url/

parser.rs

// Copyright 2013-2016 The rust-url developers.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.

use alloc::borrow::Cow;
use alloc::string::String;
use core::fmt::{self, Formatter, Write};
use core::str;

use crate::host::{Host, HostInternal};
use crate::Url;
use form_urlencoded::EncodingOverride;
use percent_encoding::{percent_encode, utf8_percent_encode, AsciiSet, CONTROLS};

/// https://url.spec.whatwg.org/#fragment-percent-encode-set
const FRAGMENT: &AsciiSet = &CONTROLS.add(b' ').add(b'"').add(b'<').add(b'>').add(b'`');

/// https://url.spec.whatwg.org/#path-percent-encode-set
const PATH: &AsciiSet = &FRAGMENT.add(b'#').add(b'?').add(b'{').add(b'}');

/// https://url.spec.whatwg.org/#userinfo-percent-encode-set
pub(crate) const USERINFO: &AsciiSet = &PATH
    .add(b'/')
    .add(b':')
    .add(b';')
    .add(b'=')
    .add(b'@')
    .add(b'[')
    .add(b'\\')
    .add(b']')
    .add(b'^')
    .add(b'|');

pub(crate) const PATH_SEGMENT: &AsciiSet = &PATH.add(b'/').add(b'%');

// The backslash (\) character is treated as a path separator in special URLs
// so it needs to be additionally escaped in that case.
pub(crate) const SPECIAL_PATH_SEGMENT: &AsciiSet = &PATH_SEGMENT.add(b'\\');

// https://url.spec.whatwg.org/#query-state
const QUERY: &AsciiSet = &CONTROLS.add(b' ').add(b'"').add(b'#').add(b'<').add(b'>');
const SPECIAL_QUERY: &AsciiSet = &QUERY.add(b'\'');

pub type ParseResult<T> = Result<T, ParseError>;

macro_rules! simple_enum_error {
    ($($name: ident => $description: expr,)+) => {
        /// Errors that can occur during parsing.
        ///
        /// This may be extended in the future so exhaustive matching is
        /// discouraged with an unused variant.
        #[derive(PartialEq, Eq, Clone, Copy, Debug)]
        #[non_exhaustive]
        pub enum ParseError {
            $(
                $name,
            )+
        }

        impl fmt::Display for ParseError {
            fn fmt(&self, fmt: &mut Formatter<'_>) -> fmt::Result {
                match *self {
                    $(
                        ParseError::$name => fmt.write_str($description),
                    )+
                }
            }
        }
    }
}

macro_rules! ascii_tab_or_new_line_pattern {
    () => {
        '\t' | '\n' | '\r'
    };
}

#[cfg(feature = "std")]
impl std::error::Error for ParseError {}

#[cfg(not(feature = "std"))]
impl core::error::Error for ParseError {}

simple_enum_error! {
    EmptyHost => "empty host",
    IdnaError => "invalid international domain name",
    InvalidPort => "invalid port number",
    InvalidIpv4Address => "invalid IPv4 address",
    InvalidIpv6Address => "invalid IPv6 address",
    InvalidDomainCharacter => "invalid domain character",
    RelativeUrlWithoutBase => "relative URL without a base",
    RelativeUrlWithCannotBeABaseBase => "relative URL with a cannot-be-a-base base",
    SetHostOnCannotBeABaseUrl => "a cannot-be-a-base URL doesn’t have a host to set",
    Overflow => "URLs more than 4 GB are not supported",
}

impl From<::idna::Errors> for ParseError {
    fn from(_: ::idna::Errors) -> Self {
        Self::IdnaError
    }
}

macro_rules! syntax_violation_enum {
    ($($name: ident => $description: literal,)+) => {
        /// Non-fatal syntax violations that can occur during parsing.
        ///
        /// This may be extended in the future so exhaustive matching is
        /// forbidden.
        #[derive(PartialEq, Eq, Clone, Copy, Debug)]
        #[non_exhaustive]
        pub enum SyntaxViolation {
            $(
                /// ```text
                #[doc = $description]
                /// ```
                $name,
            )+
        }

        impl SyntaxViolation {
            pub fn description(&self) -> &'static str {
                match *self {
                    $(
                        SyntaxViolation::$name => $description,
                    )+
                }
            }
        }
    }
}

syntax_violation_enum! {
    Backslash => "backslash",
    C0SpaceIgnored =>
        "leading or trailing control or space character are ignored in URLs",
    EmbeddedCredentials =>
        "embedding authentication information (username or password) \
         in an URL is not recommended",
    ExpectedDoubleSlash => "expected //",
    ExpectedFileDoubleSlash => "expected // after file:",
    FileWithHostAndWindowsDrive => "file: with host and Windows drive letter",
    NonUrlCodePoint => "non-URL code point",
    NullInFragment => "NULL characters are ignored in URL fragment identifiers",
    PercentDecode => "expected 2 hex digits after %",
    TabOrNewlineIgnored => "tabs or newlines are ignored in URLs",
    UnencodedAtSign => "unencoded @ sign in username or password",
}

impl fmt::Display for SyntaxViolation {
    fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result {
        fmt::Display::fmt(self.description(), f)
    }
}

#[derive(Copy, Clone, PartialEq, Eq)]
pub enum SchemeType {
    File,
    SpecialNotFile,
    NotSpecial,
}

impl SchemeType {
    pub fn is_special(&self) -> bool {
        !matches!(*self, Self::NotSpecial)
    }

    pub fn is_file(&self) -> bool {
        matches!(*self, Self::File)
    }
}

impl<T: AsRef<str>> From<T> for SchemeType {
    fn from(s: T) -> Self {
        match s.as_ref() {
            "http" | "https" | "ws" | "wss" | "ftp" => Self::SpecialNotFile,
            "file" => Self::File,
            _ => Self::NotSpecial,
        }
    }
}

pub fn default_port(scheme: &str) -> Option<u16> {
    match scheme {
        "http" | "ws" => Some(80),
        "https" | "wss" => Some(443),
        "ftp" => Some(21),
        _ => None,
    }
}

#[derive(Clone, Debug)]
pub struct Input<'i> {
    chars: str::Chars<'i>,
}

impl<'i> Input<'i> {
    pub fn new_no_trim(input: &'i str) -> Self {
        Input {
            chars: input.chars(),
        }
    }

    pub fn new_trim_tab_and_newlines(
        original_input: &'i str,
        vfn: Option<&dyn Fn(SyntaxViolation)>,
    ) -> Self {
        let input = original_input.trim_matches(ascii_tab_or_new_line);
        if let Some(vfn) = vfn {
            if input.len() < original_input.len() {
                vfn(SyntaxViolation::C0SpaceIgnored)
            }
            if input.chars().any(ascii_tab_or_new_line) {
                vfn(SyntaxViolation::TabOrNewlineIgnored)
            }
        }
        Input {
            chars: input.chars(),
        }
    }

    pub fn new_trim_c0_control_and_space(
        original_input: &'i str,
        vfn: Option<&dyn Fn(SyntaxViolation)>,
    ) -> Self {
        let input = original_input.trim_matches(c0_control_or_space);
        if let Some(vfn) = vfn {
            if input.len() < original_input.len() {
                vfn(SyntaxViolation::C0SpaceIgnored)
            }
            if input.chars().any(ascii_tab_or_new_line) {
                vfn(SyntaxViolation::TabOrNewlineIgnored)
            }
        }
        Input {
            chars: input.chars(),
        }
    }

