)

* I think there is one exception: python-sql, and then I have to start 
writing Python code to write SQL statements.


-kb