fix(v2): normalize location for route matching

[rlamana]

Motivation

Fixes #2392

Have you read the Contributing Guidelines on pull requests?

Yes.

Test Plan

The PendingNavigation component now uses a normalizeLocation function. I added a single file for the new function to easily add unit tests and check the proper normalization of the pathname in the location object before it is passed to react-router's Route components for matching:

Original Path	Normalized Path
/docs/introduction/index.html	/docs/introduction
/docs/introduction	/docs/introduction
/docs/introduction/foo.html	/docs/introduction/foo.html
/index.html	/
/	/

[docusaurus-bot]

Deploy preview for docusaurus-2 ready!

Built with commit 203adee

https://deploy-preview-2393--docusaurus-2.netlify.com

[rlamana]

Just to add more information about this change: since react-router-config.renderRoutes() is being used, we can use the root <Route> to pass a normalized location object to all its Route descendants and force the match for locations that have index.html.

A location with pathname=/docs/introduction/index.html is transformed into pathname=/docs/introduction and passed for matching to the generated routes.

[lex111]

LGTM 👍

[yangshun]

Why do we need the location to be part of the state now? I think we just have to normalize it before using it and it should be fine? Putting in state seems redundant.

[rlamana]

@yangshun the normalized location is also used in the preload and startProgressBar functions. Even though the call to normalize is very, very simple, it seemed like a good optimization to just call it once when the location changes, instead of calling it multiple times. Since its value affects rendering, the right place for it would be the component's state.

Let me know what you think, I can always change the PR to call normalize whenever is needed instead of using the state.

[yangshun]

@rlamana I would prefer to keep things stateless if possible if they can be derived. You could memoize the values instead of putting them inside state.

[rlamana]

@yangshun makes sense. I updated the PR to memoize the normalizeLocation function and cache the already calculated paths.

[yangshun]

Please address comment.

[yangshun]

There's no this.state.location

[rlamana]

😅 I missed a reference to the old state value... Now it's fixed.

[yangshun]

Thank you! 😄