perf: refactor wizard to avoid expensive imports at CLI startup

[joanise]

PR Goal?

On the login node on GPSC-C, the CLI can take 4 or 5 seconds to respond to a <tab> because it's loading a lot of stuff. The biggest culprits are typer, rich, and other things required to actually provide CLI help and autocompletion, but I noticed the wizard is doing some imports we can defer too, so I refactored it.

This results in a 15% or so speed up of everyvoice -h once already loaded from cache, and I'm not sure how much on a first try when not in disk cache yet (that's harder to measure!). Not the big difference I was hoping for, but still worth committing and pushing.

Plus, the refactored code is a little cleaner than the original imho.

Feedback sought?

Regular review

The diff will look huge, but it's not: stuff removed from wizard/__init__.py has been moved verbatim to wizard/tour.py.

Priority?

low

Tests added?

was already covered, adjusted where needed

How to test?

run everyvoice test cli and maybe everyvoice new-project and see no failures.

Confidence?

high: this is just moving definitions from one file to another, using VSCode's refactoring tools and then testing and adjusting the few little bits it didn't handle.

Note: I thought of having the Step classes in step.py and the Tour class in tour.py, but doing so created circular imports since each uses the other for typing hints, and therefore I decided it made most sense to keep them all together in tour.py.

Version change?

no

Related PRs?

[semanticdiff-com]

Review changes with  

Changed Files

File	Status
everyvoice/wizard/dataset.py	12% smaller
everyvoice/tests/test_wizard.py	1% smaller
everyvoice/tests/stubs.py	0% smaller
everyvoice/tests/test_wizard_helpers.py	0% smaller
everyvoice/wizard/__init__.py	0% smaller
everyvoice/wizard/basic.py	0% smaller
everyvoice/wizard/main_tour.py	0% smaller
everyvoice/wizard/tour.py	0% smaller

[github-actions]

CLI load time: 0:00.26
Pull Request HEAD: 65418307b7e88ec269384ed0cb620a67ed53e7ff
Imports that take more than 0.1 s:
import time: self [us] | cumulative | imported package
import time:      1027 |     102326 |     typer.main
import time:       289 |     121148 |   typer
import time:      7655 |     198155 | everyvoice.cli

[codecov]

Codecov Report

Attention: Patch coverage is 92.85714% with 17 lines in your changes missing coverage. Please review.

Project coverage is 76.77%. Comparing base (fc0dc20) to head (6541830).
Report is 4 commits behind head on main.

Files with missing lines	Patch %	Lines
everyvoice/wizard/tour.py	92.70%	11 Missing and 6 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #615      +/-   ##
==========================================
+ Coverage   76.74%   76.77%   +0.02%     
==========================================
  Files          46       47       +1     
  Lines        3445     3449       +4     
  Branches      470      470              
==========================================
+ Hits         2644     2648       +4     
  Misses        700      700              
  Partials      101      101              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[roedoejet]

Nice - thanks Eric! I agree it makes sense to move stuff out of the init module. Moving things to tour.py makes sense to me. This PR looks good, and does seem to speed things up. Thanks!