Add auto modifier to hyperlink argument to match both ls behavior and color argument behavior

[mrnossiom]

Description

Add a WHEN modifier to the --hyperlink argument. This matches both ls behavior and the --color argument behavior.

I think ls is doing it right because you might want to enable --hyperlink in a shell alias but still get plain text when piping into some other program without control chars in the way. Setting --hyperlink in a shell alias is useful for terminal integration (e.g. Kitty).

The public API is compatible since --hyperlink defaults to auto and doesn't break. However, it changes the behavior, with no value the option worked like always but it now defaults to auto. Even though I consider auto to be the right default, maybe it could be changed to always to keep old meaning.

Closes #1059 (duplicate)

[mrnossiom]

I just noticed that there is a bug in the logic of calculating items lengths with shows when combined with grid option. I guess the hyperlink feature is not used much and that no one reported that.

The target/release/eza binary is built out of 7bb01a2f (right before this message)

Argument sets used:

eza --color=auto --hyperlink -GlhFa
eza --color=auto --hyperlink=always -GlhFa
eza --color=auto --hyperlink=never -GlhFa

[PThorpe92]

I just noticed that there is a bug in the logic of calculating items lengths with shows when combined with grid option. I guess the hyperlink feature is not used much and that no one reported that.

This probably came leftover from #742, which I tested but apparently not the particular combination of hyperlinks, -F and grid because -F seems to be the problem.

with -F

without

Thanks for reporting this! 👍

[mrnossiom]

Do you know why it would trigger only when hyperlinks are enabled ?

[MartinFillon]

Do you know why it would trigger only when hyperlinks are enabled ?

As hyperlinks are not calculated and taken into account when not enabled they are not count for when taking the width

[mrnossiom]

Rebased and tested

It now works!

[nikelborm]

#1059 Is not duplicate. The changes are different. On top of the same functionality I did refactoring and shell completions

[mrnossiom]

Right, I had forgotten to update the help text.

[cafkafk]

Thanks for your contribution! We don't use merge updates, please remove the merge update commit and use rebase instead to update the branch

[MartinFillon]

lgtm

[cafkafk]

Considering the behavior change, this is breaking.

Also, I've found that some terminal emulators do in fact parse hyperlinks that are piped perfectly fine (e.g. kitty) while others (iterm2) doesn't, making the decision as to whether or not to default to auto more important.

I'm wondering what others thing re: defaulting to auto, I'm leaning defaulting to always, since to me, not supporting piped hyperlinks is worse compliance with hyperlinks, and we should support the case where a terminal emulator is doing it's best.

See also:

• bug: bad format when using --hyperlink & grep #1156
• feat: conditionally disable/enable options  #1160

[mrnossiom]

My initial argument for making hyperlink follow the color flag behavior is that I enable it in a ls alias (together with color and icons).

When piping into fzf for example, the icons and the color ANSI codes are disabled but the hyperlinks markers are still there. There is no way to enable hyperlink by default and to disable it when being pipeed in other software.

One way out to make this less breaking is just having --hyperlink default to always but adding the possibility to do --hyperlink=auto. This is still a breaking change I believe, if you had --hyperlink smth_else_not_related