Refine Platform.sh support

[ehmatthes]

#71 is getting crowded, so moving ongoing work to this issue. The goal of this issue is to bring support for Platform.sh to a level comparable to Heroku's current level of support. This work will bring up some issues that affect both platforms, as the overall structure of simple_deploy becomes more refined and consistent.

Anything that can be moved to a longer-term issue can be added to #90.

• [x] Finish updating documentation to reflect current Platform.sh implementation.
• [x] Move remaining items from #71 to here, and another longer-term issue; this one is for refinements before returning to a more general project focus, and revisiting other platforms.
• [x] Add a new developer doc, describing the deployment process as implemented between simple_deploy.py and deploy_platform.py.
• [ ] Get platform-specific confirmation (ie _confirm_preliminary()) in simple_deploy.py, before making any changes such as building log dir.
• [ ] Refine initial inspection, both generally and platform-specific.
• [ ] Require platform create to be called before running simple_deploy, without --automate-all.
• [ ] Make sure configuration changes are done in order of least impact to the current state of the project.
• [ ] Move adding simple deploy req to simple_deploy, now that inspection is done in a more consistent manner.
• [ ] Implement automate-all.
  • [ ] The platform create command can be run with a --quiet flag, and there are flags for everything that's prompted interactively. To see this, run platform help project:create.
  • [ ] Use Platform.sh integration test as a blueprint for implementing --automate-all.
• [ ] Write friendly summary log file.
• [ ] Support Poetry.
• [x] Document any general principles that have been defined or clarified through this work.
• [ ] Make sure Heroku deployments still work.
  • [ ] Some changes to simple_deploy.py will have impacts on deploy_heroku.py. Make sure any new errors are addressed, and all current tests for Heroku still pass.
  • [ ] Make issue to bring deploy_heroku.py to a consistent state with the current overall approach.
• [ ] Examine current state of support on Windows. Probably means documenting how to install and use CLI on Windows.
• [ ] Update existing documentation.

[ehmatthes]

Refine documentation

• [x] Finish updating platformsh_deployements.md.

[ehmatthes]

Working notes

• [ ] Consider bailing if git status does not return "nothing to commit, working tree clean".
  • Really want to encourage users to be able to easily undo configuration changes.
  • [ ] Consider a flag to ignore this, ie --force-git-unclean. Maybe simple_deploy is running a second time, and they don't want multiple commits for configuration? See #28.
  • [ ] Consider what this means if using --automate-all. We are making commits in this case, and we really don't want to have user changes mixed in with our changes when we make a commit.
• [x] Move confirmation of --automate-all to simple_deploy.py. Platform-specific script should not have to deal with this confirmation, other than being called by simple_deploy to generate a platform-specific message about what will be done to automate deployment.

[ehmatthes]

Move adding simple deploy req to simple_deploy

• This gets at a fundamental architectural question: should deploy_platform-name.py do any general work, or only platform-specific work. The latter seems much better. That will change a number of things, like whether _check_platform() is still a good name in simple_deploy.py.
• [x] Figure out an appropriate way to call _add_simple_deploy_req() from within simple_deploy.py.
  • [x] _check_platform() should create platform-specific deployer, and call any platform-specific checks and validations.
  • [x] Call add_simple_deploy_req() from simple_deploy.py, not platform-specific script.
  • [x] Move logging work so it occurs after platform-specific validation. See below.
  • [x] Refactor for confirmation (ie while confirmed.lower()...) to simple_deploy.
    • [x] Use for confirming --automate-all.
    • [x] Use for confirming preliminary status in platform_sh deployments.
  • [x] Add validate_platform() methods to platform-specific deployer classes. Can be pass() with DEV comment for now if needed.
    • [x] This should be called before confirming automate all; no need confirming that if the platform won't work.
    • [x] Check for platform_sh cli.
    • [x] Check for platformshconfig.
    • [x] Update unit testing approach to install platformshconfig, with remote pip call.
    • [x] Add platformshconfig to vendor/.
    • [x] Check for Heroku CLI.
  • [x] Use helper method naming consistently. Some things that were helpers are now being called through instances.
    • [x] simple_deploy.py
    • [x] deploy_heroku.py
    • [x] deploy_platformsh.py
  • [x] then call platform_deployer.deploy()
  • [x] make these changes for platform_sh, then for heroku as well.
• [x] Clean up _parse_cli_options(); it's doing more than parsing cli options. Most of this should probably be moved to handle().

[ehmatthes]

Start logging after all validation complete

This is part of making sure no changes are made until after all validation is complete. This also makes dev work easier, as a test project is not modified at all when intentionally running invalid commands.

• [x] Move creation of logging to after platform validation, before adding simple_deploy req.
• [x] _inspect_project() ignores logs; this needs to be moved, because it won't be possible before starting to log.
  • [x] Consider building a preliminary log entry; once we make the log, dump this text to the log as soon as it's created.
  • [x] Looks like there's nothing we really need to log in _get_dep_man_approach(). Skip logging here, and everything should work.
    • [x] Seems to be creating log dir, but no log file. Diagnose in _start_logging(). If you start logging before configuring the logger, it doesn't use the configured log file. Stopped all logging before calling start_logging().
    • [x] Probably need to log if we couldn't find a .git dir. No, this is reported in an error message and logging should not be critical for diagnosing anything related to this.
    • [x] Address these issues, then remove all diagnostics and address recent DEV notes.
  • [x] Make this part of unit testing, if not already tested. (Check that the info in the preliminary log entry is present in the log file.)
  • [x] How easily can I test that an invalid command results in no log dir or file?
• [x] Verify by running --platform platform_sh and denying confirmation; logs should not appear.
• [x] Verify by confirming, with exit() call before deploy() call. Logs should appear.
• [x] Move confirmation of --automate-all to simple_deploy, and move related messages out of _parse_cli_options().
  • [x] Needs to be run after the platform deployer object has been created. (No, just needs access to platform-specific messages.)
  • [x] Needs to display a platform-specific message.
  • [x] Revise _prep_automate_all() with the simplification that confirmation has already happened.
  • [x] Can I easily test any of this? Not now, but added note to #14.

[ehmatthes]

Closing

• [ ] Remove any remaining diagnostics.
• [ ] Are there any remaining DEV notes in simple_deploy.py that can be addressed before moving on?

[ehmatthes]

All unfinished tasks moved to #92.