Don't add comments to AST

[gromgit]

It just confuses the REPL into printing nothing, even when it should.

Closes #388.

[odino]

Nice idea! I want to run a couple tests but this looks good at first glance

[gromgit]

Forgot to explain the rationale for this change: Adding comments to the AST actually adds a final null object to the list of expressions that are evaluated. This means that the result of every commented statement is null, and while the effect in the REPL is mostly harmless (failure to print statement results), it will create logic bugs in any other context that tries to use the statement's results for further evaluation.

[odino]

BTW do you think we could do this in the lexer itself? So that we can ignore comments and there would be no reasons to even bring them down to the parser.

I think one of the downsides is that you might end up losing the commented blocks in error lines, but it might be that we currently don't support that either because of:

// We don't really have to do anything when comments
// come in, we can simply ignore them
func (p *Parser) parseComment() ast.Expression {
	return nil
}

[gromgit]

BTW do you think we could do this in the lexer itself? So that we can ignore comments and there would be no reasons to even bring them down to the parser.

Yeah, that might actually work even better. Let me see how best to do it.

I think one of the downsides is that you might end up losing the commented blocks in error lines

I dunno about you, but when I see:

ERROR: identifier not found: nil
	[2:7]	 1 + nil # This should return 1

the trailing comment is just noise.

[odino]

Yeah, I don't think I would be complaining if we didn't have the comments in stacktraces etc :)

…

On Wed, Aug 5, 2020, 1:42 PM Adrian Ho ***@***.***> wrote: BTW do you think we could do this in the lexer itself? So that we can ignore comments and there would be no reasons to even bring them down to the parser. Yeah, that might actually work even better. Let me see how best to do it. I think one of the downsides is that you might end up losing the commented blocks in error lines I dunno about you, but when I see: ERROR: identifier not found: nil [2:7] 1 + nil # This should return 1 the trailing comment is just noise. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#393 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACQFZDDDSUSXSPYWF63T43R7ESOXANCNFSM4PRKG3TQ> .

[gromgit]

Just submitted #394 that eliminates comments at the lexer level. These two PRs are mutually incompatible, so you can choose whichever you prefer, and close the other.

[gromgit]

Since #394 seems to be the preferred approach, I'm closing this PR.