Yeah I don't want to over complicate this because not only is very easy to make it slower for no real gain in most cases but the more complex the rules are the more edge cases they will have and with peg that can also affect other rules

Agreed on the point about complexity. Is this an improvement over the status quo though?

Is this an improvement over the status quo though?

Yes, I think it is. Compare my current branch:

Withmain:

Currently, we point at different chars in seemingly random places::,]
When the problem always starts with*

@lysnikolaou are you ok if we go with this?

Adding such test can be done, but is rather hard to do, since there's no infrastructure for such tests as far as I know.

We normally either add them in test exceptions or test syntax

I proposed a test in a PR against your branch.

I'm okay to go with this if everyone else agrees, but I'm not convinced this is clearer. The caret should point to the first character that does not adhere to our syntax and this breaks that assumption.

The current caret position is not random. It actually points to the correct place. In the above examples@sobolevn showed, ifA is anndarray, it's the] or the: that make this invalid syntax since the following would be a valid index:

A[*(1,)]

@lysnikolaou yes, this is a good point. Maybe we can come up with a better solution? Do you have an idea about what we can do?

@lysnikolaou yes, this is a good point. Maybe we can come up with a better solution? Do you have an idea about what we can do?

Ideally we would be able to place tildes below the whole expression and a caret below the position that breaks the syntax. I don't think we have that in the parser though and the complexity would be too much.

Maybe just keep this as is?

Maybe just keep this as is?

@lysnikolaou I think it's better just keep this as is. It's cleaner and clearer. There is no need to make tildes and complicate the code.

Ok, lets keep it as is :)

Maybe a test and link to this discussion would be good to prevent this changing yet again?