    #[inline]
    pub fn is_empty(&self) -> bool {
        self.clone().next().is_none()
    }

    #[inline]
    pub fn starts_with<P: Pattern>(&self, p: P) -> bool {
        p.split_prefix(&mut self.clone())
    }

    #[inline]
    pub fn split_prefix<P: Pattern>(&self, p: P) -> Option<Self> {
        let mut remaining = self.clone();
        if p.split_prefix(&mut remaining) {
            Some(remaining)
        } else {
            None
        }
    }

    #[inline]
    fn split_first(&self) -> (Option<char>, Self) {
        let mut remaining = self.clone();
        (remaining.next(), remaining)
    }

    #[inline]
    fn count_matching<F: Fn(char) -> bool>(&self, f: F) -> (u32, Self) {
        let mut count = 0;
        let mut remaining = self.clone();
        loop {
            let mut input = remaining.clone();
            if matches!(input.next(), Some(c) if f(c)) {
                remaining = input;
                count += 1;
            } else {
                return (count, remaining);
            }
        }
    }

    #[inline]
    fn next_utf8(&mut self) -> Option<(char, &'i str)> {
        loop {
            let utf8 = self.chars.as_str();
            match self.chars.next() {
                Some(c) => {
                    if !ascii_tab_or_new_line(c) {
                        return Some((c, &utf8[..c.len_utf8()]));
                    }
                }
                None => return None,
            }
        }
    }
}

pub trait Pattern {
    fn split_prefix(self, input: &mut Input) -> bool;
}

impl Pattern for char {
    fn split_prefix(self, input: &mut Input) -> bool {
        input.next() == Some(self)
    }
}

impl Pattern for &str {
    fn split_prefix(self, input: &mut Input) -> bool {
        for c in self.chars() {
            if input.next() != Some(c) {
                return false;
            }
        }
        true
    }
}

impl<F: FnMut(char) -> bool> Pattern for F {
    fn split_prefix(self, input: &mut Input) -> bool {
        input.next().map_or(false, self)
    }
}

impl Iterator for Input<'_> {
    type Item = char;
    fn next(&mut self) -> Option<char> {
        self.chars.by_ref().find(|&c| !ascii_tab_or_new_line(c))
    }

    fn size_hint(&self) -> (usize, Option<usize>) {
        (0, Some(self.chars.as_str().len()))
    }
}

pub struct Parser<'a> {
    pub serialization: String,
    pub base_url: Option<&'a Url>,
    pub query_encoding_override: EncodingOverride<'a>,
    pub violation_fn: Option<&'a dyn Fn(SyntaxViolation)>,
    pub context: Context,
}

#[derive(PartialEq, Eq, Copy, Clone)]
pub enum Context {
    UrlParser,
    Setter,
    PathSegmentSetter,
}

impl Parser<'_> {
    fn log_violation(&self, v: SyntaxViolation) {
        if let Some(f) = self.violation_fn {
            f(v)
        }
    }

    fn log_violation_if(&self, v: SyntaxViolation, test: impl FnOnce() -> bool) {
        if let Some(f) = self.violation_fn {
            if test() {
                f(v)
            }
        }
    }

    pub fn for_setter(serialization: String) -> Self {
        Parser {
            serialization,
            base_url: None,
            query_encoding_override: None,
            violation_fn: None,
            context: Context::Setter,
        }
    }

    /// https://url.spec.whatwg.org/#concept-basic-url-parser
    pub fn parse_url(mut self, input: &str) -> ParseResult<Url> {
        let input = Input::new_trim_c0_control_and_space(input, self.violation_fn);
        if let Ok(remaining) = self.parse_scheme(input.clone()) {
            return self.parse_with_scheme(remaining);
        }

        // No-scheme state
        if let Some(base_url) = self.base_url {
            if input.starts_with('#') {
                self.fragment_only(base_url, input)
            } else if base_url.cannot_be_a_base() {
                Err(ParseError::RelativeUrlWithCannotBeABaseBase)
            } else {
                let scheme_type = SchemeType::from(base_url.scheme());
                if scheme_type.is_file() {
                    self.parse_file(input, scheme_type, Some(base_url))
                } else {
                    self.parse_relative(input, scheme_type, base_url)
                }
            }
        } else {
            Err(ParseError::RelativeUrlWithoutBase)
        }
    }

    pub fn parse_scheme<'i>(&mut self, mut input: Input<'i>) -> Result<Input<'i>, ()> {
        // starts_with will also fail for empty strings so we can skip that comparison for perf
        if !input.starts_with(ascii_alpha) {
            return Err(());
        }
        debug_assert!(self.serialization.is_empty());
        while let Some(c) = input.next() {
            match c {
                'a'..='z' | '0'..='9' | '+' | '-' | '.' => self.serialization.push(c),
                'A'..='Z' => self.serialization.push(c.to_ascii_lowercase()),
                ':' => return Ok(input),
                _ => {
                    self.serialization.clear();
                    return Err(());
                }
            }
        }
        // EOF before ':'
        if self.context == Context::Setter {
            Ok(input)
        } else {
            self.serialization.clear();
            Err(())
        }
    }

    fn parse_with_scheme(mut self, input: Input<'_>) -> ParseResult<Url> {
        use crate::SyntaxViolation::{ExpectedDoubleSlash, ExpectedFileDoubleSlash};
        let scheme_end = to_u32(self.serialization.len())?;
        let scheme_type = SchemeType::from(&self.serialization);
        self.serialization.push(':');
        match scheme_type {
            SchemeType::File => {
                self.log_violation_if(ExpectedFileDoubleSlash, || !input.starts_with("//"));
                let base_file_url = self.base_url.and_then(|base| {
                    if base.scheme() == "file" {
                        Some(base)
                    } else {
                        None
                    }
                });
                self.serialization.clear();
                self.parse_file(input, scheme_type, base_file_url)
            }
            SchemeType::SpecialNotFile => {
                // special relative or authority state
                let (slashes_count, remaining) = input.count_matching(|c| matches!(c, '/' | '\\'));
                if let Some(base_url) = self.base_url {
                    if slashes_count < 2
                        && base_url.scheme() == &self.serialization[..scheme_end as usize]
                    {
                        // "Cannot-be-a-base" URLs only happen with "not special" schemes.
                        debug_assert!(!base_url.cannot_be_a_base());
                        self.serialization.clear();
                        return self.parse_relative(input, scheme_type, base_url);
                    }
                }
                // special authority slashes state
                self.log_violation_if(ExpectedDoubleSlash, || {
                    input
                        .clone()
                        .take_while(|&c| matches!(c, '/' | '\\'))
                        .collect::<String>()
                        != "//"
                });
                self.after_double_slash(remaining, scheme_type, scheme_end)
            }
            SchemeType::NotSpecial => self.parse_non_special(input, scheme_type, scheme_end),
        }
    }

