Yet another GUI toolkit question...

[Kevin Walzer]

....with a twist.

I'm undertaking my first semi-substantial Python GUI application after a
long time dabbling with the language.

I'm fairly experienced with Tcl/Tk, so Tkinter seems the obvious choice
to reduce my Python learning curve. However, my Tcl applications
typically make use of a *lot* of extensions to the core widget set, such
as BWidgets, tablelist, and Tile.

I've found minimal Tkinter wrappers for all of these extensions, but
they are all labelled "experiment al," lightly documented at best, and
little-used (as far as I can tell) by Python developers.

So: my question is, would it be more productive for me to wrestle with
these extensions when there doesn't seem to be much constituency for
their use, or would it be better/faster/more efficient for me to start
scaling the wxPython mountain?

For what it's worth, my application has to run on OS X and Windows, and
will be commercial, so this combination of requirements rules out
PyGTK/PyQt/just about every other cross-platform GUI toolkit.

Advice, especially from those Tkinter devs who look beyond the core
widget set for their apps, is appreciated!

--
Kevin Walzer
iReveal: File Search Tool
http://www.wordtech-software.com

[Peter Decker]

On 2/11/06, Kevin Walzer <[e-mail]> wrote:

Dabo looks interesting, but isn't it mainly for database applications?
Has any other kind of application been developed with it? Also, it seems
very Windows/Linux-centric. Is anyone using it on OS X?

The Dabo demo comes with several games written that have no database
stuff at all. According to the developers (one of whom develops on OS
X, to answer your other question), they started out creating a tool
for database apps, but the response has been so strong on the UI side
that they've spent most of their time making the dabo.ui module able
to be used by iteself.

I don't do any development work with databases, and I've used Dabo for
several internal apps. The stuff I used to write in wxPython I now
write in Dabo.

--

# p.d.

[Peter Decker]

On 2/11/06, Kevin Walzer <[e-mail]> wrote:

Also, it seems
very Windows/Linux-centric. Is anyone using it on OS X?

I almost forgot: take a look at this screencast:

http://leafe.com/screencasts/sizers2.html

It's the second part of a demonstration on using sizers in the Dabo
visual designer tool. The first part was done on Linux, and the saved
file was copied over unchanged to OS X, where the demo was completed.

--

# p.d.

[Chris Mellon]

On 2/11/06, Kevin Walzer <[e-mail]> wrote:

Chris Mellon wrote:

If you're planning on selling an application, especially to OS X
users, then Tk is absolutely out of the question. It doesn't have even
the slightest resemblence to native behavior, and lacks the polish and
flash that occasionally lets a non-native app get away with it. Given
the limits you've stated, I believe your only options are wxPython and
writting 2 guis using PyWin/PyObjC. The second option requires
knowledge of the native tool sets for the respective platforms, so if
you don't have that I can only suggest wxPython. If you're used to Tk
you will face a learning curve, and you will need to look for or write
replacements for the custom widgets you've been using.

Do these screenshots look "non-native" to you?

Yes, actually. But thats not the point, and a trivial dialog isn't
enough to really judge Tile anyway.

http://www.wordtech-software.com/ireveal-mac.png
http://www.wordtech-software.com/ireveal-windows.png
http://www.wordtech-software.com/ireveal-linux.png

This application was developed in Tcl/Tk.

I think Tkinter is lagging behind Tk itself in terms of its advances
with theming and native look and feel, especially on OS X and Win XP.
Certainly a lot of Python developers don't seem well-informed about the
work that has been done to bring Tk back into the modern age. The
wrappers I mentioned above (some at a site that is offline but should be
back online shortly, I'm told) are bleeding-edge in Tkinter terms but
reflect work that started a couple of years ago in Tk. (The extension is
called "Tile" and will be part of the Tk core when it's released at v. 8.5).

That doesn't mean wxPython is out of the question. What I'm trying to
figure out is whether I will be up and running faster with Tkinter +
Tile (leaving me mainly to improve my proficiency with Python itself and
translate Tk into a Python idiom) or whether I should learn wxPython,
with its vastly different GUI paradigm, as well as Python.

That depends almost totally on you and your skills, doesn't it? I'm an
unabashed fan of wxPython, so naturally I lean that way, but I
wouldn't deny that is has a learning curve. It's famous "C++ like API"
has seen vast improvement in the last few releases, by the way, and is
much more pythonic than a lot of people seem to believe - no more
from wxPython import *, for example. I certainly find it's API more
pythonic than Tkinters.
If Tkinter + Tile + Bwidgets + Tablelist does the job, as I think it may
(and as it does with regular Tcl/Tk), then I will probably stick with
that. And perhaps I can then contribute some documentation somewhere on
how to use the Tile widgets in a Tkinter application, complete with
screen shots and working code for others to use. Such documentation is
sorely lacking right now.

