Add support for local cache in the CLI

[remyleone]

As a CLI user, I want to autocomplete on the UUIDs, because I'm lazy to write/paste a UUID.

Expected behavior:

$ scw in
*tab tab*
$ scw instance
$ scw instance g
*tab tab*
$ scw instance get
$ scw instance get ser
*tab tab*
$ scw instance get server
*tab tab*
$ scw instance get server server-id=
$ scw instance get server server-id=a
*tab tab*
$ scw instance get server server-id=a863f248-dd4e-4445-ba66-afacbdcfac8d
*run*

Idea:
Having aAutoCompleteFunc and a ResolverFunc.

When trying to auto-complete a UUID field we should suggest a UUID or complete if only one match. In order to achieve this quickly, we should list the server IDs once and cache it somewhere. This catch must be updated on any create, update, list, or delete or after a cache TTL interval.

It also needs to not impact too much reactivity on auto-complete (avoiding unnecessary disk access on every completion).

This feature also needs to take into account the profile specified with the -p flag. We will get different resources with different profiles.

[jerome-quere]

We should add support for -p flag as this will affect autocompleted values

[jerome-quere]

Implementation Design Proposal 1:

• Settings
  • The cache is scoped per profile
  • It can be enabled and disabled on a profile basis in the configuration file
• Storage
  • A sqlite database seems to be a good fit because we want to people to isolate resources per namespace and per region/zone. Example: an image would have different id in different zone for the same name. We need to have flexibility and sqlite could offer it through SQL queries. feat(autocomplete): add commands to manually refresh cache #1318
  • Using an ORM such as gorm ensures that we got type structure to get data out of this database.
• Cache feeding
  • Opportunistic: We could intercept command results and automatically store in the cache all known resource types. Types could be manually registered in a WellKnownMap in cache package
    For each operation such as GetX, UpdateX, CreateX, DeleteX the call to this function would be intercepted and a consistency check with the cache will be run. At the moment having a CacheFunc field in each command that specifies how this particular command interact with the cache in case of success.
  • Complete manual refresh of the cache would also be available feat(autocomplete): add commands to manually refresh cache #1318
  • Do we need to generate code for this? Or is it manual?
• Autocompletion
  • Use autocomplete func on argSpec
  • When generating code how do we know the type of resource we are looking for? We need context for an arg spec in the generator to understand that ServerId string we should look for servers ids
• Resolver
  • The concept of ArgSpect resolver is a function that will be executed before arg unmarshalling and allow swapping the user-provided string for an argSpec by another. It would allow swapping a name by a UUID allowing the command to only support UUID only. This resolver would also support partial UUID in addition of names for completion.
• Help section
  • scw help cache should describe how the cache works, how to disable it, what customer data will be persisted on disk, etc...