    /// Scheme other than file, http, https, ws, ws, ftp.
    fn parse_non_special(
        mut self,
        input: Input<'_>,
        scheme_type: SchemeType,
        scheme_end: u32,
    ) -> ParseResult<Url> {
        // path or authority state (
        if let Some(input) = input.split_prefix("//") {
            return self.after_double_slash(input, scheme_type, scheme_end);
        }
        // Anarchist URL (no authority)
        let path_start = to_u32(self.serialization.len())?;
        let username_end = path_start;
        let host_start = path_start;
        let host_end = path_start;
        let host = HostInternal::None;
        let port = None;
        let remaining = if let Some(input) = input.split_prefix('/') {
            self.serialization.push('/');
            self.parse_path(scheme_type, &mut false, path_start as usize, input)
        } else {
            self.parse_cannot_be_a_base_path(input)
        };
        self.with_query_and_fragment(
            scheme_type,
            scheme_end,
            username_end,
            host_start,
            host_end,
            host,
            port,
            path_start,
            remaining,
        )
    }

    fn parse_file(
        mut self,
        input: Input<'_>,
        scheme_type: SchemeType,
        base_file_url: Option<&Url>,
    ) -> ParseResult<Url> {
        use crate::SyntaxViolation::Backslash;
        // file state
        debug_assert!(self.serialization.is_empty());
        let (first_char, input_after_first_char) = input.split_first();
        if matches!(first_char, Some('/') | Some('\\')) {
            self.log_violation_if(SyntaxViolation::Backslash, || first_char == Some('\\'));
            // file slash state
            let (next_char, input_after_next_char) = input_after_first_char.split_first();
            if matches!(next_char, Some('/') | Some('\\')) {
                self.log_violation_if(Backslash, || next_char == Some('\\'));
                // file host state
                self.serialization.push_str("file://");
                let scheme_end = "file".len() as u32;
                let host_start = "file://".len() as u32;
                let (path_start, mut host, remaining) =
                    self.parse_file_host(input_after_next_char)?;
                let mut host_end = to_u32(self.serialization.len())?;
                let mut has_host = !matches!(host, HostInternal::None);
                let remaining = if path_start {
                    self.parse_path_start(SchemeType::File, &mut has_host, remaining)
                } else {
                    let path_start = self.serialization.len();
                    self.serialization.push('/');
                    self.parse_path(SchemeType::File, &mut has_host, path_start, remaining)
                };

                // For file URLs that have a host and whose path starts
                // with the windows drive letter we just remove the host.
                if !has_host {
                    self.serialization
                        .drain(host_start as usize..host_end as usize);
                    host_end = host_start;
                    host = HostInternal::None;
                }
                let (query_start, fragment_start) =
                    self.parse_query_and_fragment(scheme_type, scheme_end, remaining)?;
                return Ok(Url {
                    serialization: self.serialization,
                    scheme_end,
                    username_end: host_start,
                    host_start,
                    host_end,
                    host,
                    port: None,
                    path_start: host_end,
                    query_start,
                    fragment_start,
                });
            } else {
                self.serialization.push_str("file://");
                let scheme_end = "file".len() as u32;
                let host_start = "file://".len();
                let mut host_end = host_start;
                let mut host = HostInternal::None;
                if !starts_with_windows_drive_letter_segment(&input_after_first_char) {
                    if let Some(base_url) = base_file_url {
                        let first_segment = base_url.path_segments().unwrap().next().unwrap();
                        if is_normalized_windows_drive_letter(first_segment) {
                            self.serialization.push('/');
                            self.serialization.push_str(first_segment);
                        } else if let Some(host_str) = base_url.host_str() {
                            self.serialization.push_str(host_str);
                            host_end = self.serialization.len();
                            host = base_url.host;
                        }
                    }
                }
                // If c is the EOF code point, U+002F (/), U+005C (\), U+003F (?), or U+0023 (#), then decrease pointer by one
                let parse_path_input = if let Some(c) = first_char {
                    if c == '/' || c == '\\' || c == '?' || c == '#' {
                        input
                    } else {
                        input_after_first_char
                    }
                } else {
                    input_after_first_char
                };

                let remaining =
                    self.parse_path(SchemeType::File, &mut false, host_end, parse_path_input);

                let host_start = host_start as u32;

                let (query_start, fragment_start) =
                    self.parse_query_and_fragment(scheme_type, scheme_end, remaining)?;

                let host_end = host_end as u32;
                return Ok(Url {
                    serialization: self.serialization,
                    scheme_end,
                    username_end: host_start,
                    host_start,
                    host_end,
                    host,
                    port: None,
                    path_start: host_end,
                    query_start,
                    fragment_start,
                });
            }
        }
        if let Some(base_url) = base_file_url {
            match first_char {
                None => {
                    // Copy everything except the fragment
                    let before_fragment = match base_url.fragment_start {
                        Some(i) => &base_url.serialization[..i as usize],
                        None => &*base_url.serialization,
                    };
                    self.serialization.push_str(before_fragment);
                    Ok(Url {
                        serialization: self.serialization,
                        fragment_start: None,
                        ..*base_url
                    })
                }
                Some('?') => {
                    // Copy everything up to the query string
                    let before_query = match (base_url.query_start, base_url.fragment_start) {
                        (None, None) => &*base_url.serialization,
                        (Some(i), _) | (None, Some(i)) => base_url.slice(..i),
                    };
                    self.serialization.push_str(before_query);
                    let (query_start, fragment_start) =
                        self.parse_query_and_fragment(scheme_type, base_url.scheme_end, input)?;
                    Ok(Url {
                        serialization: self.serialization,
                        query_start,
                        fragment_start,
                        ..*base_url
                    })
                }
                Some('#') => self.fragment_only(base_url, input),
                _ => {
                    if !starts_with_windows_drive_letter_segment(&input) {
                        let before_query = match (base_url.query_start, base_url.fragment_start) {
                            (None, None) => &*base_url.serialization,
                            (Some(i), _) | (None, Some(i)) => base_url.slice(..i),
                        };
                        self.serialization.push_str(before_query);
                        self.shorten_path(SchemeType::File, base_url.path_start as usize);
                        let remaining = self.parse_path(
                            SchemeType::File,
                            &mut true,
                            base_url.path_start as usize,
                            input,
                        );
                        self.with_query_and_fragment(
                            SchemeType::File,
                            base_url.scheme_end,
                            base_url.username_end,
                            base_url.host_start,
                            base_url.host_end,
                            base_url.host,
                            base_url.port,
                            base_url.path_start,
                            remaining,
                        )
                    } else {
                        self.serialization.push_str("file:///");
                        let scheme_end = "file".len() as u32;
                        let path_start = "file://".len();
                        let remaining =
                            self.parse_path(SchemeType::File, &mut false, path_start, input);
                        let (query_start, fragment_start) =
                            self.parse_query_and_fragment(SchemeType::File, scheme_end, remaining)?;
                        let path_start = path_start as u32;
                        Ok(Url {
                            serialization: self.serialization,
                            scheme_end,
                            username_end: path_start,
                            host_start: path_start,
                            host_end: path_start,
                            host: HostInternal::None,
                            port: None,
                            path_start,
                            query_start,
                            fragment_start,
                        })
                    }
                }
            }
        } else {
            self.serialization.push_str("file:///");
            let scheme_end = "file".len() as u32;
            let path_start = "file://".len();
            let remaining = self.parse_path(SchemeType::File, &mut false, path_start, input);
            let (query_start, fragment_start) =
                self.parse_query_and_fragment(SchemeType::File, scheme_end, remaining)?;
            let path_start = path_start as u32;
            Ok(Url {
                serialization: self.serialization,
                scheme_end,
                username_end: path_start,
                host_start: path_start,
                host_end: path_start,
                host: HostInternal::None,
                port: None,
                path_start,
                query_start,
                fragment_start,
            })
        }
    }

