Using -switch may lead to dead code and the Go compiler complaining

[Necoro]

I noticed that when using -switch that the resulting code seems to be sub-par and does checks multiple times. For instance, I see

switch buffer[position] {
// [...]
case ' ':
	if buffer[position] != rune(' ') {
		goto l0
	}
	position++

The if is completely redundant (and slightly removes the point of switch if the ifs are just kept but now inside a switch :))

There are other, more involved examples:

switch buffer[position] {
case '0', '1', '2', '3', '4', '5', '6', '7', '8', '9':
	{
		position5, tokenIndex5 := position, tokenIndex
		if c := buffer[position]; c < rune('0') || c > rune('9') {
			goto l6
		}
		position++
		goto l5
	l6:
		position, tokenIndex = position5, tokenIndex5
		if c := buffer[position]; c < rune('0') || c > rune('9') {
			goto l0
		}
		position++
	}
	l5:
		break

Here the whole code except for position++ is superfluous.

Example PEG:

package minimal

type Parser Peg {
}

OtherAttr <- [[A-Z_ 0-9]]+ _ ':' [^\n]*

_ <- [ \t]*

[Necoro]

Ah! Trying the main branch this issue no longer exists.
Time to cut a new release? And/or update the README to motivate people to install @main instead of @latest? :)

[Necoro]

Reopening: The issue is not completely gone in main. The second block sometimes still exists and leads to the go compiler complain about unused labels:

case '0', '1', '2', '3', '4', '5', '6', '7', '8', '9':
	{
		position39, tokenIndex39 := position, tokenIndex
		position++
		goto l39
	l40:
		position, tokenIndex = position39, tokenIndex39
		if c := buffer[position]; c < '0' || c > '9' {
			goto l34
		}
		position++
	}
l39:
	break

Here, l40 and the code it references is dead.

[dolmen]

@Necoro Could you provide a grammar that allows to reproduce your issue?

[Necoro]

@Necoro Could you provide a grammar that allows to reproduce your issue?

I provided one in the issue description. Although I haven't checked since then whether the issue still exists.