feat(ecmascript): add URI encoding/decoding support to constant evaluation (#12934)

crates/oxc_ecmascript/src/constant_evaluation/call_expr.rs

@@ -16,16 +16,49 @@ use cow_utils::CowUtils;
 
 use crate::{
     StringCharAt, StringCharAtResult, StringCharCodeAt, StringIndexOf, StringLastIndexOf,
-    StringSubstring, ToInt32, ToJsString as ToJsStringTrait, side_effects::MayHaveSideEffects,
+    StringSubstring, ToInt32, ToJsString as ToJsStringTrait,
+    constant_evaluation::url_encoding::{
+        decode_uri_chars, encode_uri_chars, is_uri_always_unescaped,
+    },
+    side_effects::MayHaveSideEffects,
 };
 
 use super::{ConstantEvaluation, ConstantEvaluationCtx, ConstantValue};
 
+fn try_fold_url_related_function<'a>(
+    ident: &IdentifierReference<'a>,
+    arguments: &Vec<'a, Argument<'a>>,
+    ctx: &impl ConstantEvaluationCtx<'a>,
+) -> Option<ConstantValue<'a>> {
+    match ident.name.as_str() {
+        "encodeURI" if ctx.is_global_reference(ident) == Some(true) => {
+            try_fold_encode_uri(arguments, ctx)
+        }
+        "encodeURIComponent" if ctx.is_global_reference(ident) == Some(true) => {
+            try_fold_encode_uri_component(arguments, ctx)
+        }
+        "decodeURI" if ctx.is_global_reference(ident) == Some(true) => {
+            try_fold_decode_uri(arguments, ctx)
+        }
+        "decodeURIComponent" if ctx.is_global_reference(ident) == Some(true) => {
+            try_fold_decode_uri_component(arguments, ctx)
+        }
+        _ => None,
+    }
+}
+
 pub fn try_fold_known_global_methods<'a>(
     callee: &Expression<'a>,
     arguments: &Vec<'a, Argument<'a>>,
     ctx: &impl ConstantEvaluationCtx<'a>,
 ) -> Option<ConstantValue<'a>> {
+    if let Expression::Identifier(ident) = callee {
+        if let Some(result) = try_fold_url_related_function(ident, arguments, ctx) {
+            return Some(result);
+        }
+        return None;
+    }
+
     let (name, object) = match callee {
         Expression::StaticMemberExpression(member) if !member.optional => {
             (member.property.name.as_str(), &member.object)
@@ -489,3 +522,102 @@ fn try_fold_math_variadic<'a>(
     };
     Some(ConstantValue::Number(result))
 }
+
+fn try_fold_encode_uri<'a>(
+    args: &Vec<'a, Argument<'a>>,
+    ctx: &impl ConstantEvaluationCtx<'a>,
+) -> Option<ConstantValue<'a>> {
+    if args.is_empty() {
+        return Some(ConstantValue::String(Cow::Borrowed("undefined")));
+    }
+    if args.len() != 1 {
+        return None;
+    }
+    let arg = args.first()?;
+    let expr = arg.as_expression()?;
