cut: fix -d= (#2424)

[jhscheer]

fix for #2424

GNU's cut supports -d= to set the delimiter to =.
Clap parsing is limited in this situation, see.
Since clap parsing handles -d= as delimiter explicitly set to "" we can use this as basis for a simple workaround.

The downside is that uu_cut loses the ability to handle empty delimiters.
I don't know how often this is used, or if it is useful at all, but I consider this an acceptable tradeoff.
GNU's behaviour for an empty delimiter is to echo the input.

Other alternatives for a workaround to fix #2424 could be:

1. parse/regex args before calling App::get_matches()
2. set Arg::with_name(options::DELIMITER).empty_values(false) use App::get_matches_safe(), handle the error manually and set the delimiter to =

[ankurdhama]

I think you can just replace "\0".to_owned() with "=".to_owned()

[jhscheer]

I think you can just replace "\0".to_owned() with "=".to_owned()

Which line are you referring to?
Also I'm confused because this PR doesn't have a line containing the change you propose!?

[ankurdhama]

I think you can just replace "\0".to_owned() with "=".to_owned()

Which line are you referring to?
Also I'm confused because this PR doesn't have a line containing the change you propose!?

Sorry about that, line number 544 in original file (line number 553 in the PR file).

[jhscheer]

Sorry about that, line number 544 in original file (line number 553 in the PR file).

Oh thanks for mentioning this, I didn't notice that.

[jhscheer]

I don't know why

coreutils/src/uu/cut/src/cut.rs

Lines 543 to 547 in 439b7e0

	let delim = if delim.is_empty() {
	"\0".to_owned()
	} else {
	delim.to_owned()
	};

is there. This doesn't seem to be the behaviour of GNU's cut:

$ echo "--libdir=./out/lib" | gcut -f2 -d "" | hexdump -C
00000000  2d 2d 6c 69 62 64 69 72  3d 2e 2f 6f 75 74 2f 6c  |--libdir=./out/l|
00000010  69 62 0a                                          |ib.|
00000013

$ echo "--libdir=./out/lib" | hexdump -C
00000000  2d 2d 6c 69 62 64 69 72  3d 2e 2f 6f 75 74 2f 6c  |--libdir=./out/l|
00000010  69 62 0a                                          |ib.|
00000013

[tertsdiepraam]

Nice! The only difference with GNU I can still find is:

$ # GNU
$ echo "a=b" | cut -f2 --delimiter=
a=b
$ # uutils
# echo "a=b" | cut -f2 --delimiter=
b

But I guess that's not useful anyway.

[moxuze]

I noticed that -d '' means '\0' (NULL character) in GNU's cut, please see:
https://github.com/coreutils/coreutils/blob/cc01b8a8f43bd1e02339322595f7a20e72a8c155/src/cut.c#L497

I tried these commands in my zsh (ArchLinux):

$ echo 'ab\0cd' | cut -f 1,2 -d '' --output-delimiter=Z | hexdump -C
00000000  61 62 5a 63 64 0a                                 |abZcd.|
00000006
$ echo 'ab\0cd' | uu-cut -f 1,2 -d '' --output-delimiter=Z | hexdump -C
00000000  61 62 00 63 64 0a                                 |ab.cd.|
0000000

Using -d '' to set the delimiter to = may not be a good idea.

coreutils/src/uu/cut/src/cut.rs

Line 544 in 439b7e0

"\0".to_owned()

And this line cannot be execute in any situation!