feat: strip ANSI escape codes for text element

[bytesnake]

What It Does

Remove ANSI escape codes for Text element. A major drawback is that this pulls in regex dependency

Related Issues

#168

[ccbrown]

I'm fine with the added dependency.

For inspiration, I took a look at what Ink does.

Similar issue: vadimdemedes/ink#362

They strip all ANSI codes out for measurement purposes, but leave certain ANSI codes in during rendering if they wouldn't break the layout. That might be something to consider later, but I'm okay stripping all of them out for now.

Their sanitization: https://github.com/vadimdemedes/ink/blob/70477953d4c97bb2e1754aef9770317d4d05f043/src/sanitize-ansi.ts#L9

Their string measurement: https://github.com/sindresorhus/string-width/blob/main/index.js

I did some benchmarks to see if this would have a notable impact on performance. For strings that don't contain ANSI sequences, the impact is pretty negligible. However, I think it is worth returning the Cow from the strip_ansi function:

diff --git a/packages/iocraft/src/components/text.rs b/packages/iocraft/src/components/text.rs
index 247c4f8..024168c 100644
--- a/packages/iocraft/src/components/text.rs
+++ b/packages/iocraft/src/components/text.rs
@@ -228,7 +228,7 @@ impl Component for Text {
             underline: props.decoration == TextDecoration::Underline,
             italic: props.italic,
         };
-        self.content = strip_ansi(&props.content);
+        self.content = strip_ansi(&props.content).into_owned();
         self.wrap = props.wrap;
         self.align = props.align;
         updater.set_measure_func(Self::measure_func(self.content.clone(), props.wrap));
diff --git a/packages/iocraft/src/strip_ansi.rs b/packages/iocraft/src/strip_ansi.rs
index 56f711e..2fcbd28 100644
--- a/packages/iocraft/src/strip_ansi.rs
+++ b/packages/iocraft/src/strip_ansi.rs
@@ -1,5 +1,5 @@
-use std::sync::LazyLock;
 use regex::Regex;
+use std::{borrow::Cow, sync::LazyLock};
 
 pub const ANSI_REGEX_PATTERN: &str = concat!(
     // OSC branch
@@ -28,6 +28,6 @@ pub const ANSI_REGEX_PATTERN: &str = concat!(
 static ANSI_REGEX: LazyLock<Regex> =
     LazyLock::new(|| Regex::new(ANSI_REGEX_PATTERN).expect("valid ANSI regex"));
 
-pub(crate) fn strip_ansi(string: &str) -> String {
-    ANSI_REGEX.replace_all(string, "").to_string()
+pub(crate) fn strip_ansi(string: &str) -> Cow<str> {
+    ANSI_REGEX.replace_all(string, "")
 }

That makes strip_ansi ~30% faster in the fast case. I don't think this actually matters, since we immediately call into_owned in the text component, but still, it might be a little cleaner and may matter for future uses of the function.

It seems like it's probably worth stripping ANSI in MixedText too?

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.06%. Comparing base (5a3d4c6) to head (621e336).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #169   +/-   ##
=======================================
  Coverage   90.06%   90.06%           
=======================================
  Files          32       33    +1     
  Lines        5333     5337    +4     
  Branches     5333     5337    +4     
=======================================
+ Hits         4803     4807    +4     
  Misses        427      427           
  Partials      103      103           

Files with missing lines	Coverage Δ
packages/iocraft/src/components/text.rs	98.95% <100.00%> (ø)
packages/iocraft/src/strip_ansi.rs	100.00% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.