document const semantics

[andychu]

• We want const because POSIX shell has readonly. We don't want to lose functionality (or make people use old syntax?)
• However the whole "declaration" thing is weird for dynamic languages ...
  • especially since we don't have block scope. But should we?
• idea: const can be dynamic globally, but static locally? Is that too complex?

https://oilshell.zulipchat.com/#narrow/stream/121540-oil-discuss/topic/nested.20procs.20(Notes.20for.20Oil.200.2E8.2E6)

Considerations:

• Oil blocks, configuration. Const is the COMMON case here.

[andychu]

Related #814 -- var at the top level is inconvenient

• Should we go back to Python-like semantics?
  • we could have ONLY setvar and setglobal ? (instead of the global statement)

Or maybe local consts only? Top level should be setvar ?

[andychu]

• Idea 1: keep things as is.
  • declarations checked statically. mutation of const is a dynamic check, like readonly check
• Idea 2: const is ONLY static. set/setlocal and setvar can't mutate it, because they're known statically. Others can?

---

Why we need const: for Oil blocks. Those shouldn't be variables!

[andychu]

As of 451ed31

we have static checks for Oil assignment keywords. But there will be an ADDITIONAL dynamic check for const (the one that's already there). I think that's a good compromise.

[andychu]

TODO: test it both locally and globally. I think the dynamic check kicks in globally, but we'll see

[andychu]

Done with 0.8.pre11

[andychu]

https://www.oilshell.org/blog/2021/07/release-0.8.12.html