Pull out wx to top level import, improve typing for core

[seanbudd]

Link to issue number:

Follow up of #13699

Summary of the issue:

NVDA core is dependent on wx in many implicit cases.

Description of how this pull request fixes the issue:

Make the wx dependency explicit.
Add additional linting and typing for work done in #13699.

Future work is needed to make core no longer dependent on wx.

Testing strategy:

Test starting NVDA with a built launcher and running from source

Known issues with pull request:

Change log entries:

No user facing changes

Code Review Checklist:

• Pull Request description:
  • description is up to date
  • change log entries
• Testing:
  • Unit tests
  • System (end to end) tests
  • Manual testing
• API is compatible with existing add-ons.
• Documentation:
  • User Documentation
  • Developer / Technical Documentation
  • Context sensitive help for GUI changes
• UX of all users considered:
  • Speech
  • Braille
  • Low Vision
  • Different web browsers
  • Localization in other languages / culture than English

[AppVeyorBot]

• PASS: Translation comments check.
• PASS: Unit tests.
• PASS: Lint check.
• FAIL: System tests. See test results for more information.
• Build (for testing PR): https://ci.appveyor.com/api/buildjobs/7rlmeag4wuvlt2pw/artifacts/output/nvda_snapshot_pr13701-25425,1bff2015.exe

See test results for failed build of commit 1bff20152c

[lukaszgo1]

How you've determined that not importing wxlazily is safe?

[seanbudd]

How you've determined that not importing wxlazily is safe?

I've run the launcher and run from source. I'd be incredibly surprised if it wasn't safe. Our main reason for deferred imports is circular dependencies.
This is an external dependency, and as such shouldn't execute code that depends on NVDA when being imported.

[seanbudd]

Another reason for deferred imports is one dependency may need another dependency initialized first. wx doesn't rely on any NVDA modules to be initialized first.

[feerrenrut]

Another thing to consider here, where some other module needs to import core before wx is imported. A reason for this may be the time it takes to import wx.

[seanbudd]

running from source, with the following code at the start of core.py

"""NVDA core"""
import time
s = time.time()
import wx
print(time.time()-s)

Time spent importing:
0.037993 (seconds)

[seanbudd]

This is an increase to the sum of the other imports: 0.00799

[LeonarddeR]

Will it also be considered to improve the code style of core, i.e. by taking the class definitions of App, CorePump and MessageWindow out of the main function?

[seanbudd]

Another thing to consider here, where some other module needs to import core before wx is imported. A reason for this may be the time it takes to import wx.

I believe imports are cached, and wx must be imported in core.main, before NVDA starts.

For the python console, this is already imported in the namespace: https://www.nvaccess.org/files/nvda/documentation/developerGuide.html#toc47

[michaelDCurran]

Way back in NVDA's infancy, we did deliberately import wx after initializing speech, so that on slow machines or when running off a USB key, we could speak a message first, as loading wx would add up to 10 seconds of time back then. But I really don't think this is relevant anymore, and no doubt we already broke this rule a long time back.

[feerrenrut]

I believe imports are cached, and wx must be imported in core.main, before NVDA starts.

When NVDA starts, multiple NVDA processes are triggered (due to the way we use Ease of Access), this is unrelated, probably unwanted behavior. However, in those cases, NVDA quite quickly determines that another NVDA process is already running, then exits.

Moving the import of wx may impact this.

For the python console, this is already imported in the namespace:

Sorry, I don't follow this.

[seanbudd]

Another approach can be started to add typing to core, without moving the wx import.