+    let string_value = expr.get_side_free_string_value(ctx)?;
+
+    // SAFETY: should_encode only returns false for ascii chars
+    let encoded = unsafe {
+        encode_uri_chars(
+            string_value,
+            #[inline(always)]
+            |c| match c {
+                c if is_uri_always_unescaped(c) => false,
+                b';' | b'/' | b'?' | b':' | b'@' | b'&' | b'=' | b'+' | b'$' | b',' | b'#' => false,
+                _ => true,
+            },
+        )
+    };
+    Some(ConstantValue::String(encoded))
+}
+
+fn try_fold_encode_uri_component<'a>(
+    args: &Vec<'a, Argument<'a>>,
+    ctx: &impl ConstantEvaluationCtx<'a>,
+) -> Option<ConstantValue<'a>> {
+    if args.is_empty() {
+        return Some(ConstantValue::String(Cow::Borrowed("undefined")));
+    }
+    if args.len() != 1 {
+        return None;
+    }
+    let arg = args.first()?;
+    let expr = arg.as_expression()?;
+    let string_value = expr.get_side_free_string_value(ctx)?;
+
+    // SAFETY: should_encode only returns false for ascii chars
+    let encoded = unsafe {
+        encode_uri_chars(
+            string_value,
+            #[inline(always)]
+            |c| !is_uri_always_unescaped(c),
+        )
+    };
+    Some(ConstantValue::String(encoded))
+}
+
+fn try_fold_decode_uri<'a>(
+    args: &Vec<'a, Argument<'a>>,
+    ctx: &impl ConstantEvaluationCtx<'a>,
+) -> Option<ConstantValue<'a>> {
+    if args.is_empty() {
+        return Some(ConstantValue::String(Cow::Borrowed("undefined")));
+    }
+    if args.len() != 1 {
+        return None;
+    }
+    let arg = args.first()?;
+    let expr = arg.as_expression()?;
+    let string_value = expr.get_side_free_string_value(ctx)?;
+
+    let decoded = decode_uri_chars(
+        string_value,
+        #[inline(always)]
+        |c| matches!(c, b';' | b',' | b'/' | b'?' | b':' | b'@' | b'&' | b'=' | b'+' | b'$' | b'#'),
+    )?;
+    Some(ConstantValue::String(decoded))
+}
+
+fn try_fold_decode_uri_component<'a>(
+    args: &Vec<'a, Argument<'a>>,
+    ctx: &impl ConstantEvaluationCtx<'a>,
+) -> Option<ConstantValue<'a>> {
+    if args.is_empty() {
+        return Some(ConstantValue::String(Cow::Borrowed("undefined")));
+    }
+    if args.len() != 1 {
+        return None;
+    }
+    let arg = args.first()?;
+    let expr = arg.as_expression()?;
+    let string_value = expr.get_side_free_string_value(ctx)?;
+
+    // decodeURIComponent decodes all percent-encoded sequences
+    let decoded = decode_uri_chars(
+        string_value,
+        #[inline(always)]
+        |_| false,
+    )?;
+    Some(ConstantValue::String(decoded))
+}