    fn parse_relative(
        mut self,
        input: Input<'_>,
        scheme_type: SchemeType,
        base_url: &Url,
    ) -> ParseResult<Url> {
        // relative state
        debug_assert!(self.serialization.is_empty());
        let (first_char, input_after_first_char) = input.split_first();
        match first_char {
            None => {
                // Copy everything except the fragment
                let before_fragment = match base_url.fragment_start {
                    Some(i) => &base_url.serialization[..i as usize],
                    None => &*base_url.serialization,
                };
                self.serialization.push_str(before_fragment);
                Ok(Url {
                    serialization: self.serialization,
                    fragment_start: None,
                    ..*base_url
                })
            }
            Some('?') => {
                // Copy everything up to the query string
                let before_query = match (base_url.query_start, base_url.fragment_start) {
                    (None, None) => &*base_url.serialization,
                    (Some(i), _) | (None, Some(i)) => base_url.slice(..i),
                };
                self.serialization.push_str(before_query);
                let (query_start, fragment_start) =
                    self.parse_query_and_fragment(scheme_type, base_url.scheme_end, input)?;
                Ok(Url {
                    serialization: self.serialization,
                    query_start,
                    fragment_start,
                    ..*base_url
                })
            }
            Some('#') => self.fragment_only(base_url, input),
            Some('/') | Some('\\') => {
                let (slashes_count, remaining) = input.count_matching(|c| matches!(c, '/' | '\\'));
                if slashes_count >= 2 {
                    self.log_violation_if(SyntaxViolation::ExpectedDoubleSlash, || {
                        input
                            .clone()
                            .take_while(|&c| matches!(c, '/' | '\\'))
                            .collect::<String>()
                            != "//"
                    });
                    let scheme_end = base_url.scheme_end;
                    debug_assert!(base_url.byte_at(scheme_end) == b':');
                    self.serialization
                        .push_str(base_url.slice(..scheme_end + 1));
                    if let Some(after_prefix) = input.split_prefix("//") {
                        return self.after_double_slash(after_prefix, scheme_type, scheme_end);
                    }
                    return self.after_double_slash(remaining, scheme_type, scheme_end);
                }
                let path_start = base_url.path_start;
                self.serialization.push_str(base_url.slice(..path_start));
                self.serialization.push('/');
                let remaining = self.parse_path(
                    scheme_type,
                    &mut true,
                    path_start as usize,
                    input_after_first_char,
                );
                self.with_query_and_fragment(
                    scheme_type,
                    base_url.scheme_end,
                    base_url.username_end,
                    base_url.host_start,
                    base_url.host_end,
                    base_url.host,
                    base_url.port,
                    base_url.path_start,
                    remaining,
                )
            }
            _ => {
                let before_query = match (base_url.query_start, base_url.fragment_start) {
                    (None, None) => &*base_url.serialization,
                    (Some(i), _) | (None, Some(i)) => base_url.slice(..i),
                };
                self.serialization.push_str(before_query);
                // FIXME spec says just "remove last entry", not the "pop" algorithm
                self.pop_path(scheme_type, base_url.path_start as usize);
                // A special url always has a path.
                // A path always starts with '/'
                if self.serialization.len() == base_url.path_start as usize
                    && (SchemeType::from(base_url.scheme()).is_special() || !input.is_empty())
                {
                    self.serialization.push('/');
                }
                let remaining = match input.split_first() {
                    (Some('/'), remaining) => self.parse_path(
                        scheme_type,
                        &mut true,
                        base_url.path_start as usize,
                        remaining,
                    ),
                    _ => {
                        self.parse_path(scheme_type, &mut true, base_url.path_start as usize, input)
                    }
                };
                self.with_query_and_fragment(
                    scheme_type,
                    base_url.scheme_end,
                    base_url.username_end,
                    base_url.host_start,
                    base_url.host_end,
                    base_url.host,
                    base_url.port,
                    base_url.path_start,
                    remaining,
                )
            }
        }
    }

    fn after_double_slash(
        mut self,
        input: Input<'_>,
        scheme_type: SchemeType,
        scheme_end: u32,
    ) -> ParseResult<Url> {
        self.serialization.push('/');
        self.serialization.push('/');
        // authority state
        let before_authority = self.serialization.len();
        let (username_end, remaining) = self.parse_userinfo(input, scheme_type)?;
        let has_authority = before_authority != self.serialization.len();
        // host state
        let host_start = to_u32(self.serialization.len())?;
        let (host_end, host, port, remaining) =
            self.parse_host_and_port(remaining, scheme_end, scheme_type)?;
        if host == HostInternal::None && has_authority {
            return Err(ParseError::EmptyHost);
        }
        // path state
        let path_start = to_u32(self.serialization.len())?;
        let remaining = self.parse_path_start(scheme_type, &mut true, remaining);
        self.with_query_and_fragment(
            scheme_type,
            scheme_end,
            username_end,
            host_start,
            host_end,
            host,
            port,
            path_start,
            remaining,
        )
    }

    /// Return (username_end, remaining)
    fn parse_userinfo<'i>(
        &mut self,
        mut input: Input<'i>,
        scheme_type: SchemeType,
    ) -> ParseResult<(u32, Input<'i>)> {
        let mut last_at = None;
        let mut remaining = input.clone();
        let mut char_count = 0;
        while let Some(c) = remaining.next() {
            match c {
                '@' => {
                    if last_at.is_some() {
                        self.log_violation(SyntaxViolation::UnencodedAtSign)
                    } else {
                        self.log_violation(SyntaxViolation::EmbeddedCredentials)
                    }
                    last_at = Some((char_count, remaining.clone()))
                }
                '/' | '?' | '#' => break,
                '\\' if scheme_type.is_special() => break,
                _ => (),
            }
            char_count += 1;
        }
        let (mut userinfo_char_count, remaining) = match last_at {
            None => return Ok((to_u32(self.serialization.len())?, input)),
            Some((0, remaining)) => {
                // Otherwise, if one of the following is true
                // c is the EOF code point, U+002F (/), U+003F (?), or U+0023 (#)
                // url is special and c is U+005C (\)
                // If @ flag is set and buffer is the empty string, validation error, return failure.
                if let (Some(c), _) = remaining.split_first() {
                    if c == '/' || c == '?' || c == '#' || (scheme_type.is_special() && c == '\\') {
                        return Err(ParseError::EmptyHost);
                    }
                }
                return Ok((to_u32(self.serialization.len())?, remaining));
            }
            Some(x) => x,
        };

        let mut username_end = None;
        let mut has_password = false;
        let mut has_username = false;
        while userinfo_char_count > 0 {
            let (c, utf8_c) = input.next_utf8().unwrap();
            userinfo_char_count -= 1;
            if c == ':' && username_end.is_none() {
                // Start parsing password
                username_end = Some(to_u32(self.serialization.len())?);
                // We don't add a colon if the password is empty
                if userinfo_char_count > 0 {
                    self.serialization.push(':');
                    has_password = true;
                }
            } else {
                if !has_password {
                    has_username = true;
                }
                self.check_url_code_point(c, &input);
                self.serialization
                    .extend(utf8_percent_encode(utf8_c, USERINFO));
            }
        }
        let username_end = match username_end {
            Some(i) => i,
            None => to_u32(self.serialization.len())?,
        };
        if has_username || has_password {
            self.serialization.push('@');
        }
        Ok((username_end, remaining))
    }

