Fri, 26 Sep 2003
Some random stuff:
I've also just noticed that PyPI
seems to be getting more popular - yay! If only I had some time to work on it...
Thu, 25 Sep 2003
Rachel's just cooked some chocolate cookies. The smell is extremely
distracting. Must go eat warm cookies :)
Astute viewers might notice that along with the normal MD5 checksum, the
Python 2.3.1 downloads
also have a GPG signature (the "sig" link):
The signatures above were generated with
GPG using the release manager's
(Anthony Baxter)
public key
which has a key id of 6A45C816.
And thanks to a key chain of Anthony → Me → Jim Fulton (we
signed each other's keys at the recent Sprint here - thanks Jim!)
→ Barry Warsaw → <a whole lot of people>, there's a large number of people
out there that can verify that Anthony really did generate the file they're
downloading from python.org.
Jeremy Hylton has written a
PyPI
Tutorial:
PyPI: the Python Package Index is the latest attempt to create a
comprehensive catalog of third-party Python packages. The catalog is
integrated with distutils. This tutorial explains how to use setup.py to
create PyPI entries.
Thanks, Jeremy!
Tue, 23 Sep 2003
We've just bought "Monty Python and the Holy Grail" collector's
edition DVD - this is what DVD was invented for:
- Commentaries by the cast (two of them - one with Gilliam and Jones, the
other with Cleese, Idle and Palin).
- Normal sub-titles plus special sub-titles for people who don't like the
film (taken from Shakespeare's Henry IV, Part III)
- Really special sub-titles so you can read the screenplay and watch the
film at the same time
- 'Monty Python and the Holy Grail in LEGO'
- and a bunch of documentaries, etc.
Yes, it really does have the LEGO version :)
We also got a Goodies compilation which includes episodes from before I
was born up to ones made while I was watching the show. Including Kitten Kong.
Happy :)
Last weekend was spent mostly in the garden. It's now early spring, so
the weather's warming up and the plants are going ape and are generally in
need of some TLC. Oh, and Rachel and I picked out about a couple of dozen
new trees to plant, so they had to go into the ground too.
I'm in terrible shape though, so on Saturday I planned ahead and we went
out for dinner (Shakahari, one of the best vegie restaurants
in Australia) which was yummy, as always. After a night's sleep, I was
reminded just how unfit I really am, so I spent Sunday doing
mostly small cleanup work instead of more weed removal.
Eventually, I stopped that and played with another programming
side-project which I decided needed a GUI. I've previously used PyQt for various things, but
this time I had a potential audience that couldn't be stuffed installing
PyQt. So I tried tkinter. Ugh. Never again. Ugh.
A couple of hours later, I'd used Qt Designer to create five interfaces
and I'd wired most of my existing code into those. Painless! Except now I
need to get my audience to install PyQt, which can be anything but
painless. I can't even get sip to compile on OSX, let alone trying for PyQt
itself. Ah well, it's a good thing I've got Linux too :)
Sun, 14 Sep 2003
Rachel's venerable old printer died a while back, and we've just
gotten around to replacing it. We like the colour printer / scanner / copier
combos, so
after shopping around, we found that the HP offerings
were quite nice. I did a quickie check, and found that the particular model
(PSC
2110)
we wanted to get was supported under Linux.
What support though! The
printer and scanner drivers are
maintained in sourceforge projects supported by HP. And they just
plain work out of the box!
The quality is good too - I printed one of my old photos
out on one of the sample photo paper sheets that came with the printer, and
it's pretty good. There's banding when you look closely, but it's not
noticeable at a viewing distance.
The Sydney Morning Herald has a web diary
with a number
of contributors (from around the country, not just Sydney).
I've only just noticed it, and it has some interesting
entries. Stuff like "Howard
on the ropes: Labor's three chances for a knockout blow". Going to keep
an eye on that one...
Some random links (short list this time):
Fri, 12 Sep 2003
Well, this week I demoed the new version of the
software
I'm working on to
the company as a whole, and then to some outside interested parties.
Stuart (the other developer), the lucky blighter, was scuba diving in Thailand during the lead
up (and possibly on the first demo day too). The demo went really well
though, with positive noises from all involved. That's always a nice
outcome :)
The code has a couple of utility modules that I'll release if I can find
the time. I've only been able to scrape together the most
superficial personal involvement in Roundup recently, so I'm not sure whether I
really should release any more code... The first module takes
several chunks of text and formats them into columns (plain text, of
course). At the most basic level, it does what the Python2.3 textwrap
module does, but extends that to handle the multiple columns. The other
module is a Zope interface to the GPG.py module that I snarfed from amk and extended and contributed back to the pycrypto project.
If anyone's interested in this code, let me know and I'll fire the module
your way...
And now for something completely different:
Aoccdrnig to a rscheearch at an Elingsh uinervtisy, it deosn't mttaer in
waht oredr the ltteers in a wrod are, the olny iprmoetnt tihng is taht
frist and lsat ltteer is at the rghit pclae. Therset can be a toatl mses
and you can sitll raed it wouthit porbelm. Tihs is bcuseae we do not raed
ervey lteter by it slef but the wrod asa wlohe.
Thanks, Anthony,
for that. He also pointed to some related
news.
