4

I am finishing up a decently sized Python/Shell project, and I'm wondering if there's a 'best practice' list of things to do when finishing up development.

So far I've done:

  • pylint
  • pycallgraph
  • grep'ed for :TODO: and :FIXME:
  • find orphan code
  • some more documentation
  • install scripts for client and server
  • make sure comments match the reality (before I forget the finer points)

What other things do you do, generally, or Python-specific?

Marcin
  • 181

3 Answers3

5
  • make sure you have enough automatic tests
  • run them!
  • also test things for which you don't have automatic regression tests (for example, your install scripts)
  • and make sure you have everything checked in into your SCCS (test that by checking everything out on a clean machine and see if there is something missing)
  • make sure version numbers are up-to-date
  • update your change log (the document for your client, not the comments in your SCCS)
  • update your list of open issues / list of requirements for the next release / backlog / issue tracker (whatever you use for this purpose)

And one thing: IMHO some of the things you mentioned should be always in order at the end of the day - adding or correcting comments only before going to production is far too late. Make your comments match reality immediately, whenever you change a function, later you will forget about it.

Doc Brown
  • 218,378
1

Spend good amount of time on back-up plans to roll-back deployments in the case of failure or un-expected events.

Back-up :

  • Prepare a checklist and add items to it
  • Make sure you have a back-up plans for Database roll-back, scripts, etc.
  • Remove logging code or check the production/testing setting of the app.
  • Simulate the worse scenario of deployment
  • Check the failing points of the overall system (network connectivity, firewalls, etc..)
  • Prepare a roll-back scenario and simulate it
Yusubov
  • 21,498
0

I haven't seen this listed explicitly (though Doc Brown's answer touched on it, regarding version numbers), but IMHO the single most important part to get right in a release process is the Configuration Management. You should to be able to identify exactly what artifacts were deployed, and be able to reproduce those artifacts on demand from source.

If a customer comes to you with a bug, you'll really want to know precisely the source under which the bug was created.

Community
  • 1
Aidan Cully
  • 3,506