Re: Replacing the "if" statement with an "if" expression which supersedes it

Le 27/06/2026 à 15:12, Luther Thompson a écrit :
> I should point out that Lua does not allow an expression to be a statement (unless it's a function call), so letting the last statement of a block be the value of that block would be a major change in the grammar.
>
> Also, while I'd love to see a true ternary operator in Lua, it is fundamentally different from an if-else statement, and I don't think they should be unified. A ternary operator is a single expression resolving to a single value. An if-else can contain multiple statements, which makes side-effects more likely to happen.
>
> Luther


Hello!

That is a good point.
Precisely due to the strange exception that routine ("function") calls are both expressions and statements, I had not noticed that and was assuming that expressions worked as statements, reinforced in that misunderstanding by the REPL's behaviour!

Do you think this (yes, profound) change of accepting expressions as statements would be a bad one?
If so, I would agree with that, but then I would immediately draw the logical consequences and want to fix that strange exception you noted, about the callable entities declared with the keyword `function`, whose calls can be valid statements despite being (also?) expressions.

Maybe Lua needs to acquire a `proc` or `procedure` keyword that declares works as a statement, and to make calls to `function`s invalid as statements.
Functions would only be allowed to contain an expression — no statements.
...but *that* would be a backwards-incompatible, breaking change. And probably regarded as a very bad move.

# (abandoning this line of thought and returning to my initial proposal's consequences) #

At least, accepting expressions as statements would restore consistency in a backwards-compatible way.
Warning when a non-nil value was ignored could also make a lot of sense.
Thinking about backwards compatibility, this would cause new warnings in programs that called "functions" that return a value without using it. This would probably be a good thing, too: I can imagine this leading to welcome bug fixes.

Currently, it is indeed a syntax error as you pointed out, which is different in nature.
But I think the language would be cleaner, more regular, if it systematically accepted expressions that evaluate to nil as statements.

Cheers!


-- 
You received this message because you are subscribed to the Google Groups "lua-l" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].
To view this discussion visit https://groups.google.com/d/msgid/lua-l/461e2afa-85fc-5d9f-5e94-ac5eefc4959e%40free.fr.