client/parser/scanner.go: invalid rune quoting

[3052]

runes should be quoted with %q, not '%c'

https://github.com/alda-lang/alda/blob/8608a982bdc5522ed9b6f697ad294bff61323903/client/parser/scanner.go#L34-L51

%q  a single-quoted character literal safely escaped with Go syntax.

https://godocs.io/fmt#hdr-Printing

[daveyarwood]

Hi, thanks for the discussion!

I don't fully understand the issue. I played around with this a bit in Go Playground, and as far as I can tell, the code we have now where we're printing '%c' (with added single quotes) produces identical results to what we would get if we switched to %q (without quotes). For example:

package main

import "fmt"

func main() {
    fmt.Printf("%%c is %c, %%q is %q\n", '🐧', '🐧')
}

Produces the following output:

%c is 🐧, %q is '🐧'

Is there an edge case that I'm missing?

[3052]

yeah the single quote itself, '

[daveyarwood]

I'm inclined to leave this the way it is. The explicit single quotes in the Alda error message are unrelated to Go's choice to put single quotes around a rune when you print it via %q. Hypothetically, if we wanted to change the single quotes in the error message in the future, the way we have it now would make it easier to understand how to do that.

[3052]

suit yourself, but in some cases you are going to end up with ''', which is obviously invalid

[daveyarwood]

That's a good example to consider!

I tried that in Go Playground just now, and I got this output:

%c is ', %q is '\''

Given the choice between presenting end users with ''' or '\'', I think I would actually prefer '''. Yes, it's weird-looking, but I think the backslash could be potentially more confusing.

[daveyarwood]

Another option worth considering here is to remove the single-quotes entirely and just display %c. For example, in the case of a single quote, an error message might look like:

Unexpected ' at the top level

whereas currently, it would be:

Unexpected ''' at the top level

[3052]

%c is bad, because any control, white space, or otherwise non printing characters are not going to produce helpful output.

[daveyarwood]

Agreed, although we do have explicit checks for those kinds of characters before we fall back to the %c representation.

I appreciate the suggestion to use %q instead. I don't see an immediate need to switch, but I will keep it in mind! Thanks again for the discussion.

[3052]

as mentioned, the current checks are not robust, so other characters are going to fall through:

package main

import "fmt"

func main() {
   fmt.Printf("'%c'\n", 0x200E) 
   fmt.Printf("%q\n", 0x200E) 
}

result:

'‎'
'\u200e'

I would again argue that %q is more helpful because of this reason, and because as mentioned, '%c' produces invalid Go code if the input is '.