4 ms·

Python in production engineering

40 points by samber 10y ago

sandGorgon 10y ago
Millions of lines of Python code in production at Facebook. 5% of it is Python3. Interesting. I wonder if there are any interesting learnings around Unicode handling in Python 2 at this scale and use cases. I'm fairly surprised that someone like Facebook wouldn't have a Unicode monkeypatch for Python 2.
- rtpg 10y ago
  While a loooot of libraries are Python 3 compatible now, There are holdouts among python libraries with heavy C FFI elements. So if you have just one of those in your stack, it becomes really hard to use Python 3. Though honestly, I feel like Facebook could, you know, pay people to port the needed projects to Python 3.
  - rebootthesystem 10y ago
    > So if you have just one of those in your stack, it becomes really hard to use Python 3. That, exactly that, is the mess that was created when Python 3 was released as it was. When facing a new project you don't always know exactly what libraries you'll need to tap into. And so, from both a business and technical perspective, the correct decision is to go for the version that will not shaft you half way during development. And that's 2.7.x. I'm sure they folks behind Python are very smart but they have shown to not have a solid understanding of how business works. Unfortunate because the impasse it has created is almost impossible to fix.
    - xapata 10y ago
      Though some projects are still written in COBOL, sometimes we get to pick better technology. It takes time. Python has been around for more than 25 years. They can plan for the future.
- tyingq 10y ago
  Reading the article, it looks like the vast majority of the Python 2 code is around infrastructure, configuration, etc...areas where unicode support aren't terribly important. Then, in the list of Python 3 deployments, you see things that support advertising and irc, where unicode matters more. It looks like they just avoided using Python 2 anywhere near customer facing code.

4 ms·

Python in production engineering

40 points by samber 10y ago

sandGorgon 10y ago
Millions of lines of Python code in production at Facebook. 5% of it is Python3. Interesting. I wonder if there are any interesting learnings around Unicode handling in Python 2 at this scale and use cases. I'm fairly surprised that someone like Facebook wouldn't have a Unicode monkeypatch for Python 2.
- rtpg 10y ago
  While a loooot of libraries are Python 3 compatible now, There are holdouts among python libraries with heavy C FFI elements. So if you have just one of those in your stack, it becomes really hard to use Python 3. Though honestly, I feel like Facebook could, you know, pay people to port the needed projects to Python 3.
  - rebootthesystem 10y ago
    > So if you have just one of those in your stack, it becomes really hard to use Python 3. That, exactly that, is the mess that was created when Python 3 was released as it was. When facing a new project you don't always know exactly what libraries you'll need to tap into. And so, from both a business and technical perspective, the correct decision is to go for the version that will not shaft you half way during development. And that's 2.7.x. I'm sure they folks behind Python are very smart but they have shown to not have a solid understanding of how business works. Unfortunate because the impasse it has created is almost impossible to fix.
    - xapata 10y ago
      Though some projects are still written in COBOL, sometimes we get to pick better technology. It takes time. Python has been around for more than 25 years. They can plan for the future.
- tyingq 10y ago
  Reading the article, it looks like the vast majority of the Python 2 code is around infrastructure, configuration, etc...areas where unicode support aren't terribly important. Then, in the list of Python 3 deployments, you see things that support advertising and irc, where unicode matters more. It looks like they just avoided using Python 2 anywhere near customer facing code.