Remove more usages of JsStr in favor of just JsString (#4578)

Move the following to `JsString` (from `JsStr`):

- `get` and `get_expect` now return `JsString`. Uses `code_point_at()`
instead of `get(usize)`
- Move the display types to using a ref to the `JsString` being
displayed. Using a `JsString` reference instead of a clone makes it
clearer that calling `string.display()` should be temporary (and has a
lifetime).
- Move trim to `JsString` and return `JsString` directly.
- Have the code points iterator be part of the `JsString` API and its
`vtable`.
- Add inlines and other small time optimizations, like `trim()` not
slicing twice.
- This results in less conversions from JsStr to JsString. It's more
performant because creating a `JsString` from a slice is as fast (thanks
to `SliceString`) than a `JsStr` on a slice, but creating a `JsString`
from it is faster now.

The end goal is to get rid of `JsStr` in its current form.

Benchmarks on my machine:

| Benchmark    | Main | This PR |
|--------------|------|---------|
| Richards     | 235  | 255     |
| DeltaBlue    | 245  | 254     |
| Crypto       | 207  | 219     |
| RayTrace     | 485  | 525     |
| EarleyBoyer  | 594  | 632     |
| RegExp       | 87.2 | 90.4    |
| Splay        | 789  | 896     |
| NavierStokes | 484  | 490     |
| **Total**    | 322  | 342     |

Author: hansl

core/engine/src/builtins/intl/segmenter/iterator.rs

@@ -127,7 +127,7 @@ impl SegmentIterator {
                 .segmenter
                 .downcast_ref::<Segmenter>()
                 .expect("segment iterator object should contain a segmenter");
-            let mut segments = segmenter.native.segment(string);
+            let mut segments = segmenter.native.segment(string.variant());
             // the first elem is always 0.
             segments.next();
             segments

core/engine/src/builtins/intl/segmenter/mod.rs

@@ -1,15 +1,7 @@
 use std::ops::Range;
 
-use boa_gc::{Finalize, Trace};
-use icu_collator::provider::CollationDiacriticsV1;
-use icu_locale::Locale;
-use icu_segmenter::{
-    GraphemeClusterSegmenter, SentenceSegmenter, WordSegmenter,
-    options::{SentenceBreakOptions, WordBreakOptions},
-};
-
 use crate::{
-    Context, JsArgs, JsData, JsNativeError, JsResult, JsStr, JsString, JsSymbol, JsValue,
+    Context, JsArgs, JsData, JsNativeError, JsResult, JsString, JsSymbol, JsValue,
     builtins::{
         BuiltInBuilder, BuiltInConstructor, BuiltInObject, IntrinsicObject,
         options::{get_option, get_options_object},
@@ -21,6 +13,14 @@ use crate::{
     realm::Realm,
     string::StaticJsStrings,
 };
+use boa_gc::{Finalize, Trace};
+use boa_string::JsStrVariant;
+use icu_collator::provider::CollationDiacriticsV1;
+use icu_locale::Locale;
+use icu_segmenter::{
+    GraphemeClusterSegmenter, SentenceSegmenter, WordSegmenter,
+    options::{SentenceBreakOptions, WordBreakOptions},
+};
 
 mod iterator;
 mod options;
@@ -62,9 +62,12 @@ impl NativeSegmenter {
 
     /// Segment the passed string, returning an iterator with the index boundaries
     /// of the segments.
-    pub(crate) fn segment<'l, 's>(&'l self, input: JsStr<'s>) -> NativeSegmentIterator<'l, 's> {
-        match input.variant() {
-            crate::string::JsStrVariant::Latin1(input) => match self {
+    pub(crate) fn segment<'l, 's>(
+        &'l self,
+        input: JsStrVariant<'s>,
+    ) -> NativeSegmentIterator<'l, 's> {
+        match input {
+            JsStrVariant::Latin1(input) => match self {
                 Self::Grapheme(g) => {
                     NativeSegmentIterator::GraphemeLatin1(g.as_borrowed().segment_latin1(input))
                 }
@@ -75,7 +78,7 @@ impl NativeSegmenter {
                     NativeSegmentIterator::SentenceLatin1(s.as_borrowed().segment_latin1(input))
                 }
             },
-            crate::string::JsStrVariant::Utf16(input) => match self {
+            JsStrVariant::Utf16(input) => match self {
                 Self::Grapheme(g) => {
                     NativeSegmentIterator::GraphemeUtf16(g.as_borrowed().segment_utf16(input))
                 }

core/engine/src/builtins/intl/segmenter/segments.rs

@@ -88,7 +88,7 @@ impl Segments {
         // 8. Let startIndex be ! FindBoundary(segmenter, string, n, before).
         // 9. Let endIndex be ! FindBoundary(segmenter, string, n, after).
         let (range, is_word_like) = {
-            let mut segments = segmenter.native.segment(segments.string.as_str());
+            let mut segments = segmenter.native.segment(segments.string.variant());
             std::iter::from_fn(|| segments.next().map(|i| (i, segments.is_word_like())))
                 .tuple_windows()
                 .find(|((i, _), (j, _))| (*i..*j).contains(&n))

core/engine/src/builtins/json/mod.rs

@@ -386,7 +386,7 @@ impl Json {
         // 7. Else if Type(space) is String, then
         } else if let Some(s) = space.as_string() {
             // a. If the length of space is 10 or less, let gap be space; otherwise let gap be the substring of space from 0 to 10.
-            js_string!(s.get(..10).unwrap_or(s.as_str()))
+            s.get(..10).unwrap_or(s)
         // 8. Else,
         } else {
             // a. Let gap be the empty String.

