safe_rewrite should actually do some validation about bud

[palvaro]

safe rewrite will take this block, try to parse it, extract some information and possibly transform the parse tree, and then rewrite the block as ruby again. some illegal ruby will break this and we'll catch it then. some illegal ruby will look 'legal enough' to get parsed and rewritten and will fail at runtime, as it would have if written in 'pure' ruby. wouldn't doing what you suggest ("catch stuff" instead of letting ruby catch it) duplicate functionality?

[billmarczak]

it duplicates functionality in exactly the same way a static type system duplicates the functionality of a runtime exception with a compile-time exception. in other words, what i'm proposing is more compile-time safety. in general, i think that's a good thing because catching errors at runtime may be very costly, but the specific reason i opened this ticket is because in the REPL, no execution happens by default until the user explicitly types a "/tick" command. this means they can enter arbitrarily many messed up statements (undefined identifiers, type errors, etc), and nothing will fail until they hit "/tick". what do you do when something fails on /tick? if you have static checking, then nothing can "fail" on tick. if you don't have static checking, then either the REPL has to understand the error thrown by Ruby, and understand how to fix the program, or the user has to understand the error message and do some back and forth where they remove and re-add rules to get it working. the last way seems kind of ugly, but then again i suppose it's what you'd expect from an interpreted, duck-typed language like Ruby.

i just thought of another way to fix this in the REPL, I could tick by default whenever any code was added.