On 01/02/2017 04:25, Jonathan Hartley wrote: > Thanks all. > > Hansel - Thank you, that makes sense. I actually already do a mini > version of that at the place I'm leaving, but devs are only using > Linux/Mac host machines, and we only a Linux VM. It's reassuring to hear > that sort of setup is still feasible when extended to Windows hosts, and > Windows VMs too. > > I'm tempted to reduce the number of dev configurations we need to > maintain by just holding my nose and using a Windows laptop, same as > everyone else, with a Linux VM on it. That way, I can easily replicate > my setup on any other dev's machine if they want to get involved in the > Python. Would get us up and running quicker, rather than figuring out > every combo of host and VM OS. But maybe expand into doing the full > monty you describe if there's ever more than just me who would like to > work from Linux (or if I get sick of working in a VM the whole time) > > Jonathan > >
Jonathan, Have you considered that python is extremely cross platform. Provided you are careful in the areas of: - Paths - Direct hardware interfaces - Shell operations - Maintaining your requires.txt file(s) - A tiny number of libraries that are not (easily) available for Windows, (maybe it is worth having a blacklist). On that front Christoph Gohlke has http://www.lfd.uci.edu/~gohlke/pythonlibs/ which can save many hours of fun. Note that the introduction of the wheel library format has made things a lot simpler. You should be able to develop & test python code on Linux (where you are comfortable) including writing test suites - fire up a Windows VM with python installed, (complete with populating you library from requires.txt) & run your test suite. Then, assuming the tests all pass, either deploy to Windows machines with python already installed as a prerequisite or while you are in your VM use py2exe/pyinstaller or cxfreeze to produce a Windows deployment for machines that don't have python installed. If you stick to good, cross platform, programming standards you should find it a breeze - pylint is your friend on this. Just as an example I had been developing a suite of (wx)Python based programs/utilities at work for quite a few years under Windows 7 and deploying it. as executables, to machines running XP though 10. When the whole lot was at about 80k lines I had an opportunity to try it for the first time on a Linux box - I had to change 3 lines of code to get it working. Of course this was helped by the fact that I was testing both as source and as bundled executable. Many people swear by Anaconda for simplifying this sort of stuff you may wish to consider going that way personally I stick with pip and sometimes, rarely, have to re-think which library I am going to have to use. Of course deploying a docker using vagrant does give you a lot more control and Linux dockers with python & your libraries installed can be quite light weight. -- Steve (Gadget) Barnes Any opinions in this message are my personal opinions and do not reflect those of my employer. _______________________________________________ python-uk mailing list python-uk@python.org https://mail.python.org/mailman/listinfo/python-uk
