Treat newlines as width 0 in the 0.1 stream, publish 0.1.14 (#67)

Manishearth · web-flow · commit 9eaafa5f91b7 · 2024-09-19T19:06:03.000Z
* Treat newlines as width 0

* Publish 0.1.14
diff --git a/Cargo.toml b/Cargo.toml
@@ -1,7 +1,7 @@
 [package]
 
 name = "unicode-width"
-version = "0.1.13"
+version = "0.1.14"
 authors = [
     "kwantam <kwantam@gmail.com>",
     "Manish Goregaokar <manishsmail@gmail.com>",
diff --git a/scripts/unicode.py b/scripts/unicode.py
@@ -1281,7 +1281,10 @@ def lookup_fns(
     s += """
     if c <= '\\u{A0}' {
         match c {
-            '\\n' => (1, WidthInfo::LINE_FEED),
+            // According to the spec, LF should be width 1, which is how it is often rendered when it is forced to have a single-line rendering
+            // However, this makes it harder to use this crate to calculate line breaks, and breaks assumptions of downstream crates.
+            // https://github.com/unicode-rs/unicode-width/issues/60
+            '\\n' => (0, WidthInfo::LINE_FEED),
             '\\r' if next_info == WidthInfo::LINE_FEED => (0, WidthInfo::DEFAULT),
             _ => (1, WidthInfo::DEFAULT),
         }
diff --git a/src/tables.rs b/src/tables.rs
@@ -215,7 +215,10 @@ fn width_in_str(c: char, mut next_info: WidthInfo) -> (i8, WidthInfo) {
     }
     if c <= '\u{A0}' {
         match c {
-            '\n' => (1, WidthInfo::LINE_FEED),
+            // According to the spec, LF should be width 1, which is how it is often rendered when it is forced to have a single-line rendering
+            // However, this makes it harder to use this crate to calculate line breaks, and breaks assumptions of downstream crates.
+            // https://github.com/unicode-rs/unicode-width/issues/60
+            '\n' => (0, WidthInfo::LINE_FEED),
             '\r' if next_info == WidthInfo::LINE_FEED => (0, WidthInfo::DEFAULT),
             _ => (1, WidthInfo::DEFAULT),
         }
@@ -507,7 +510,10 @@ fn width_in_str_cjk(c: char, mut next_info: WidthInfo) -> (i8, WidthInfo) {
     }
     if c <= '\u{A0}' {
         match c {
-            '\n' => (1, WidthInfo::LINE_FEED),
+            // According to the spec, LF should be width 1, which is how it is often rendered when it is forced to have a single-line rendering
+            // However, this makes it harder to use this crate to calculate line breaks, and breaks assumptions of downstream crates.
+            // https://github.com/unicode-rs/unicode-width/issues/60
+            '\n' => (0, WidthInfo::LINE_FEED),
             '\r' if next_info == WidthInfo::LINE_FEED => (0, WidthInfo::DEFAULT),
             _ => (1, WidthInfo::DEFAULT),
         }
diff --git a/tests/tests.rs b/tests/tests.rs
@@ -214,18 +214,23 @@ fn test_control_line_break() {
     assert_width!('\r', None, None);
     assert_width!('\n', None, None);
     assert_width!("\r", 1, 1);
-    assert_width!("\n", 1, 1);
-    assert_width!("\r\n", 1, 1);
+    // This is 0 due to #60
+    assert_width!("\n", 0, 0);
+    assert_width!("\r\n", 0, 0);
     assert_width!("\0", 1, 1);
-    assert_width!("1\t2\r\n3\u{85}4", 7, 7);
-    assert_width!("\r\u{FE0F}\n", 2, 2);
-    assert_width!("\r\u{200D}\n", 2, 2);
+    assert_width!("1\t2\r\n3\u{85}4", 6, 6);
+    assert_width!("\r\u{FE0F}\n", 1, 1);
+    assert_width!("\r\u{200D}\n", 1, 1);
 }
 
 #[test]
 fn char_str_consistent() {
     let mut s = String::with_capacity(4);
     for c in '\0'..=char::MAX {
+        // Newlines are special cased (#60)
+        if c == '\n' {
+            continue;
+        }
         s.clear();
         s.push(c);
         assert_eq!(c.width().unwrap_or(1), s.width());
@@ -418,6 +423,10 @@ fn test_khmer_coeng() {
             assert_width!(format!("\u{17D2}{c}"), 0, 0);
             assert_width!(format!("\u{17D2}\u{200D}\u{200D}{c}"), 0, 0);
         } else {
+            // Newlines are special cased (#60)
+            if c == '\n' {
+                continue;
+            }
             assert_width!(
                 format!("\u{17D2}{c}"),
                 c.width().unwrap_or(1),
@@ -588,6 +597,11 @@ fn emoji_test_file() {
     }
 }
 
+#[test]
+fn test_newline_zero_issue_60() {
+    assert_width!("a\na", 2, 2);
+}
+
 // Test traits are unsealed
 
 #[cfg(feature = "cjk")]

Original file line number	Diff line number	Diff line change
`@@ -215,7 +215,10 @@ fn width_in_str(c: char, mut next_info: WidthInfo) -> (i8, WidthInfo) {`
`215`	`215`	`}`
`216`	`216`	`if c <= '\u{A0}' {`
`217`	`217`	`match c {`
`218`		`- '\n' => (1, WidthInfo::LINE_FEED),`
	`218`	`+ // According to the spec, LF should be width 1, which is how it is often rendered when it is forced to have a single-line rendering`
	`219`	`+ // However, this makes it harder to use this crate to calculate line breaks, and breaks assumptions of downstream crates.`
	`220`	`+ // https://github.com/unicode-rs/unicode-width/issues/60`
	`221`	`+ '\n' => (0, WidthInfo::LINE_FEED),`
`219`	`222`	`'\r' if next_info == WidthInfo::LINE_FEED => (0, WidthInfo::DEFAULT),`
`220`	`223`	`_ => (1, WidthInfo::DEFAULT),`
`221`	`224`	`}`
`@@ -507,7 +510,10 @@ fn width_in_str_cjk(c: char, mut next_info: WidthInfo) -> (i8, WidthInfo) {`
`507`	`510`	`}`
`508`	`511`	`if c <= '\u{A0}' {`
`509`	`512`	`match c {`
`510`		`- '\n' => (1, WidthInfo::LINE_FEED),`
	`513`	`+ // According to the spec, LF should be width 1, which is how it is often rendered when it is forced to have a single-line rendering`
	`514`	`+ // However, this makes it harder to use this crate to calculate line breaks, and breaks assumptions of downstream crates.`
	`515`	`+ // https://github.com/unicode-rs/unicode-width/issues/60`
	`516`	`+ '\n' => (0, WidthInfo::LINE_FEED),`
`511`	`517`	`'\r' if next_info == WidthInfo::LINE_FEED => (0, WidthInfo::DEFAULT),`
`512`	`518`	`_ => (1, WidthInfo::DEFAULT),`
`513`	`519`	`}`