The vast range of available GUI toolkits is one thing that partially
offsets Python's many virtues as a programming language. I realize I was
stepping into a fire swamp by even broaching the question. However, the
application I'm envisioning (accessing web services from a desktop GUI)
will benefit from the abundant Python libraries/wrappers that have
already been written, which are mostly lacking (or which would require
me to roll my own) in Tcl.

I can't possibly see how an available range of GUI toolkits is a
detriment to a *language*. I can see (but don't neccesarily agree)
that is detrimental to a desktop environment or whatever, but not a
language. By that standard C is the worst language to ever exist, C++
only marginally better, and stuff like Visual FoxPro the best of all.


--
Kevin Walzer
iReveal: File Search Tool
http://www.wordtech-software.com

[Neil Hodgson]

Kevin Walzer:

Do these screenshots look "non-native" to you?

http://www.wordtech-software.com/ireveal-mac.png
Yes. I don't use a Mac much but the following are really obvious.
The column header icons don't fit within the header. The scroll bar is
squished up too thin. The icons are from Windows.
http://www.wordtech-software.com/ireveal-windows.png
Yes. The archetype here is Windows XP Explorer. Icons too large.
Icons jaggy due to lack of alpha compositing. Icon text too long but no
label for entry field. Weird discontinuity between toolbar and list.
List header too tall. List has alternating background. Status bar not
indented. No resize indicator in status bar. Line between columns.
Vertical scroll bar ends at bottom of horizontal scroll bar rather than
top. Scroll bars use dithered background rather than graduated smooth
background.
http://www.wordtech-software.com/ireveal-linux.png

Totally native. If you are still using a 10 year old desktop
environment.

Neil

[John J. Lee]

Kevin Walzer <[e-mail]> writes:
[...]

Commercial Qt is a little out of my price range.

Commercial *PyQt* (including a license for Qt for use only with PyQt)
is $400 (USD) per developer (plus an extra $300/year if you want
upgrades). That's compared to Qt license for use *with C++* varying
from $1690 to $6260. So PyQt is 4-15 times cheaper than old-fashioned
C++ Qt!

http://www.thekompany.com/products/blackadder/
Apparently you get mxODBC in that price, too. And the Blackadder
development environment itself, of course, though personally I
wouldn't use it.

PyQt 4 now seems to exist, though not as a stable release yet, so I
imagine it'll be a bit longer untill there's a release of Blackadder
that supports Qt 4. I recall the PyQt 2 --> PyQt 3 upgrade as being
fairly painless (in terms of code changes), though.
John

[Phil Thompson]

On Monday 13 February 2006 12:33 am, John J. Lee wrote:

Kevin Walzer <[e-mail]> writes:
[...]

Commercial Qt is a little out of my price range.

Commercial *PyQt* (including a license for Qt for use only with PyQt)
is $400 (USD) per developer (plus an extra $300/year if you want
upgrades). That's compared to Qt license for use *with C++* varying
from $1690 to $6260. So PyQt is 4-15 times cheaper than old-fashioned
C++ Qt!

http://www.thekompany.com/products/blackadder/
Apparently you get mxODBC in that price, too. And the Blackadder
development environment itself, of course, though personally I
wouldn't use it.

PyQt 4 now seems to exist, though not as a stable release yet, so I
imagine it'll be a bit longer untill there's a release of Blackadder
that supports Qt 4. I recall the PyQt 2 --> PyQt 3 upgrade as being
fairly painless (in terms of code changes), though.

There will never be a release of Blackadder that supports PyQt4.

Phil

[John J. Lee]

Phil Thompson <ph**@riverbank computing.co.uk > writes:

On Monday 13 February 2006 12:33 am, John J. Lee wrote:

Kevin Walzer <[e-mail]> writes:
[...]

Commercial Qt is a little out of my price range.

Commercial *PyQt* (including a license for Qt for use only with PyQt)
is $400 (USD) per developer (plus an extra $300/year if you want
upgrades). That's compared to Qt license for use *with C++* varying
from $1690 to $6260. So PyQt is 4-15 times cheaper than old-fashioned
C++ Qt! [...] PyQt 4 now seems to exist, though not as a stable release yet, so I
imagine it'll be a bit longer untill there's a release of Blackadder
that supports Qt 4. I recall the PyQt 2 --> PyQt 3 upgrade as being
fairly painless (in terms of code changes), though.

There will never be a release of Blackadder that supports PyQt4.

Will there be a licensing deal that allows commercial use of PyQt
cheaper than the standard Qt C++ licenses?
John