    fn parse_host_and_port<'i>(
        &mut self,
        input: Input<'i>,
        scheme_end: u32,
        scheme_type: SchemeType,
    ) -> ParseResult<(u32, HostInternal, Option<u16>, Input<'i>)> {
        let (host, remaining) = Parser::parse_host(input, scheme_type)?;
        write!(&mut self.serialization, "{host}").unwrap();
        let host_end = to_u32(self.serialization.len())?;
        if let Host::Domain(h) = &host {
            if h.is_empty() {
                // Port with an empty host
                if remaining.starts_with(":") {
                    return Err(ParseError::EmptyHost);
                }
                if scheme_type.is_special() {
                    return Err(ParseError::EmptyHost);
                }
            }
        };

        let (port, remaining) = if let Some(remaining) = remaining.split_prefix(':') {
            let scheme = || default_port(&self.serialization[..scheme_end as usize]);
            let (port, remaining) = Parser::parse_port(remaining, scheme, self.context)?;
            if let Some(port) = port {
                self.serialization.push(':');
                let mut buffer = [0u8; 5];
                let port_str = fast_u16_to_str(&mut buffer, port);
                self.serialization.push_str(port_str);
            }
            (port, remaining)
        } else {
            (None, remaining)
        };
        Ok((host_end, host.into(), port, remaining))
    }

    pub fn parse_host(
        mut input: Input<'_>,
        scheme_type: SchemeType,
    ) -> ParseResult<(Host<Cow<'_, str>>, Input<'_>)> {
        if scheme_type.is_file() {
            return Parser::get_file_host(input);
        }
        // Undo the Input abstraction here to avoid allocating in the common case
        // where the host part of the input does not contain any tab or newline
        let input_str = input.chars.as_str();
        let mut inside_square_brackets = false;
        let mut has_ignored_chars = false;
        let mut non_ignored_chars = 0;
        let mut bytes = 0;
        for c in input_str.chars() {
            match c {
                ':' if !inside_square_brackets => break,
                '\\' if scheme_type.is_special() => break,
                '/' | '?' | '#' => break,
                ascii_tab_or_new_line_pattern!() => {
                    has_ignored_chars = true;
                }
                '[' => {
                    inside_square_brackets = true;
                    non_ignored_chars += 1
                }
                ']' => {
                    inside_square_brackets = false;
                    non_ignored_chars += 1
                }
                _ => non_ignored_chars += 1,
            }
            bytes += c.len_utf8();
        }
        let host_str;
        {
            let host_input = input.by_ref().take(non_ignored_chars);
            if has_ignored_chars {
                host_str = Cow::Owned(host_input.collect());
            } else {
                for _ in host_input {}
                host_str = Cow::Borrowed(&input_str[..bytes]);
            }
        }
        if scheme_type == SchemeType::SpecialNotFile && host_str.is_empty() {
            return Err(ParseError::EmptyHost);
        }
        if !scheme_type.is_special() {
            let host = Host::parse_opaque_cow(host_str)?;
            return Ok((host, input));
        }
        let host = Host::parse_cow(host_str)?;
        Ok((host, input))
    }

    fn get_file_host(input: Input<'_>) -> ParseResult<(Host<Cow<'_, str>>, Input<'_>)> {
        let (_, host_str, remaining) = Parser::file_host(input)?;
        let host = match Host::parse(&host_str)? {
            Host::Domain(ref d) if d == "localhost" => Host::Domain(Cow::Borrowed("")),
            Host::Domain(s) => Host::Domain(Cow::Owned(s)),
            Host::Ipv4(ip) => Host::Ipv4(ip),
            Host::Ipv6(ip) => Host::Ipv6(ip),
        };
        Ok((host, remaining))
    }

    fn parse_file_host<'i>(
        &mut self,
        input: Input<'i>,
    ) -> ParseResult<(bool, HostInternal, Input<'i>)> {
        let has_host;
        let (_, host_str, remaining) = Parser::file_host(input)?;
        let host = if host_str.is_empty() {
            has_host = false;
            HostInternal::None
        } else {
            match Host::parse_cow(host_str)? {
                Host::Domain(ref d) if d == "localhost" => {
                    has_host = false;
                    HostInternal::None
                }
                host => {
                    write!(&mut self.serialization, "{host}").unwrap();
                    has_host = true;
                    host.into()
                }
            }
        };
        Ok((has_host, host, remaining))
    }

    pub fn file_host(input: Input<'_>) -> ParseResult<(bool, Cow<'_, str>, Input<'_>)> {
        // Undo the Input abstraction here to avoid allocating in the common case
        // where the host part of the input does not contain any tab or newline
        let input_str = input.chars.as_str();
        let mut has_ignored_chars = false;
        let mut non_ignored_chars = 0;
        let mut bytes = 0;
        for c in input_str.chars() {
            match c {
                '/' | '\\' | '?' | '#' => break,
                ascii_tab_or_new_line_pattern!() => has_ignored_chars = true,
                _ => non_ignored_chars += 1,
            }
            bytes += c.len_utf8();
        }
        let host_str;
        let mut remaining = input.clone();
        {
            let host_input = remaining.by_ref().take(non_ignored_chars);
            if has_ignored_chars {
                host_str = Cow::Owned(host_input.collect());
            } else {
                for _ in host_input {}
                host_str = Cow::Borrowed(&input_str[..bytes]);
            }
        }
        if is_windows_drive_letter(&host_str) {
            return Ok((false, "".into(), input));
        }
        Ok((true, host_str, remaining))
    }

    pub fn parse_port<P>(
        mut input: Input<'_>,
        default_port: P,
        context: Context,
    ) -> ParseResult<(Option<u16>, Input<'_>)>
    where
        P: Fn() -> Option<u16>,
    {
        let mut port: u32 = 0;
        let mut has_any_digit = false;
        while let (Some(c), remaining) = input.split_first() {
            if let Some(digit) = c.to_digit(10) {
                port = port * 10 + digit;
                if port > u16::MAX as u32 {
                    return Err(ParseError::InvalidPort);
                }
                has_any_digit = true;
            } else if context == Context::UrlParser && !matches!(c, '/' | '\\' | '?' | '#') {
                return Err(ParseError::InvalidPort);
            } else {
                break;
            }
            input = remaining;
        }

        if !has_any_digit && context == Context::Setter && !input.is_empty() {
            return Err(ParseError::InvalidPort);
        }

        let mut opt_port = Some(port as u16);
        if !has_any_digit || opt_port == default_port() {
            opt_port = None;
        }
        Ok((opt_port, input))
    }

    pub fn parse_path_start<'i>(
        &mut self,
        scheme_type: SchemeType,
        has_host: &mut bool,
        input: Input<'i>,
    ) -> Input<'i> {
        let path_start = self.serialization.len();
        let (maybe_c, remaining) = input.split_first();
        // If url is special, then:
        if scheme_type.is_special() {
            if maybe_c == Some('\\') {
                // If c is U+005C (\), validation error.
                self.log_violation(SyntaxViolation::Backslash);
            }
            // A special URL always has a non-empty path.
            if !self.serialization.ends_with('/') {
                self.serialization.push('/');
                // We have already made sure the forward slash is present.
                if maybe_c == Some('/') || maybe_c == Some('\\') {
                    return self.parse_path(scheme_type, has_host, path_start, remaining);
                }
            }
            return self.parse_path(scheme_type, has_host, path_start, input);
        } else if maybe_c == Some('?') || maybe_c == Some('#') {
            // Otherwise, if state override is not given and c is U+003F (?),
            // set url’s query to the empty string and state to query state.
            // Otherwise, if state override is not given and c is U+0023 (#),
            // set url’s fragment to the empty string and state to fragment state.
            // The query and path states will be handled by the caller.
            return input;
        }