crates/oxc_ecmascript/src/constant_evaluation/mod.rs

@@ -2,6 +2,7 @@ mod call_expr;
 mod equality_comparison;
 mod is_int32_or_uint32;
 mod is_literal_value;
+mod url_encoding;
 mod value;
 mod value_type;
 

crates/oxc_ecmascript/src/constant_evaluation/url_encoding/dec.rs

@@ -0,0 +1,104 @@
+// based on https://github.com/kornelski/rust_urlencoding/blob/a617c89d16f390e3ab4281ea68c514660b111301/src/dec.rs#L21
+// MIT license: https://github.com/kornelski/rust_urlencoding/blob/a617c89d16f390e3ab4281ea68c514660b111301/LICENSE
+
+use std::borrow::Cow;
+use std::panic::panic_any;
+
+/// Implements <https://tc39.es/ecma262/2025/multipage/global-object.html#sec-decode>
+#[inline]
+pub fn decode(
+    data_str: Cow<'_, str>,
+    should_not_decode: impl Fn(u8) -> bool,
+) -> Option<Cow<'_, str>> {
+    let data = data_str.as_bytes();
+    let offset = data.iter().take_while(|&&c| c != b'%').count();
+    if offset >= data.len() {
+        return Some(data_str);
+    }
+
+    let mut decoded = Vec::new();
+    if decoded.try_reserve(data.len()).is_err() {
+        panic_any("OOM"); // more efficient codegen than built-in OOM handler
+    }
+    let mut out = NeverRealloc(&mut decoded);
+
+    let (ascii, mut data) = data.split_at(offset);
+    out.extend_from_slice(ascii);
+
+    loop {
+        let mut parts = data.splitn(2, |&c| c == b'%');
+        // first the decoded non-% part
+        let non_escaped_part = parts.next().unwrap();
+        let rest = parts.next();
+        if rest.is_none() && out.0.is_empty() {
+            // if empty there were no '%' in the string
+            return Some(data_str);
+        }
+        out.extend_from_slice(non_escaped_part);
+
+        // then decode one %xx
+        match rest {
+            Some(rest) => {
+                let Some(&[first, second]) = rest.get(0..2) else {
+                    // 4.c.i.
+                    return None;
+                };
+                let (Some(first_val), Some(second_val)) =
+                    (from_hex_digit(first), from_hex_digit(second))
+                else {
+                    // 4.c.iii.
+                    return None;
+                };
+                let char = (first_val << 4) | second_val;
+                if should_not_decode(char) {
+                    out.extend_from_slice(&[b'%', first, second]);
+                } else {
+                    out.push(char);
+                }
+                data = &rest[2..];
+            }
+            None => break,
+        }
+    }
+    Some(Cow::Owned(String::from_utf8(decoded).ok()?))
+}
+
+#[inline]
+fn from_hex_digit(digit: u8) -> Option<u8> {
+    match digit {
+        b'0'..=b'9' => Some(digit - b'0'),
+        b'A'..=b'F' => Some(digit - b'A' + 10),
+        b'a'..=b'f' => Some(digit - b'a' + 10),
+        _ => None,
+    }
+}
+
+struct NeverRealloc<'a, T>(pub &'a mut Vec<T>);
+
+impl<T> NeverRealloc<'_, T> {
+    #[inline]
+    pub fn push(&mut self, val: T) {
+        // these branches only exist to remove redundant reallocation code
+        // (the capacity is always sufficient)
+        if self.0.len() != self.0.capacity() {
+            self.0.push(val);
+        }
+    }
+
+    #[inline]
+    pub fn extend_from_slice(&mut self, val: &[T])
+    where
+        T: Clone,
+    {
+        if self.0.capacity() - self.0.len() >= val.len() {
+            self.0.extend_from_slice(val);
+        }
+    }
+}
+
+#[test]
+fn dec_borrows() {
+    assert!(matches!(decode("hello".into(), |_| false), Some(Cow::Borrowed("hello"))));
+    assert!(matches!(decode("hello%20".into(), |_| false), Some(Cow::Owned(s)) if s == "hello "));
+    assert!(matches!(decode("%20hello".into(), |_| false), Some(Cow::Owned(s)) if s == " hello"));
+}

