improve the heuristics to detect commented out code

[sylr]

Follow up of #6 (comment)

I like to make a distinction between comments and commented code, i.e.:

// This is a comment
if thisiscommentedcode == false {

}

//// This is a comment
//if thisiscommentedcode == true {
//
//}

This way I can grep for commented code without having comments returned and vice versa.

Regards.

[mvdan]

What are you seeing exactly? Gofumpt will not add any spaces to the code you showed as of the latest master version, and the latest stable version is the same.

$ cat /tmp/f.go
package p

func f() {
	// This is a comment
	if thisiscommentedcode == false {

	}

	//// This is a comment
	//if thisiscommentedcode == true {
	//
	//}
}
$ gofumpt -version
v0.3.1
$ gofumpt /tmp/f.go
package p

func f() {
	// This is a comment
	if thisiscommentedcode == false {
	}

	//// This is a comment
	//if thisiscommentedcode == true {
	//
	//}
}
$ gofumpt -d /tmp/f.go
diff -u /tmp/f.go.orig /tmp/f.go
--- /tmp/f.go.orig	2022-06-10 13:25:25.200203324 +0100
+++ /tmp/f.go	2022-06-10 13:25:25.200203324 +0100
@@ -3,7 +3,6 @@
 func f() {
 	// This is a comment
 	if thisiscommentedcode == false {
-
 	}
 
 	//// This is a comment

[sylr]

package main

import "fmt"

func main() {
	pwet := true
	fmt.Printf("%b", pwet)
	if pwet {
		//fmt.Printf("%b", pwet)
		return
	}
}

$ ~/git/gofumpt/gofumpt -w .
$ git --no-parger diff

diff --git main.go main.go
index 8f26369..22084d7 100644
--- main.go
+++ main.go
@@ -6,7 +6,7 @@ func main() {
 	pwet := true
 	fmt.Printf("%b", pwet)
 	if pwet {
-		//fmt.Printf("%b", pwet)
+		// fmt.Printf("%b", pwet)
 		return
 	}
 }

[mvdan]

The heuristic is at

gofumpt/format/format.go

Lines 415 to 419 in 8f1392a

	r, _ := utf8.DecodeRuneInString(body)
	if !unicode.IsLetter(r) && !unicode.IsNumber(r) && !unicode.IsSpace(r) {
	// this line could be code like "//{"
	continue groupLoop
	}

; if any of the lines in a contiguous block of comments matches, then we conservatively treat the entire thing as code, so we don't add any spaces.

In your case, there is just one comment and it doesn't match our heuristic - which is currently "is the first character special?".

I guess one way to make the heuristic a bit more clever is to also treat a comment line as code if it parses as a valid Go statement or declaration. That's not particularly easy to do unfortunately, given that https://pkg.go.dev/go/parser only allows parsing files or expressions, but it is doable.