[EVALUATION] - E01: The Java Failure - May Python Helps?

[Jeremy Bowers]

On Thu, 03 Feb 2005 09:26:08 +0200, Ilias Lazaridis wrote:
> My question is essentially:
> 
> How many of those constructs are already supported by python (and the 
> surrounding open-source-projects):
> 
> http://lazaridis.com/case/stack/index.html

This post is hard to follow, but I'm going to assume this is the core
question, as it is labelled as such.

The first thing that leaps to mind is that you need to play with Python
for a bit to get a full analysis of it. Due to the nature of Python, some
of the things you have in that list don't really apply. 

The most obvious example of this is "code generation": Assuming you mean
what I think you mean, the sort of thing you have in the C++ or Java world
where you click a couple of boxes, push "next" a few times, and have
several hundred kilobytes of impenetrable source code pop out as your
project's "framework", it doesn't generally apply. In Python, you can
actually write frameworks that can be made to do useful things in a small
handful of lines.

For a good example of that, I'd recommend the Twisted tutorial:
http://twistedmatrix.com/documents/current/howto/tutorial/index . In
Python, if a framework makes you write reams of boilerplate... it's
probably because someone just came from Java and isn't thinking in Python
yet. :-) Code generation in Python is generally a non-starter. (Few rare
counterexamples, but it's nothing like the Necessary Crutch it is in the
C++ world.) Generally, every line in that Twisted tutorial, even in the
final program, is *doing something*, not satisfying the framework with
necessary boilerplate.

As for the rest of your requirements, you have specified your "technology
stack" in terms of *goals*, not capabilities, so for quite a lot of them,
there is no answer except, "Yes, Python can do that, and generally easier
than most anything else". For instance, "Use of metadata within the design
(on text level [code file])" can mean a thousand things. For what it's
worth, Python tends to make it so easy I do it all the time, but for any
given way you mean it, I can't promise there exists a pre-rolled framework
for you.

So I can only speak generally. Given your list, you may find that Python
is weak in the "graphical programming" department; drop-and-drop GUI
generation isn't as well-tuned as VB. (I, for one, consider that
development methodology toxic and actively dangerous, but I can understand
why you miss it.)

Skipping down to your evaluation sequence:

* Create a class: Well, I'll show you this one:

class Something: pass

There, a class.

* Simple GUI: You may wish to check out Boa Constructor, as the closest
thing to the GUI generator you probably want. There are at least 3 major
viable GUI toolkits for Python and several other minor (but still capable)
ones.

* Embedded DBs: I don't know, Google for your embedded DB name + Python.
Failing that, there are several ways to wrap your embedded DB such that a
Python program can use it.

* Web GUI: There are more Python web frameworks than you can shake a stick
at, and I don't mean "some guys hacked together templating system" either;
there are a lot of very mature systems out there, expressing a lot of
different philosophies. Given some of your other requirements, for a
web-based application I'd recommend examining Zope.

* Deployment: I don't generally have enough problems with this to be worth
thinking about. I don't know what the state of the remote debugging is on
Python; Google "remote debugging Python".

* For your "complex updates", I see two distinct things there; half of the
depend on the database, not Python. For the other half, it depends on if
you mean "while the program is still running". If so, they are
challenging. If not, they are trivial.

* I'd recommend adding "testability" to your stack, and refer you to the
"unittest" module; others can suggest their preferred testing modules, but
I tend to stick with that as the general framework is available in a lot
of languages that I use. If by evolutive development you are including the
ideas behind "agile software development" (hard to tell if you mean that,
or a community open to change*)

Based on what I see here, I have two basic comments. First, yes, Python is
probably as close as you are going to get in an existing system to what I
think you are looking for. It isn't perfect, but you will also be able to
reasonably extend it for your needs. Second, Python has a different
philosophy; you recognize the need for a different philosophy and I
extremely strongly suggest that you take the time with Python to sort out
what you truly need from what you think you need due to still thinking
with your old philosophy. (For example, given my aforementioned
reasons why code generation isn't an issue in Python, I strongly suggest
you embrace Python and give it a chance, rather than force 'code
generation' on top of it.)

You've got some exciting learning to do, and some interesting philosophy
shifts to absorb, but I think you'll find, as a lot of use have, that even
if it isn't perfect it's a helluva lot better than Java!

*: Community open to change: One caveat, from what I've seen when other
people talk about this. The Python community is open to change, but it is
still a meritocracy, and you will still need to convince others your
change is good. For example, getting something into the standard library
is not something that everybody does routinely. But as it is open source,
if you want to take responsibility for your changes, you can do as you
like. I *think*, based on your reading, that you are focussing on the
latter, in which case you are OK. But do not expect to be able to dump
things off onto "the Python community". 

Also, not getting into the standard library isn't that big a deal anyhow;
a lot of very powerful and popular frameworks and modules don't go in,
often because they conflict with the needs of a Python release (slow rate
of change, mature interface, general utility, need to prevent a Python
download ballooning to 250 MB, etc.).