        if maybe_c.is_some() && maybe_c != Some('/') {
            self.serialization.push('/');
        }
        // Otherwise, if c is not the EOF code point:
        self.parse_path(scheme_type, has_host, path_start, input)
    }

    pub fn parse_path<'i>(
        &mut self,
        scheme_type: SchemeType,
        has_host: &mut bool,
        path_start: usize,
        mut input: Input<'i>,
    ) -> Input<'i> {
        // it's much faster to call utf8_percent_encode in bulk
        fn push_pending(
            serialization: &mut String,
            start_str: &str,
            remaining_len: usize,
            context: Context,
            scheme_type: SchemeType,
        ) {
            let text = &start_str[..start_str.len() - remaining_len];
            if text.is_empty() {
                return;
            }
            if context == Context::PathSegmentSetter {
                if scheme_type.is_special() {
                    serialization.extend(utf8_percent_encode(text, SPECIAL_PATH_SEGMENT));
                } else {
                    serialization.extend(utf8_percent_encode(text, PATH_SEGMENT));
                }
            } else {
                serialization.extend(utf8_percent_encode(text, PATH));
            }
        }

        // Relative path state
        loop {
            let mut segment_start = self.serialization.len();
            let mut ends_with_slash = false;
            let mut start_str = input.chars.as_str();
            loop {
                let input_before_c = input.clone();
                // bypass input.next() and manually handle ascii_tab_or_new_line
                // in order to encode string slices in bulk
                let c = if let Some(c) = input.chars.next() {
                    c
                } else {
                    push_pending(
                        &mut self.serialization,
                        start_str,
                        0,
                        self.context,
                        scheme_type,
                    );
                    break;
                };
                match c {
                    ascii_tab_or_new_line_pattern!() => {
                        push_pending(
                            &mut self.serialization,
                            start_str,
                            input_before_c.chars.as_str().len(),
                            self.context,
                            scheme_type,
                        );
                        start_str = input.chars.as_str();
                    }
                    '/' if self.context != Context::PathSegmentSetter => {
                        push_pending(
                            &mut self.serialization,
                            start_str,
                            input_before_c.chars.as_str().len(),
                            self.context,
                            scheme_type,
                        );
                        self.serialization.push(c);
                        ends_with_slash = true;
                        break;
                    }
                    '\\' if self.context != Context::PathSegmentSetter
                        && scheme_type.is_special() =>
                    {
                        push_pending(
                            &mut self.serialization,
                            start_str,
                            input_before_c.chars.as_str().len(),
                            self.context,
                            scheme_type,
                        );
                        self.log_violation(SyntaxViolation::Backslash);
                        self.serialization.push('/');
                        ends_with_slash = true;
                        break;
                    }
                    '?' | '#' if self.context == Context::UrlParser => {
                        push_pending(
                            &mut self.serialization,
                            start_str,
                            input_before_c.chars.as_str().len(),
                            self.context,
                            scheme_type,
                        );
                        input = input_before_c;
                        break;
                    }
                    _ => {
                        self.check_url_code_point(c, &input);
                        if scheme_type.is_file()
                            && self.serialization.len() > path_start
                            && is_normalized_windows_drive_letter(
                                &self.serialization[path_start + 1..],
                            )
                        {
                            push_pending(
                                &mut self.serialization,
                                start_str,
                                input_before_c.chars.as_str().len(),
                                self.context,
                                scheme_type,
                            );
                            start_str = input_before_c.chars.as_str();
                            self.serialization.push('/');
                            segment_start += 1;
                        }
                    }
                }
            }

            let segment_before_slash = if ends_with_slash {
                &self.serialization[segment_start..self.serialization.len() - 1]
            } else {
                &self.serialization[segment_start..self.serialization.len()]
            };
            match segment_before_slash {
                // If buffer is a double-dot path segment, shorten url’s path,
                ".." | "%2e%2e" | "%2e%2E" | "%2E%2e" | "%2E%2E" | "%2e." | "%2E." | ".%2e"
                | ".%2E" => {
                    debug_assert!(self.serialization.as_bytes()[segment_start - 1] == b'/');
                    self.serialization.truncate(segment_start);
                    if self.serialization.ends_with('/')
                        && Parser::last_slash_can_be_removed(&self.serialization, path_start)
                    {
                        self.serialization.pop();
                    }
                    self.shorten_path(scheme_type, path_start);

                    // and then if neither c is U+002F (/), nor url is special and c is U+005C (\), append the empty string to url’s path.
                    if ends_with_slash && !self.serialization.ends_with('/') {
                        self.serialization.push('/');
                    }
                }
                // Otherwise, if buffer is a single-dot path segment and if neither c is U+002F (/),
                // nor url is special and c is U+005C (\), append the empty string to url’s path.
                "." | "%2e" | "%2E" => {
                    self.serialization.truncate(segment_start);
                    if !self.serialization.ends_with('/') {
                        self.serialization.push('/');
                    }
                }
                _ => {
                    // If url’s scheme is "file", url’s path is empty, and buffer is a Windows drive letter, then
                    if scheme_type.is_file()
                        && segment_start == path_start + 1
                        && is_windows_drive_letter(segment_before_slash)
                    {
                        // Replace the second code point in buffer with U+003A (:).
                        if let Some(c) = segment_before_slash.chars().next() {
                            self.serialization.truncate(segment_start);
                            self.serialization.push(c);
                            self.serialization.push(':');
                            if ends_with_slash {
                                self.serialization.push('/');
                            }
                        }
                        // If url’s host is neither the empty string nor null,
                        // validation error, set url’s host to the empty string.
                        if *has_host {
                            self.log_violation(SyntaxViolation::FileWithHostAndWindowsDrive);
                            *has_host = false; // FIXME account for this in callers
                        }
                    }
                }
            }
            if !ends_with_slash {
                break;
            }
        }
        if scheme_type.is_file() {
            // while url’s path’s size is greater than 1
            // and url’s path[0] is the empty string,
            // validation error, remove the first item from url’s path.
            //FIXME: log violation
            let path = self.serialization.split_off(path_start);
            self.serialization.push('/');
            self.serialization.push_str(path.trim_start_matches('/'));
        }

        input
    }

    fn last_slash_can_be_removed(serialization: &str, path_start: usize) -> bool {
        let url_before_segment = &serialization[..serialization.len() - 1];
        if let Some(segment_before_start) = url_before_segment.rfind('/') {
            // Do not remove the root slash
            segment_before_start >= path_start
                // Or a windows drive letter slash
                && !path_starts_with_windows_drive_letter(&serialization[segment_before_start..])
        } else {
            false
        }
    }

    /// https://url.spec.whatwg.org/#shorten-a-urls-path
    fn shorten_path(&mut self, scheme_type: SchemeType, path_start: usize) {
        // If path is empty, then return.
        if self.serialization.len() == path_start {
            return;
        }
        // If url’s scheme is "file", path’s size is 1, and path[0] is a normalized Windows drive letter, then return.
        if scheme_type.is_file()
            && is_normalized_windows_drive_letter(&self.serialization[path_start..])
        {
            return;
        }
        // Remove path’s last item.
        self.pop_path(scheme_type, path_start);
    }

