Speed up interpreter selection (with cached & or partial lists)

[DonJayamanne]

Currently the performance selection of interpreters is left at the mercy of external tools such as conda and the like. On Windows these tools can be extremely slow.
This will only get slower as we support more sources for interpreters or if conda and other tools get slower).

Suggestion:

• Cache list of interpreters (filter to exclude invalid items, easy file check)
• Cache list of conda environments (re-query only if we have new items in environments.txt)
• Display partial list of interpreters while conda and the like run in background
• Display Loading Conda Envs in list for those not yet completed (could apply to registry searches as well)

This should improve ux significantly (at least in Windows).

[DonJayamanne]

Brett Cannon (@brettcannon) Dan Taylor (@qubitron) any suggestions.
Dan Taylor (@qubitron) you have experienced delays here.

[qubitron]

Don Jayamanne (@DonJayamanne) I like your suggestion, I think it results in the best possible user experience given the constraints. Just don't know how complex it is to implement. At a minimum, displaying "Loading..." any time we are waiting is a good idea.

This currently an issue only when selecting an interpreter soon after first loading the extension, correct?

[DonJayamanne]

This currently an issue only when selecting an interpreter soon after first loading the extension, correct?

Yes, but appliers to all instances of VS Code. I.e. if you were to open another instance of VS Code, this delay would be expected.

[brettcannon]

The "Loading ..." message seems reasonable. Caching also seems reasonable as long as we provide a way to blow the cache out manually, e.g. Select Interpreters has a "Re-check for interpreters" option or something.

[brettcannon]

And luckily this only affects selecting the interpreter, so if you already have a conda environment selected then there's no penalty there, correct?

[DonJayamanne]

And luckily this only affects selecting the interpreter, so if you already have a conda environment selected then there's no penalty there, correct?

Yes