I don’t see us changing to py3 at work any time soon. We have a large set of “rest-ish” internal and external services all running py2. We are faced with a huge amount of work to upgrade to py3 and the benefit for us is… dubious.
py3 was slower for quite a while (not sure if it still is).
we are not fans of twisted style concurrency (which trollius/tulip was based on) preferring greenlet coroutine style, so that isn’t an especially compelling feature.
So instead we are looking at sticking with the py2 series (and pypy) in the near term, and possibly rewriting some things in Go in the long term. If one of the py2 forks ever ends up taking off, we may switch to that.
We are faced with a huge amount of work to upgrade to py3 and the benefit for us is… dubious.
Same here. I’ve done the analysis and I can’t see why I would migrate. Better Unicode? It already works. Async? That style of coding isn’t pleasant (and the code isn’t async now, so migration + rewrite?).
I suspect a lot of code bases are in this situation.
It’s a long story. My mail client is written in ruby. Real world mail is frequently incorrectly encoded, which causes ruby 1.9+ to throw panics left and right. Technically all the code runs, and it works for 99% of mail, but I can’t get by parsing only 99% of my mail. Spam is the most likely culprit, but even Amazon will frequently send me messages where the message header says Encoding: utf8 but the message body has an html meta encoding=latin1 tag. Ruby does not like “both” strings.
Heh. You definitely don’t want this. I’ve slowly been working on a replacement since it’s clearly a deadend. I may, ironically, publish source at that point because I think the warning “I used this. Then I stopped.” carries a little more weight than the typical beta warning.
Cool data! Python 3 feels so much more polished once you put it to work. But I think asyncio will be the killer feature that gets people to switch in time. I gave a talk on it here.
Still not the only holdout! I think I may be the last inhabitant of ruby 1.8.7 isle, but Python 2.7 isn’t quite so lonely.
I don’t see us changing to py3 at work any time soon. We have a large set of “rest-ish” internal and external services all running py2. We are faced with a huge amount of work to upgrade to py3 and the benefit for us is… dubious.
So instead we are looking at sticking with the py2 series (and pypy) in the near term, and possibly rewriting some things in Go in the long term. If one of the py2 forks ever ends up taking off, we may switch to that.
Same here. I’ve done the analysis and I can’t see why I would migrate. Better Unicode? It already works. Async? That style of coding isn’t pleasant (and the code isn’t async now, so migration + rewrite?).
I suspect a lot of code bases are in this situation.
At my old job, they said they’d re-write in Java before moving to Python 3.
That sounds like it’s motivated by something else?
It’s more like “If we’re going to have to put in this much effort, I want way more advantage than I’m gonna get with Python 3.”
Why stay on rb 1.8.7?
It’s a long story. My mail client is written in ruby. Real world mail is frequently incorrectly encoded, which causes ruby 1.9+ to throw panics left and right. Technically all the code runs, and it works for 99% of mail, but I can’t get by parsing only 99% of my mail. Spam is the most likely culprit, but even Amazon will frequently send me messages where the message header says Encoding: utf8 but the message body has an html meta encoding=latin1 tag. Ruby does not like “both” strings.
Ah interesting. Is your mail client up on Github or a similar site? I’m curious to see what it looks like.
Heh. You definitely don’t want this. I’ve slowly been working on a replacement since it’s clearly a deadend. I may, ironically, publish source at that point because I think the warning “I used this. Then I stopped.” carries a little more weight than the typical beta warning.
Cool data! Python 3 feels so much more polished once you put it to work. But I think asyncio will be the killer feature that gets people to switch in time. I gave a talk on it here.
I’ve been using it for all new code for a while (3.0.x or 3.1.x, maybe) and haven’t looked back. Life is a lot better with reasonable UTF-8 support.