The production version of Python 3 has been available for about two and a half years. Since it breaks backwards compatibility with the Python 2.x series there has been a lot of mixed reactions to it. To get a developer’s perspective on Python 3 we decided to interview Virgil Dupras.
InfoQ: Can you tell us a little about yourself and how you use Python in your job?
Virgil Dupras: I develop open source cross platform desktop apps with PyObjC and PyQt and I publish them on my website at http://www.hardcoded.net/. I used to work with Delphi, but I switched to Python in 2005 and never looked back.
InfoQ: What was the major reason you switched to Python 3 a year ago?
Virgil Dupras: I love Python and I always wanted to contribute back to it. I'd love to be more involved with core development, but I think I'm just not smart enough. Sure, I can submit patches (and I did submit a few) but the fact of the matter is I seldom encounter bugs in Python, so I seldom have a patch to submit.
So, with all that in mind, I figured that the best way for me to help Python was to be an early Python 3 adopter, and I think I do help. I didn't encounter bugs in Python itself, but I encountered plenty of them in libraries I use. I also did a fair amount of py3k porting for libraries.
InfoQ: Did any of the breaking changes that caused you significant problems when you migrated your code?
Virgil Dupras: The only big problem you'll encounter with Python 3 migration is the str/bytes thing. Don't worry about anything else, it is all easily ported by a adding all appropriate __future__ imports before porting. Well, ok, maybe divisions can be a problem, but cmon! weren't you using "from __future__ import division" already?
But the str/bytes thing, it's real tricky. You really have to think about how you use your strings before you port. One nasty gotcha: the "re" library only works with strings, not bytes. So if you were using regular expressions to parse "semi-binary" data, you're in trouble.
But unless you're in one of those "gotcha" situations, taking care of str/bytes is relatively straightforward: Take care of your literals, and then review all your "open()" calls and see if they should be opened in binary or text mode.
InfoQ: Are they Python 2 features that you like that didn’t make it into Python 3?
Virgil Dupras: No, it's the same language, with less kinks.
For more on cross platform development UI development see Virgil’s article Cross-Toolkit Software.
Community comments
I was wrong about bytes and regexp
by Virgil Dupras,
I was wrong about bytes and regexp
by Virgil Dupras,
Your message is awaiting moderation. Thank you for participating in the discussion.
I'm a bit embarrassed because it's not true that you can't do regexp on bytes. I mentioned it because that was my problem of the moment and I didn't do proper research first.
You can perform regexp operations on bytes, but you have to use byte patterns. That also cause its own share of problems because many regexp-based libraries assume strings for patterns.