core/engine/src/builtins/number/globals.rs

@@ -247,7 +247,7 @@ pub(crate) fn parse_int(_: &JsValue, args: &[JsValue], context: &mut Context) ->
     //     0 digit, at the option of the implementation; and if R is not 2, 4, 8, 10, 16, or 32, then
     //     mathInt may be an implementation-approximated value representing the integer value that is
     //     represented by Z in radix-R notation.)
-    let math_int = from_js_str_radix(z, r).expect("Already checked");
+    let math_int = from_js_str_radix(z.as_str(), r).expect("Already checked");
 
     // 15. If mathInt = 0, then
     //     a. If sign = -1, return -0𝔽.
@@ -303,11 +303,6 @@ pub(crate) fn parse_float(
     args: &[JsValue],
     context: &mut Context,
 ) -> JsResult<JsValue> {
-    const PLUS_CHAR: u16 = b'+' as u16;
-    const MINUS_CHAR: u16 = b'-' as u16;
-    const LOWER_CASE_I_CHAR: u16 = b'i' as u16;
-    const UPPER_CASE_I_CHAR: u16 = b'I' as u16;
-
     let Some(string) = args.first() else {
         return Ok(JsValue::nan());
     };
@@ -333,18 +328,34 @@ pub(crate) fn parse_float(
     // 5. Let parsedNumber be ParseText(trimmedPrefix, StrDecimalLiteral).
     // 6. Assert: parsedNumber is a Parse Node.
     // 7. Return the StringNumericValue of parsedNumber.
-    let (positive, prefix) = match trimmed_string.get(0) {
-        Some(PLUS_CHAR) => (true, trimmed_string.get(1..).unwrap_or(JsStr::latin1(&[]))),
-        Some(MINUS_CHAR) => (false, trimmed_string.get(1..).unwrap_or(JsStr::latin1(&[]))),
-        _ => (true, trimmed_string),
+    let (positive, prefix) = match trimmed_string
+        .code_unit_at(0)
+        .and_then(|c| char::from_u32(u32::from(c)))
+    {
+        Some('+') => (
+            true,
+            trimmed_string
+                .get(1..)
+                .unwrap_or(StaticJsStrings::EMPTY_STRING),
+        ),
+        Some('-') => (
+            false,
+            trimmed_string
+                .get(1..)
+                .unwrap_or(StaticJsStrings::EMPTY_STRING),
+        ),
+        _ => (true, trimmed_string.clone()),
     };
 
     if prefix.starts_with(js_str!("Infinity")) {
         if positive {
             return Ok(JsValue::positive_infinity());
         }
         return Ok(JsValue::negative_infinity());
-    } else if let Some(LOWER_CASE_I_CHAR | UPPER_CASE_I_CHAR) = prefix.get(0) {
+    } else if let Some('I' | 'i') = prefix
+        .code_unit_at(0)
+        .and_then(|c| char::from_u32(u32::from(c)))
+    {
         return Ok(JsValue::nan());
     }
 

core/engine/src/builtins/regexp/mod.rs

@@ -1788,7 +1788,7 @@ impl RegExp {
         // 17. Return the string-concatenation of accumulatedResult and the substring of S from nextSourcePosition.
         Ok(js_string!(
             &JsString::from(&accumulated_result[..]),
-            s.get_expect(next_source_position..)
+            &s.get_expect(next_source_position..)
         )
         .into())
     }

core/engine/src/builtins/string/mod.rs

@@ -620,11 +620,10 @@ impl String {
 
         match position {
             // 4. Let size be the length of S.
-            IntegerOrInfinity::Integer(i) if i >= 0 => {
+            IntegerOrInfinity::Integer(i) if i >= 0 && i < string.len() as i64 => {
                 // 6. Return the Number value for the numeric value of the code unit at index position within the String S.
-                Ok(string
-                    .get(i as usize)
-                    .map_or_else(JsValue::nan, JsValue::from))
+                // SAFETY: already validated the index.
+                Ok(unsafe { string.code_unit_at(i as usize).unwrap_unchecked() }.into())
             }
             // 5. If position < 0 or position ≥ size, return NaN.
             _ => Ok(JsValue::nan()),
@@ -1043,7 +1042,7 @@ impl String {
         };
 
         // 10. Let preserved be the substring of string from 0 to position.
-        let preserved = JsString::from(string.get_expect(..position));
+        let preserved = string.get_expect(..position);
 
         let replacement = match replace_value {
             // 11. If functionalReplace is true, then
@@ -1080,7 +1079,7 @@ impl String {
         Ok(js_string!(
             &preserved,
             &replacement,
-            &JsString::from(string.get_expect(position + search_length..))
+            &string.get_expect(position + search_length..)
         )
         .into())
     }
@@ -1675,7 +1674,7 @@ impl String {
         // 2. Return ? TrimString(S, end).
         let object = this.require_object_coercible()?;
         let string = object.to_string(context)?;
-        Ok(js_string!(string.trim_end()).into())
+        Ok(string.trim_end().into())
     }
 
