Grammar: Statements.

[ehuss]

Cleans up "Statement" definition.
Adds "let" statement.

[ehuss]

There's an issue with this PR. Some statements allow outer attributes, but that is not indicated in the grammar. I see two ways to fix that:

1. Just slap OuterAttribute* on all statements, and then explain which ones are restricted in the text.
2. Carefully annotate which statements accept outer attributes.

Statements that I know of that don't support attributes:

• ;
• if and if let
• Compound expressions like most Operator Expressions and Range Expressions (requires stmt_expr_attributes feature), presumably this is due to precedence.

I lean towards 1, but please let me know which you think would be better.

[matthewjasper]

I lean towards 1, but please let me know which you think would be better.

I think that's probably the right choice.

[ehuss]

I decided to document attributes in a slightly different way. It was more complex than I first thought, as it was accidentally stabilized for expressions in some contexts.

I am a little uneasy about the way statements are defined. Technically the ; is not part of the statement. Trying to define it without made it needlessly more complicated, though.

[Havvy]

What I see LGTM. What I don't see, doesn't. attributes.md should list all places attributes are allowed, and you've shown a bunch of expression locations that I've missed. Can you add them to the list please?

[ehuss]

attributes.md should list all places attributes are allowed

I added to the list, though in a brief form. Hopefully that's ok, it seemed more natural that way instead of stuffing all the rules in one place.

[matthewjasper]

There should be a note that * is lazy here.

[ehuss]

Just to be clear, are you saying that because block-expressions can be confused as to whether it is a statement or a tail expression?

This kinda gets back to my uneasiness with how statements are defined. Technically they do not include the trailing semicolon, they are separated by semicolons. I'm thinking about rewriting the grammar to express it that way. What do you think?

[matthewjasper]

yes { { 2 } } has a tail expression, not a tail-statement.

If you can get statements without ; to work well then I would prefer that but I couldn't work out how to do it myself.

[ehuss]

Ok, I have pushed an update to rewrite statements in such a way that I think is more correct (less wrong?). I also came across some related attributes that were missing.

Please let me know if there's anything I can do to make it easier to review these things. Sample code, links to rustc source, whatever.

[matthewjasper]

I think this is good now. One thing that is maybe worth changing.

[matthewjasper]

You could move these down to Expression statements, also this arguably should be Expression since { { 1 } let x = 0; } does not type check while { { 1 }; let x = 0; } does.

[ehuss]

Ok, moved it down.

It can't be just Expression because { { () } let x = 0; } is valid. That's why I put the sentence about blocks must have the unit type. Generally I don't think the grammar should include type checking rules.

[matthewjasper]

Thanks for this!