I recently started writing Python code. And I love it.
If you’d ask me a week ago about Python, I’d say “meh.” Python3K? I wouldn’t be happy about it. Sunday evening? I love both.
One project, rewritten
I wrote a new project. Or rather re-wrote a Perl “project”.
The project, once a little help for building AUR packages, now is almost a fully-featured AUR helper (it has no update function, but I will write one soon). It’s the PKGBUILDer.
The Perl version (search in the repo, linked above) had 56 lines. In short, it did something like this (rewritten to bash):
This code is really, REALLY bad. But it worked for me, because the “normal” AUR helpers were slow. I wanted to do something about it. I put an entry on my TODO list about it. A few months later I decided to do it. The TODO list entry said “write build.py”. I wanted to use Python because I wanted to learn it. In fact, I began loving Python.
The Perl version had 56 lines. A shortened version of it in Bash took only 10 (I skipped a few features, the full version would be around 20 lines or so.) Take a guess: how long is the Python version? 30 lines? 100? No. 300 lines. How could this happen? No, not because Python is a pain in the ass to write. It was because I could implement new, great features EASILY. The original version could only download a package and build it. What if the package didn’t exist? The library responsible for untarring it would throw an error. And even if makepkg had a problem with building the package, the script would happily inform the user that it was successfully bulit… What are the new features, you may ask? Install validation, i.e. checking if the package is installed or not. Package searching, sanity checks, dependency solving… This is great. If I’d like to write it in Perl, it will take me ages and I’m not sure if there is any libalpm wrapper.
If you think that you can rewrite it in Perl, sure, go for it, if you will:
find a working libalpm wrapper or write one yourself
port python3-aur (it heps with the XML-RPC of the AUR) to perl
implement EVERY feature of the Py3K version
give me the code and tell me how long did you write it
Done? Great then, contact me!
Python has the friendliest web documentation ever. PHP’s looks a bit harsh. Perl’s is not easy to search.
Time for a real world example: I want to learn how to write a specific function in Perl, PHP and Python. This function would print the argument. For example, in C:
Notice: by “searching” in docs I mean reading the page and looking for a thing.
Perl: Let’s begin at <http://perl.org>. Documentation tab, Tutorials. I need to define a function. Nothing seems to help me. I look at the sidebar and find Reference/Functions. Great, that’s what I need, so i click it… I can’t see anything about functions. Langauge reference? Nothing. I ask Google and I learn that Perl names them subroutines. I check the Language reference: it’s the sub function, now I can define my function and call it. printf? Let’s look it up in the Functions list. We’re done.
Perl’s documentation is anywhere near user-friendliness.
PHP: <http://php.net>. Why is the documentation link so small? Anyways, I need functions. Language Refernce/Functions. Here we go, one more click and I know how to make a function. And I guess that I’ll have to search the Function Reference. I find text processing, go for Strings and I can happily see printf. Take a look and we can write this:
Python: <http://python.org/>. Documentation element exists in the menu. I click it. They offer me a nice tutorial, so I’ll check it out. I scan through the Table of Contents and I see a chapter called Defining Functions. Great, it will work. Now I go back to the ToC and, because this is a tutorial rather than a reference, and I can see chapter 7.1: Fancier output formatting. I want to have %s as in other languages, so I skip this one and see Old string formatting, which uses the %s. Now, assuming they indented the code on purpose because there are no braces, I can write:
All of them work and output
Input: some stuff to print followed by a
newline. The original C example had 8 lines. Perl made it in 5, PHP
in 6 (or 4 if you won’t count the PHP tags), Python used only 3.
Which documentation is the most HUMAN-friendly? Python’s. Which is the worst? Perl’s.
Nothing is flawless
Everything has some flaws. What is it in Python, then? Existence of two concurrent versions.
Most distros and projects use Py2K, while some of them offer Py3K (or both.)
The PKGBUILDer is in Py3K, because it requires
pyalpm and the
module (I could rewrite the AUR module in Py2K, but pyalpm is much harder to
modify. UPDATE 2012-08-04: it was re-implemented by me several releases ago,
in version 2.1.0, released over a month after this post) My other projects
(like KWDv2, another rewrite, this time with minimal changes and 30% less code
or my first ever Python project, trash.py, a partial XDG trash standard
implementation) use the old Py2K (usually v2.6, because I need compatibility
with my shell server.)
I would rewrite this blog into Django if I’d own a VPS or a dedicated server. (update 2011-10-20: this blog is now based on hyde, jekyll’s evil twin in Python. Update 2013-02-08: Now using Nikola, even better engine, yet still in Python.)