Open agentm opened 1 year ago
TutorialD (master/main): x:=with (x as true) x
TutorialD (master/main): :showexpr x
┌┐
││
├┤
└┘
It seems fine on my machine. I would like to give opinions but I found myself not understand the issue here at all. What's the risk?
Indeed, it's just on a future branch, but what's the point of supporting that expression? It makes substitution much more difficult and optimizations have to track shadowed names recursively.
I see. I can't think of a useful example either.
Name shadowing can be quite a footgun. Example:
The original expression should almost certainly be rejected, but the relational expression validation is sound, so we would need to check that the new relvarname "x" is not mentioned in the with clause names.
I am toying with the idea of preventing name shadowing of relvars or other with clause macro names (even macros shadowing macros). It does seem that name shadowing can be sometimes useful when moving subexpressions around, but the risk and cost of mistakes here is very high.
Any objection?