NoName
Posts: 2729 Joined: Mar. 2013
|
Quote (Wesley R. Elsberry @ May 02 2016,08:44) | Quote (NoName @ May 02 2016,06:50) | Quote (Wesley R. Elsberry @ May 02 2016,07:40) | NoName, that's an interesting comment about the Anaconda distribution of Python. I consider my life to have been made better by using Anaconda. What route is giving you a better, solid Python programming basis? I'd like to give it a try if it is FOSS. |
It was test install for a project that never came to be -- the Mac OS X Anaconda install completely screwed up my dev environment (largely Ruby and RubyMotion but with a variety of other tools and languages). It was a bugger to uninstall (although that's true for OS X regardless), and never worked 'as described'. I still have artifacts lingering scattered across my system; I'll put some time and effort into clearing them on 'real soon now'. If the project had materialized, I'd have looked at building a Docker container to run it in. I've got Parallels, but it so badly bogs down the performance of my machine that I hate using it. It's been at least a year since I've launched my Windows install. I don't miss it ;-) On the other hand, plain old ordinary Python from the standard distribution has been trouble-free from 2.3 on. I don't use it often, having developed a strong preference for Ruby, but I try to stay within a rev or two of 'latest release'. Your mileage will likely vary, especially if you're not running on OS X. |
I recently got a SkyTree ML Express download. It comes as a Virtual Box box. This may be the future of development environments: containerized deployments that avoid unpleasant interactions.
I did install Anaconda on my Macbook Pro, but since that was about the first thing I did after getting it upgraded to Yosemite, I didn't run into the sort of bad behavior that you did. My work machine runs Windows 7 and most of my deployments are to Linux, and Anaconda just works on those two platforms so far as I can tell. I can develop on Windows and move to Linux without a hitch, usually. I'm using Python 2.7 to have the most package compatibility. That's in contrast to building up an environment from base Python, which on Windows typically had me scrounging Gohlke's site for packages that worked there, and having to harmonize everything with what was available for Linux. Anaconda's packages work directly and ease the cross-platform dev cycle for me.
It sounds like you have moved on from Python, but you probably could set up another user account and install Anaconda in user space with the new account without endangering your build environment again, though I admit I'd be very leery of anything that actually messed up stuff as it installed. That is pretty egregious. |
If I ever have to get back into Python, I plan on going with a container or installing on Parallels/Bootcamp. Sooner or later I'm going to be upgrading hardware and that should help performance issues.
I've done a bit of work building Docker containers for clients, and liked the experience. It seems a very good way to go, standardizing development environments, making DevOps easier on the way to deployment, deployment on AWS or Digital Ocean or inside the corporate firewall, etc.
What little coding I do these days, though, is in Ruby, RubyMotion, Hanami, Sinatra/Padrino (or Rails if I must) and so on. It's a congenial fit for me. But if I were doing any kind of significant or meaningful science coding, I'd probably be back on Python in a flash. C# is about the only alternative I could consider, and that would require a paying client ;-) But unlike Gary, I have a clear appreciation for when, where, and how modeling fits into the broader arena of science research*. I can keep very busy (and useful and productive) without ever needing to build a software model.
*Of course, to be fair, Gary has to look at the arena(s) of science with a high-power telescope. It appears to have a number of significant optical defects, mapped isomorphically onto Gary's delusions. Or, more likely, vice versa.
|