does anyone successfully use pants keeping multipl...
# general
does anyone successfully use pants keeping multiple versions of 3rdparty python libraries around? it seemed to kinda work, but when our ci script ran
./pants test lint ::
(roughly) it tried to import both versions of the library and failed.
i believe there are two answers to your question, one involving the python pipeline porting effort to the new engine which helps to isolate python code (in progress:, and as for incrementally allowing the action you describe, something like
I don’t think the second PR is quite the same thing. I think it’s more targeted around having one version of a tool like MyPy or Pylint, and only running that took over a certain set of files; as opposed to allowing you to install two versions of that tool
👍 2
This is a general Pants deficiency not related directly to python. In general, :: will fail for many pants goals against python targets, jvm targets, etc. See here for a way to do 'Continuous Integration': We're working on supporting :: generally, but it's a ways off still.
👍 2
interesting. just out of curiosity, is there a summary on the issues with a top-level
somewhere? I guess this applies to any glob that touches “sufficiently many” targets to expose whatever issues are being run into?
The summary is just that most tasks were written to batch back in the day. Among these, dependency resolves. It's mainly (all I think) surrounding fixing resolves for repos that have no global resolve solution - most don't.
For historical clarity, the repo pants grew up in held a single global resolve solution for ~2 years before split Hadoop deployment cluster versions crept into reality and 1st broke global resolves on the jvm side.