    /// https://url.spec.whatwg.org/#pop-a-urls-path
    fn pop_path(&mut self, scheme_type: SchemeType, path_start: usize) {
        if self.serialization.len() > path_start {
            let slash_position = self.serialization[path_start..].rfind('/').unwrap();
            // + 1 since rfind returns the position before the slash.
            let segment_start = path_start + slash_position + 1;
            // Don’t pop a Windows drive letter
            if !(scheme_type.is_file()
                && is_normalized_windows_drive_letter(&self.serialization[segment_start..]))
            {
                self.serialization.truncate(segment_start);
            }
        }
    }

    pub fn parse_cannot_be_a_base_path<'i>(&mut self, mut input: Input<'i>) -> Input<'i> {
        loop {
            let input_before_c = input.clone();
            match input.next_utf8() {
                Some(('?', _)) | Some(('#', _)) if self.context == Context::UrlParser => {
                    return input_before_c
                }
                Some((c, utf8_c)) => {
                    self.check_url_code_point(c, &input);
                    self.serialization
                        .extend(utf8_percent_encode(utf8_c, CONTROLS));
                }
                None => return input,
            }
        }
    }

    #[allow(clippy::too_many_arguments)]
    fn with_query_and_fragment(
        mut self,
        scheme_type: SchemeType,
        scheme_end: u32,
        username_end: u32,
        host_start: u32,
        host_end: u32,
        host: HostInternal,
        port: Option<u16>,
        mut path_start: u32,
        remaining: Input<'_>,
    ) -> ParseResult<Url> {
        // Special case for anarchist URL's with a leading empty path segment
        // This prevents web+demo:/.//not-a-host/ or web+demo:/path/..//not-a-host/,
        // when parsed and then serialized, from ending up as web+demo://not-a-host/
        // (they end up as web+demo:/.//not-a-host/).
        //
        // If url’s host is null, url does not have an opaque path,
        // url’s path’s size is greater than 1, and url’s path[0] is the empty string,
        // then append U+002F (/) followed by U+002E (.) to output.
        let scheme_end_as_usize = scheme_end as usize;
        let path_start_as_usize = path_start as usize;
        if path_start_as_usize == scheme_end_as_usize + 1 {
            // Anarchist URL
            if self.serialization[path_start_as_usize..].starts_with("//") {
                // Case 1: The base URL did not have an empty path segment, but the resulting one does
                // Insert the "/." prefix
                self.serialization.insert_str(path_start_as_usize, "/.");
                path_start += 2;
            }
            assert!(!self.serialization[scheme_end_as_usize..].starts_with("://"));
        } else if path_start_as_usize == scheme_end_as_usize + 3
            && &self.serialization[scheme_end_as_usize..path_start_as_usize] == ":/."
        {
            // Anarchist URL with leading empty path segment
            // The base URL has a "/." between the host and the path
            assert_eq!(self.serialization.as_bytes()[path_start_as_usize], b'/');
            if self
                .serialization
                .as_bytes()
                .get(path_start_as_usize + 1)
                .copied()
                != Some(b'/')
            {
                // Case 2: The base URL had an empty path segment, but the resulting one does not
                // Remove the "/." prefix
                self.serialization
                    .replace_range(scheme_end_as_usize..path_start_as_usize, ":");
                path_start -= 2;
            }
            assert!(!self.serialization[scheme_end_as_usize..].starts_with("://"));
        }

        let (query_start, fragment_start) =
            self.parse_query_and_fragment(scheme_type, scheme_end, remaining)?;
        Ok(Url {
            serialization: self.serialization,
            scheme_end,
            username_end,
            host_start,
            host_end,
            host,
            port,
            path_start,
            query_start,
            fragment_start,
        })
    }

    /// Return (query_start, fragment_start)
    fn parse_query_and_fragment(
        &mut self,
        scheme_type: SchemeType,
        scheme_end: u32,
        mut input: Input<'_>,
    ) -> ParseResult<(Option<u32>, Option<u32>)> {
        let mut query_start = None;
        match input.next() {
            Some('#') => {}
            Some('?') => {
                query_start = Some(to_u32(self.serialization.len())?);
                self.serialization.push('?');
                let remaining = self.parse_query(scheme_type, scheme_end, input);
                if let Some(remaining) = remaining {
                    input = remaining
                } else {
                    return Ok((query_start, None));
                }
            }
            None => return Ok((None, None)),
            _ => panic!("Programming error. parse_query_and_fragment() called without ? or #"),
        }

        let fragment_start = to_u32(self.serialization.len())?;
        self.serialization.push('#');
        self.parse_fragment(input);
        Ok((query_start, Some(fragment_start)))
    }

    pub fn parse_query<'i>(
        &mut self,
        scheme_type: SchemeType,
        scheme_end: u32,
        input: Input<'i>,
    ) -> Option<Input<'i>> {
        struct QueryPartIter<'i, 'p> {
            is_url_parser: bool,
            input: Input<'i>,
            violation_fn: Option<&'p dyn Fn(SyntaxViolation)>,
        }

        impl<'i> Iterator for QueryPartIter<'i, '_> {
            type Item = (&'i str, bool);

            fn next(&mut self) -> Option<Self::Item> {
                let start = self.input.chars.as_str();
                // bypass self.input.next() in order to get string slices
                // which are faster to operate on
                while let Some(c) = self.input.chars.next() {
                    match c {
                        ascii_tab_or_new_line_pattern!() => {
                            return Some((
                                &start[..start.len() - self.input.chars.as_str().len() - 1],
                                false,
                            ));
                        }
                        '#' if self.is_url_parser => {
                            return Some((
                                &start[..start.len() - self.input.chars.as_str().len() - 1],
                                true,
                            ));
                        }
                        c => {
                            if let Some(vfn) = &self.violation_fn {
                                check_url_code_point(vfn, c, &self.input);
                            }
                        }
                    }
                }
                if start.is_empty() {
                    None
                } else {
                    Some((start, false))
                }
            }
        }

        let mut part_iter = QueryPartIter {
            is_url_parser: self.context == Context::UrlParser,
            input,
            violation_fn: self.violation_fn,
        };
        let set = if scheme_type.is_special() {
            SPECIAL_QUERY
        } else {
            QUERY
        };
        let query_encoding_override = self.query_encoding_override.filter(|_| {
            matches!(
                &self.serialization[..scheme_end as usize],
                "http" | "https" | "file" | "ftp"
            )
        });

        while let Some((part, is_finished)) = part_iter.next() {
            match query_encoding_override {
                // slightly faster to be repetitive and not convert text to Cow
                Some(o) => self.serialization.extend(percent_encode(&o(part), set)),
                None => self
                    .serialization
                    .extend(percent_encode(part.as_bytes(), set)),
            }
            if is_finished {
                return Some(part_iter.input);
            }
        }