crates/oxc_ecmascript/src/constant_evaluation/url_encoding/enc.rs

@@ -0,0 +1,79 @@
+// Based on https://github.com/kornelski/rust_urlencoding/blob/a617c89d16f390e3ab4281ea68c514660b111301/src/enc.rs
+// MIT license: https://github.com/kornelski/rust_urlencoding/blob/a617c89d16f390e3ab4281ea68c514660b111301/LICENSE
+
+use std::borrow::Cow;
+
+/// Implements <https://tc39.es/ecma262/2025/multipage/global-object.html#sec-encode>
+/// # Safety
+/// `should_encode` should only return false for characters that are ascii
+#[must_use]
+pub unsafe fn encode(data_str: Cow<'_, str>, should_encode: impl Fn(u8) -> bool) -> Cow<'_, str> {
+    let data = data_str.as_bytes();
+    // add maybe extra capacity, but try not to exceed allocator's bucket size
+    let mut escaped = String::new();
+    let _ = escaped.try_reserve(data.len() | 15);
+    let unmodified = encode_into(data, should_encode, |s| {
+        escaped.push_str(s);
+    });
+    if unmodified {
+        return data_str;
+    }
+    Cow::Owned(escaped)
+}
+
+fn encode_into(
+    mut data: &[u8],
+    should_encode: impl Fn(u8) -> bool,
+    mut push_str: impl FnMut(&str),
+) -> bool {
+    let mut pushed = false;
+    loop {
+        // Fast path to skip over safe chars at the beginning of the remaining string
+        let ascii_len = data.iter().take_while(|&&c| !should_encode(c)).count();
+
+        let (safe, rest) = if ascii_len >= data.len() {
+            if !pushed {
+                return true;
+            }
+            (data, &[][..]) // redundant to optimize out a panic in split_at
+        } else {
+            data.split_at(ascii_len)
+        };
+        pushed = true;
+        if !safe.is_empty() {
+            // SAFETY: should_encode has checked it's ASCII
+            push_str(unsafe { str::from_utf8_unchecked(safe) });
+        }
+        if rest.is_empty() {
+            break;
+        }
+
+        match rest.split_first() {
+            Some((byte, rest)) => {
+                let enc = &[b'%', to_hex_digit(byte >> 4), to_hex_digit(byte & 15)];
+                // SAFETY: `%` is a valid UTF-8 char and to_hex_digit returns a valid UTF-8 char
+                push_str(unsafe { str::from_utf8_unchecked(enc) });
+                data = rest;
+            }
+            None => break,
+        }
+    }
+    false
+}
+
+#[inline]
+fn to_hex_digit(digit: u8) -> u8 {
+    match digit {
+        0..=9 => b'0' + digit,
+        10..=255 => b'A' - 10 + digit,
+    }
+}
+
+/// `alwaysUnescaped` in `Encode`
+/// <https://tc39.es/ecma262/2025/multipage/global-object.html#sec-encode>
+const URI_ALWAYS_UNESCAPED: &[u8] =
+    b"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789_-.!~*'()";
+
+pub fn is_uri_always_unescaped(c: u8) -> bool {
+    URI_ALWAYS_UNESCAPED.contains(&c)
+}

crates/oxc_ecmascript/src/constant_evaluation/url_encoding/mod.rs

@@ -0,0 +1,5 @@
+mod dec;
+mod enc;
+
+pub use dec::decode as decode_uri_chars;
+pub use enc::{encode as encode_uri_chars, is_uri_always_unescaped};

