monadic / selective names lookup for parsing & pretty-printing

[aryairani]

As codebases are growing, it's increasingly intractable to pre-load all names, even with history excluded. (This applies to both UCM and Share.)

We want to just load the names that are needed for parsing (for specific lexed identifiers) or rendering (for specific dependencies).

[pchiusano]

I thought Share already did something fancy for this... no?

[aryairani]

Err, right, good point. Share does something fancy for this (not positive if it covers all use cases, though it may), but UCM doesn't yet.

[mitchellwrosen]

Share does load all names in order to construct the pretty-print environment.

[aryairani]

@ChrisPenner This is already done right?

[ChrisPenner]

@aryairani only on Share, we still load all the whole codebase root in ucm. Project roots will help, but ucm doesn't do anything sophisticated like Share does

[aryairani]

Gotcha, thanks.