     /// [`String.prototype.toUpperCase()`][upper] and [`String.prototype.toLowerCase()`][lower]
@@ -1957,9 +1956,8 @@ impl String {
         if separator_length == 0 {
             // a. Let head be the substring of S from 0 to lim.
             // b. Let codeUnits be a List consisting of the sequence of code units that are the elements of head.
-            let head = this_str
-                .get(..lim)
-                .unwrap_or(this_str.as_str())
+            let head_str = this_str.get(..lim).unwrap_or(this_str);
+            let head = head_str
                 .iter()
                 .map(|code| js_string!(std::slice::from_ref(&code)).into());
 

core/engine/src/builtins/string/tests.rs

@@ -761,7 +761,7 @@ fn char_at() {
 #[test]
 fn char_code_at() {
     run_test_actions([
-        TestAction::assert_eq("'abc'.charCodeAt-1", f64::NAN),
+        TestAction::assert_eq("'abc'.charCodeAt(-1)", f64::NAN),
         TestAction::assert_eq("'abc'.charCodeAt(1)", 98),
         TestAction::assert_eq("'abc'.charCodeAt(9)", f64::NAN),
         TestAction::assert_eq("'abc'.charCodeAt()", 97),

core/engine/src/builtins/uri/mod.rs

@@ -309,7 +309,7 @@ where
         }
 
         // b. Let C be the code unit at index k within string.
-        let c = string.get_expect(k);
+        let c = string.code_unit_at(k).expect("Bounds were verified");
 
         // c. If C is in unescapedSet, then
         if unescaped_set(c) {
@@ -384,7 +384,7 @@ where
         }
 
         // b. Let C be the code unit at index k within string.
-        let c = string.get_expect(k);
+        let c = string.code_point_at(k).as_u32() as u16;
 
         // c. If C is not the code unit 0x0025 (PERCENT SIGN), then
         #[allow(clippy::if_not_else)]
@@ -406,10 +406,17 @@ where
             // iii. If the code units at index (k + 1) and (k + 2) within string do not represent
             // hexadecimal digits, throw a URIError exception.
             // iv. Let B be the 8-bit value represented by the two hexadecimal digits at index (k + 1) and (k + 2).
-            let b = decode_hex_byte(string.get_expect(k + 1), string.get_expect(k + 2))
-                .ok_or_else(|| {
-                    JsNativeError::uri().with_message("invalid hexadecimal digit found")
-                })?;
+
+            // SAFETY: the indices have been verified as valid already.
+            let (high, low) = unsafe {
+                (
+                    string.code_unit_at(k + 1).unwrap_unchecked(),
+                    string.code_unit_at(k + 2).unwrap_unchecked(),
+                )
+            };
+            let b = decode_hex_byte(high, low).ok_or_else(|| {
+                JsNativeError::uri().with_message("invalid hexadecimal digit found")
+            })?;
 
             // v. Set k to k + 2.
             k += 2;
@@ -457,18 +464,24 @@ where
                     k += 1;
 
                     // b. If the code unit at index k within string is not the code unit 0x0025 (PERCENT SIGN), throw a URIError exception.
-                    if string.get_expect(k) != 0x0025 {
+                    if string.code_unit_at(k) != Some(0x0025) {
                         return Err(JsNativeError::uri()
                             .with_message("escape characters must be preceded with a % sign")
                             .into());
                     }
 
                     // c. If the code units at index (k + 1) and (k + 2) within string do not represent hexadecimal digits, throw a URIError exception.
                     // d. Let B be the 8-bit value represented by the two hexadecimal digits at index (k + 1) and (k + 2).
-                    let b = decode_hex_byte(string.get_expect(k + 1), string.get_expect(k + 2))
-                        .ok_or_else(|| {
-                            JsNativeError::uri().with_message("invalid hexadecimal digit found")
-                        })?;
+                    // SAFETY: the indices have been verified as valid already.
+                    let (high, low) = unsafe {
+                        (
+                            string.code_unit_at(k + 1).unwrap_unchecked(),
+                            string.code_unit_at(k + 2).unwrap_unchecked(),
+                        )
+                    };
+                    let b = decode_hex_byte(high, low).ok_or_else(|| {
+                        JsNativeError::uri().with_message("invalid hexadecimal digit found")
+                    })?;
 
                     // e. Set k to k + 2.
                     k += 2;

core/engine/src/vm/opcode/get/property.rs

@@ -114,7 +114,7 @@ fn get_by_value<const PUSH_KEY: bool>(
                 }
             } else if let Some(string) = base.as_string() {
                 let value = string
-                    .get(index.get() as usize)
+                    .code_unit_at(index.get() as usize)
                     .map_or_else(JsValue::undefined, |char| {
                         js_string!([char].as_slice()).into()
                     });