crates/oxc_minifier/src/peephole/replace_known_methods.rs

@@ -1624,4 +1624,102 @@ mod test {
             "const node_env = 'production'; v = !0",
         );
     }
+
+    #[test]
+    fn test_fold_encode_uri() {
+        test("x = encodeURI()", "x = 'undefined'");
+        test("x = encodeURI('hello')", "x = 'hello'");
+        test("x = encodeURI('hello world')", "x = 'hello%20world'");
+        test(
+            "x = encodeURI('http://example.com/path?a=1&b=2#hash')",
+            "x = 'http://example.com/path?a=1&b=2#hash'",
+        );
+        test("x = encodeURI('a;b,c/d?e:f@g&h=i+j$k')", "x = 'a;b,c/d?e:f@g&h=i+j$k'");
+        test("x = encodeURI('ABC-_abc.!~*()123')", "x = 'ABC-_abc.!~*()123'");
+        test("x = encodeURI('hello<>\"')", "x = 'hello%3C%3E%22'");
+        test("x = encodeURI('hello\\t\\n')", "x = 'hello%09%0A'");
+        test("x = encodeURI('café')", "x = 'caf%C3%A9'"); // spellchecker:disable-line
+        test("x = encodeURI('测试')", "x = '%E6%B5%8B%E8%AF%95'");
+
+        test_same("encodeURI('a', 'b')");
+        test_same("encodeURI(x)");
+    }
+
+    #[test]
+    fn test_fold_encode_uri_component() {
+        test("x = encodeURIComponent()", "x = 'undefined'");
+        test("x = encodeURIComponent('hello')", "x = 'hello'");
+        test("x = encodeURIComponent('ABC-_abc.!~*()123')", "x = 'ABC-_abc.!~*()123'");
+        test(
+            "x = encodeURIComponent('a;b,c/d?e:f@g&h=i+j$k')",
+            "x = 'a%3Bb%2Cc%2Fd%3Fe%3Af%40g%26h%3Di%2Bj%24k'",
+        );
+        test("x = encodeURIComponent('#')", "x = '%23'");
+        test("x = encodeURIComponent('hello world')", "x = 'hello%20world'");
+        test("x = encodeURIComponent('hello<>\"')", "x = 'hello%3C%3E%22'");
+        test("x = encodeURIComponent('café')", "x = 'caf%C3%A9'"); // spellchecker:disable-line
+        test("x = encodeURIComponent('测试')", "x = '%E6%B5%8B%E8%AF%95'");
+
+        test_same("encodeURIComponent('a', 'b')");
+        test_same("encodeURIComponent(x)");
+    }
+
+    #[test]
+    fn test_fold_decode_uri() {
+        test("x = decodeURI()", "x = 'undefined'");
+        test("x = decodeURI('hello%20world')", "x = 'hello world'");
+        test("x = decodeURI('hello')", "x = 'hello'");
+        test(
+            "x = decodeURI('a%3Bb%2Cc%2Fd%3Fe%3Af%40g%26h%3Di%2Bj%24k')",
+            "x = 'a%3Bb%2Cc%2Fd%3Fe%3Af%40g%26h%3Di%2Bj%24k'",
+        );
+        test("x = decodeURI('%2f')", "x = '%2f'"); // `/`, lower case
+        test("x = decodeURI('%23')", "x = '%23'"); // `#`
+        test("x = decodeURI('%23hash')", "x = '%23hash'");
+        test("x = decodeURI('hello%3C%3E%22')", "x = 'hello<>\"'");
+        test("x = decodeURI('hello%09%0A')", "x = 'hello\\t\\n'");
+        test("x = decodeURI('caf%C3%A9')", "x = 'café'"); // spellchecker:disable-line
+        test("x = decodeURI('%E6%B5%8B%E8%AF%95')", "x = '测试'");
+
+        test_same("decodeURI('%ZZ')"); // URIError
+        test_same("decodeURI('%A')"); // URIError
+
+        test_same("decodeURI('a', 'b')");
+        test_same("decodeURI(x)");
+    }
+
+    #[test]
+    fn test_fold_decode_uri_component() {
+        test("x = decodeURIComponent()", "x = 'undefined'");
+        test("x = decodeURIComponent('hello%20world')", "x = 'hello world'");
+        test("x = decodeURIComponent('hello')", "x = 'hello'");
+        test(
+            "x = decodeURIComponent('a%3Bb%2Cc%2Fd%3Fe%3Af%40g%26h%3Di%2Bj%24k')",
+            "x = 'a;b,c/d?e:f@g&h=i+j$k'",
+        );
+        test("x = decodeURIComponent('%23')", "x = '#'");
+        test("x = decodeURIComponent('%23hash')", "x = '#hash'");
+        test("x = decodeURIComponent('hello%3C%3E%22')", "x = 'hello<>\"'");
+        test("x = decodeURIComponent('hello%09%0A')", "x = 'hello\\t\\n'");
+        test("x = decodeURIComponent('caf%C3%A9')", "x = 'café'"); // spellchecker:disable-line
+        test("x = decodeURIComponent('%E6%B5%8B%E8%AF%95')", "x = '测试'");
+
+        test_same("decodeURIComponent('%ZZ')"); // URIError
+        test_same("decodeURIComponent('%A')"); // URIError
+
+        test_same("decodeURIComponent('a', 'b')");
+        test_same("decodeURIComponent(x)");
+    }
+
+    #[test]
+    fn test_fold_uri_roundtrip() {
+        test("x = decodeURI(encodeURI('hello world'))", "x = 'hello world'");
+        test("x = decodeURIComponent(encodeURIComponent('hello world'))", "x = 'hello world'");
+        test(
+            "x = decodeURIComponent(encodeURIComponent('a;b,c/d?e:f@g&h=i+j$k'))",
+            "x = 'a;b,c/d?e:f@g&h=i+j$k'",
+        );
+        test("x = decodeURI(encodeURI('café'))", "x = 'café'");
+        test("x = decodeURIComponent(encodeURIComponent('测试'))", "x = '测试'");
+    }
 }