Update ignore to v5

[fisker]

Previous attempt #6975

• I’ve added tests to confirm my change works.
• (If changing the API or CLI) I’ve documented the changes I’ve made (in the docs/ directory)
• (If the change is user-facing) I’ve added my changes to changelog_unreleased/*/pr-XXXX.md file following changelog_unreleased/TEMPLATE.md.
• I’ve read the contributing guidelines.

✨Try the playground for this PR✨

[thorn0]

I'd like to take a deeper look at this PR, but I need time for that. Please don't merge for now.

[alexander-akait]

Ready for review?

[fisker]

It's ready long time ago, just solved the conflicts, still waiting for @thorn0 review.

[alexander-akait]

Looks good for me

[thorn0]

Here's what bothers me about this change.

./foo/ignore.txt

*.js

./bar.js

1

prettier --ignore-path foo/ignore.txt .

Result with Prettier 2.0.5: bar.js is ignored.
Result with this PR: bar.js isn't ignored.

So this change is breaking.

[fisker]

@thorn0 ignore@5 works like this, and I think it's reasonable, add I did add a test to confirm that "Should not ignore files in parent dir".

---

I did edit this comment several times, get little confused by this problem, I think this(original) is right explanation.

[fisker]

FYI, the related file path pass to ignore.ignores is ../bar.js, so ignore didn't ignore files in parent dir. (This change was made in #7588)

[thorn0]

I understand why it works this way, but this is still a breaking change.

[fisker]

If the breaking change is what you worried about, we can merge later.

Most important is, which behavior do you prefer? I think the new one not bad.

[thorn0]

The new behavior is aligned with Git. It looks simpler and cleaner to me, especially if support for multiple ignore files is going to be added. But as long as Prettier doesn't have that support and because we don't know how and why people usually use --ignore-path with subdirectories, I'd prefer to keep the current behavior to stay on the safe side.

In order to both keep the behavior and update ignore, a workaround seems to be possible: a separate ignorer for paths that start with ../. It should use only patterns that don't have a slash at the beginning or middle and patterns that start with **/. It would be safe then to strip .. from the relative paths passed to it.

[fisker]

Use ignore@5.2.0 with allowRelativePaths: true options, we don't need this anymore. See #11990