        None
    }

    fn fragment_only(mut self, base_url: &Url, mut input: Input<'_>) -> ParseResult<Url> {
        let before_fragment = match base_url.fragment_start {
            Some(i) => base_url.slice(..i),
            None => &*base_url.serialization,
        };
        debug_assert!(self.serialization.is_empty());
        self.serialization
            .reserve(before_fragment.len() + input.chars.as_str().len());
        self.serialization.push_str(before_fragment);
        self.serialization.push('#');
        let next = input.next();
        debug_assert!(next == Some('#'));
        self.parse_fragment(input);
        Ok(Url {
            serialization: self.serialization,
            fragment_start: Some(to_u32(before_fragment.len())?),
            ..*base_url
        })
    }

    pub fn parse_fragment(&mut self, input: Input<'_>) {
        struct FragmentPartIter<'i, 'p> {
            input: Input<'i>,
            violation_fn: Option<&'p dyn Fn(SyntaxViolation)>,
        }

        impl<'i> Iterator for FragmentPartIter<'i, '_> {
            type Item = &'i str;

            fn next(&mut self) -> Option<Self::Item> {
                let start = self.input.chars.as_str();
                // bypass self.input.next() in order to get string slices
                // which are faster to operate on
                while let Some(c) = self.input.chars.next() {
                    match c {
                        ascii_tab_or_new_line_pattern!() => {
                            return Some(
                                &start[..start.len() - self.input.chars.as_str().len() - 1],
                            );
                        }
                        '\0' => {
                            if let Some(vfn) = &self.violation_fn {
                                vfn(SyntaxViolation::NullInFragment);
                            }
                        }
                        c => {
                            if let Some(vfn) = &self.violation_fn {
                                check_url_code_point(vfn, c, &self.input);
                            }
                        }
                    }
                }
                if start.is_empty() {
                    None
                } else {
                    Some(start)
                }
            }
        }

        let part_iter = FragmentPartIter {
            input,
            violation_fn: self.violation_fn,
        };

        for part in part_iter {
            self.serialization
                .extend(utf8_percent_encode(part, FRAGMENT));
        }
    }

    #[inline]
    fn check_url_code_point(&self, c: char, input: &Input<'_>) {
        if let Some(vfn) = self.violation_fn {
            check_url_code_point(vfn, c, input)
        }
    }
}

fn check_url_code_point(vfn: &dyn Fn(SyntaxViolation), c: char, input: &Input<'_>) {
    if c == '%' {
        let mut input = input.clone();
        if !matches!((input.next(), input.next()), (Some(a), Some(b))
                             if a.is_ascii_hexdigit() && b.is_ascii_hexdigit())
        {
            vfn(SyntaxViolation::PercentDecode)
        }
    } else if !is_url_code_point(c) {
        vfn(SyntaxViolation::NonUrlCodePoint)
    }
}

// Non URL code points:
// U+0000 to U+0020 (space)
// " # % < > [ \ ] ^ ` { | }
// U+007F to U+009F
// surrogates
// U+FDD0 to U+FDEF
// Last two of each plane: U+__FFFE to U+__FFFF for __ in 00 to 10 hex
#[inline]
fn is_url_code_point(c: char) -> bool {
    matches!(c,
        'a'..='z' |
        'A'..='Z' |
        '0'..='9' |
        '!' | '$' | '&' | '\'' | '(' | ')' | '*' | '+' | ',' | '-' |
        '.' | '/' | ':' | ';' | '=' | '?' | '@' | '_' | '~' |
        '\u{A0}'..='\u{D7FF}' | '\u{E000}'..='\u{FDCF}' | '\u{FDF0}'..='\u{FFFD}' |
        '\u{10000}'..='\u{1FFFD}' | '\u{20000}'..='\u{2FFFD}' |
        '\u{30000}'..='\u{3FFFD}' | '\u{40000}'..='\u{4FFFD}' |
        '\u{50000}'..='\u{5FFFD}' | '\u{60000}'..='\u{6FFFD}' |
        '\u{70000}'..='\u{7FFFD}' | '\u{80000}'..='\u{8FFFD}' |
        '\u{90000}'..='\u{9FFFD}' | '\u{A0000}'..='\u{AFFFD}' |
        '\u{B0000}'..='\u{BFFFD}' | '\u{C0000}'..='\u{CFFFD}' |
        '\u{D0000}'..='\u{DFFFD}' | '\u{E1000}'..='\u{EFFFD}' |
        '\u{F0000}'..='\u{FFFFD}' | '\u{100000}'..='\u{10FFFD}')
}

/// https://url.spec.whatwg.org/#c0-controls-and-space
#[inline]
fn c0_control_or_space(ch: char) -> bool {
    ch <= ' ' // U+0000 to U+0020
}

/// https://infra.spec.whatwg.org/#ascii-tab-or-newline
#[inline]
fn ascii_tab_or_new_line(ch: char) -> bool {
    matches!(ch, ascii_tab_or_new_line_pattern!())
}

/// https://url.spec.whatwg.org/#ascii-alpha
#[inline]
pub fn ascii_alpha(ch: char) -> bool {
    ch.is_ascii_alphabetic()
}

#[inline]
pub fn to_u32(i: usize) -> ParseResult<u32> {
    if i <= u32::MAX as usize {
        Ok(i as u32)
    } else {
        Err(ParseError::Overflow)
    }
}

fn is_normalized_windows_drive_letter(segment: &str) -> bool {
    is_windows_drive_letter(segment) && segment.as_bytes()[1] == b':'
}

/// Whether the scheme is file:, the path has a single segment, and that segment
/// is a Windows drive letter
#[inline]
pub fn is_windows_drive_letter(segment: &str) -> bool {
    segment.len() == 2 && starts_with_windows_drive_letter(segment)
}

/// Whether path starts with a root slash
/// and a windows drive letter eg: "/c:" or "/a:/"
fn path_starts_with_windows_drive_letter(s: &str) -> bool {
    if let Some(c) = s.as_bytes().first() {
        matches!(c, b'/' | b'\\' | b'?' | b'#') && starts_with_windows_drive_letter(&s[1..])
    } else {
        false
    }
}

fn starts_with_windows_drive_letter(s: &str) -> bool {
    s.len() >= 2
        && ascii_alpha(s.as_bytes()[0] as char)
        && matches!(s.as_bytes()[1], b':' | b'|')
        && (s.len() == 2 || matches!(s.as_bytes()[2], b'/' | b'\\' | b'?' | b'#'))
}

/// https://url.spec.whatwg.org/#start-with-a-windows-drive-letter
fn starts_with_windows_drive_letter_segment(input: &Input<'_>) -> bool {
    let mut input = input.clone();
    match (input.next(), input.next(), input.next()) {
        // its first two code points are a Windows drive letter
        // its third code point is U+002F (/), U+005C (\), U+003F (?), or U+0023 (#).
        (Some(a), Some(b), Some(c))
            if ascii_alpha(a) && matches!(b, ':' | '|') && matches!(c, '/' | '\\' | '?' | '#') =>
        {
            true
        }
        // its first two code points are a Windows drive letter
        // its length is 2
        (Some(a), Some(b), None) if ascii_alpha(a) && matches!(b, ':' | '|') => true,
        _ => false,
    }
}

#[inline]
fn fast_u16_to_str(
    // max 5 digits for u16 (65535)
    buffer: &mut [u8; 5],
    mut value: u16,
) -> &str {
    let mut index = buffer.len();

    loop {
        index -= 1;
        buffer[index] = b'0' + (value % 10) as u8;
        value /= 10;
        if value == 0 {
            break;
        }
    }

    // SAFETY: we know the values in the buffer from the
    // current index on will be a number
    unsafe { core::str::from_utf8_unchecked(&buffer[index..]) }
}