Moved from unicode-width to unicode-display-width for visual grapheme width estimation #210
Conversation
… width estimation
|
Pulling in @djc as requested |
| { | ||
| use unicode_width::UnicodeWidthStr; | ||
| s.width() | ||
| unicode_display_width::width(s) as usize |
There was a problem hiding this comment.
Note: this usize cast is here to avoid changing APIs. It may not be what is wanted?
There was a problem hiding this comment.
This seems okay to me. Would sure be nice to avoid breaking compatibility for this.
| @@ -13,13 +13,13 @@ readme = "README.md" | |||
| rust-version = "1.56.0" | |||
|
|
|||
| [features] | |||
| default = ["unicode-width", "ansi-parsing"] | |||
| default = ["unicode-display-width", "ansi-parsing"] | |||
There was a problem hiding this comment.
If we use a feature named unicode-width that then enables ["dep:unicode-display-width"] this might be backwards incompatible?
src/utils.rs
Outdated
| { | ||
| use unicode_width::UnicodeWidthChar; | ||
| c.width().unwrap_or(0) | ||
| match unicode_display_width::is_double_width(c) { |
There was a problem hiding this comment.
For compatibility, maybe we should replicate the logic from unicode-width here for control characters? Maybe like:
match c.is_control() {
true if c == '\x00' => ?,
true => 0,
false => match unicode_display_width::is_double_width(c) {
true => 2,
false => 1,
}
}
|
Worth an extra note: I'm not sure if |
| use unicode_width::UnicodeWidthChar; | ||
| c.width().unwrap_or(0) | ||
| #[cfg(feature = "unicode-display-width")] | ||
| if c < '\u{7F}' { |
There was a problem hiding this comment.
Why did you go this route? I'd like to avoid referring to explicit code points as much as possible as it would be likely for these to fall out of sync for Unicode updates -- that's what we're trying to rely on other crates for.
This change had to be made in both crates in order to be updated correctly.
The console-rs crate has been bumped an extra version as a result of a behavior change of
char_width()with theunicode-display-widthfeature enabled - it now returns always 1 or 2, never 0 as it used to.The indicatif crate may also need to be bumped a full version as a result of the dependency change, but I'm not sure if you'd consider this a breaking change to the same degree, as it's not an advertised feature.
Related issue: console-rs/indicatif#638
Related PR: console-rs/indicatif#639