Fri, 05 Sep 2003
More random links (not many today):
Tue, 02 Sep 2003
The following are a quick (random, off the top of various heads) list of
things that I think are
anti-pitfalls in Python. That is, because the language has these
features, it is harder to make programming mistakes.
Thanks to Dougal, Anthony and Toby for running their eyes over this
list, and for their suggestions.
- 1. Everything's a reference to an object (no magic line-noise)
-
Python doesn't have a separate scalar / reference syntax! Celebrate!
Dereferencing is one of the hardest concepts for a new programmer to learn,
and I know from personal experience that some people gave up on the whole
programming caper just because they didn't understand pointer
arithmetic and dereferencing in C.
Aside: perhaps it might be because it's such a fundamental pitfall in
other languages that people coming from them have such trouble
coping with python's "names for objects" approach. How many times have we
seen new Python programmers despairingly ask "how do I pass by reference"?
Pitfall avoided:
Many and various errors due to not getting the magic line-noise quite
right or not fully understanding the concept of dereferencing.
- 2. Real polymorphism
- (or rather, not that jumping-through-hoops, interface-tied,
type-casting-up-the-wazoo anticipate-all-use-cases polymorphism that some languages force on
users)
Everything's an object and there's no built-in type checking on arguments.
Interfaces generally evolve through consensus - if a function is written
that accepts a file object argument to read data from, then any
object that implements a read() method will do as a substitute argument - a
StringIO instance, the result of a urllib.urlopen() call, a socket, etc.
Combine with built-in variable argument lists and keyword argument lists (with
default values) for a really satisfying taste!
Pitfall avoided:
Inflexibility due to not anticipating some new use of your code.
- 3. Sequence unpacking
-
I use it so much. Mostly just automatically too when returning
multiple values from a method/function. And "a,b = b,a" does the right
thing.
Pitfall avoided:
Messy code.
- 4. The call operator
-
Bound methods, functions, classes... all use exactly
the same interface and all are first-class objects.
Pitfalls avoided:
Having to deal with (and remember) the different ways of calling different
callable things. Also, like #2 you don't have to worry about what kind
of callable you might get if you accept one as a function argument.
- 5. Automatic memory management
-
W00t!
Also under this heading could be the complete inability of user
input to mangle code and take over your system (caveat: no use of
exec/eval).
Pitfalls avoided:
Memory usage going out of control. Your system being compromised because
you didn't bounds-check some string from the Real World.
- 6. The print statement
-
A lot of other languages have a print statement or function but Python's is
particularly good at just dumping a bunch of information in human-readable
form. It just works. This is mostly because many/most builtin
objects have useful repr methods.
BTW, I think it's a good thing that it adds spaces, even with a dangling
comma. If I want fine control over output, I'll use sys.stdout.write().
Pitfall avoided:
Inability to just have a gander at what's going on.
- 7. Indentation for structure
-
Not only does this save typing, but it also means that code is far more
readable than in some other languages.
Pitfalls avoided:
Unreadable code. Strange errors due to not using block delimeters when you
should have used them.
- 8. Ubiquitous, unencumbering exception system
-
It's available and used everywhere, and you don't have to trap exceptions
if you don't want to. Trapping exceptions can be as fine-grained as you
like, and you can send an arbitrary object of information along with the
exception. Now that's nice.
- 9. Introspection
-
From Dougal: you can poke into any object. You can test whether objects
have attributes, or whether they're callable, or...
Pitfall avoided:
Like #6, you can get in there to understand what's
really going on in your own code.
- 10. Feature-full, rock-solid, simple built-in types
-
It can be argued that a lot of problems are averted just by having the
solid built-in string, tuple, list and dictionary types. The potential
problems in other languages being errors in
implementing those types or using half-assed attempts at implementing them.
Pitfall avoided:
Half-assed basic type (re-)implementations.
- 11. No namespace pollution (unless you really want to)
-
Another from Dougal: Python doesn't automatically pollute namespaces unless
you specifically force it to with statements like "from foo import *".
Pitfall avoided:
Namespace pollution, and the strange side-effects that occur.
- 12. The python shell (interactive interpreter)
-
From Toby: "if for no other reason than that i don't need a calculator
any more ;)". I definitely echo that sentiment, and more...
The interactive interpreter prompt lets new and experienced Python users
alike muck around with code to see what it does. I couldn't imagine
developing regular expressions without the interactive shell :)
Pitfall avoided:
This is more a tool for avoiding a bunch of errors you might have run into
without the ability to test ideas and code fragments out.
- 13. PDB
- From Anthony: PDB, the built-in debugger. I'm not sure this consitutes
an anti-pitfall, but it's definitely on his list. The magic string you're
looking for is "import pdb ; pdb.set_trace()"
When the code hits that you'll be dropped into the pdb shell. Nice.
Pitfall avoided:
When things do go wrong, you've got the direct ability to get
elbow-deep in the live program to see why it went wrong. And to change that
live program then and there if necessary.
14. Treat programmers with respect
Python treats programmers as intelligent people. It doesn't specifically
put barriers
in their way which can cause confusion (eg. public/private/protected
attribute distinctions that are circumventable in some other language
implementations anyway, leading to horrible, horrible code).
Pitfall avoided:
Treating programmers like babies, asking them to be overly explicit in
every little thing they do, and locking them out from things that they
might hurt themselves on.
