feldera-sqllib 0.331.0

//! SQL String operations.
//!

// SqlString is implemented based on the ArcStr crate.
// The module contains various string operations.

#![allow(non_snake_case)]
use crate::{
    Variant, array::Array, some_function1, some_function2, some_function3, some_function4,
    some_polymorphic_function1, some_polymorphic_function2, some_polymorphic_null_function1,
    string_interner::*,
};

use arcstr::ArcStr;
use core::fmt::Error;
use feldera_macros::IsNone;
use feldera_types::{deserialize_without_context, serialize_without_context};
use itertools::Itertools;
use like::{Escape, Like};
use md5::{Digest, Md5};
use regex::Regex;
use rkyv::{
    DeserializeUnsized, Fallible, SerializeUnsized,
    string::{ArchivedString, StringResolver},
};
use serde::{Deserialize, Serialize};
use size_of::{Context, SizeOf};
use std::{
    cmp::max,
    fmt::{Display, Formatter},
    mem::{MaybeUninit, transmute},
    sync::Arc,
};
use tracing::warn;
use xxhash_rust::xxh64::xxh64;

type StringRef = ArcStr;
pub type InternedString = InternedStringId;

/// An immutable reference counted string.
#[derive(
    Clone, Default, Debug, Eq, Hash, Ord, PartialEq, PartialOrd, Serialize, Deserialize, IsNone,
)]
#[serde(transparent)]
pub struct SqlString(StringRef);

/// String representation used by the Feldera SQL runtime
impl SqlString {
    pub fn new() -> Self {
        Self::default()
    }

    pub fn from_ref(value: &str) -> Self {
        SqlString(StringRef::from(value.to_string()))
    }

    /// Constructs a `SqlString` that contains `value`.
    ///
    /// If `candidate` is the same as `value`, then reuses it instead of
    /// allocating a new string and copying into it.
    pub fn maybe_reuse(value: &str, candidate: &SqlString) -> Self {
        if value.as_ptr() == candidate.str().as_ptr() && value.len() == candidate.len() {
            candidate.clone()
        } else {
            SqlString(StringRef::from(value.to_string()))
        }
    }

    /// Constructs a `SqlString` that contains all of `strings` concatenated.
    ///
    /// This is more efficient than concatenating into a `String` and then
    /// converting it to a `SqlString` because `ArcStr::from` always allocates
    /// and copies.
    pub fn from_concat(strings: &[&str]) -> Self {
        Self::from_concat_iterator(strings.iter().copied())
    }

    /// Constructs a `SqlString` that contains the concatenation of the strings
    /// produced by the iterator.
    ///
    /// The iterator gets iterated twice and must produce the same total length
    /// (normally, the same strings) each time.
    pub fn from_concat_iterator<'a, I>(strings: I) -> Self
    where
        I: IntoIterator<Item = &'a str>,
        I::IntoIter: Clone,
    {
        fn copy_bytes_uninit(dst: &mut [MaybeUninit<u8>], src: &[u8]) {
            // SAFETY: &[T] and &[MaybeUninit<T>] have the same layout.
            let src: &[MaybeUninit<u8>] = unsafe { transmute(src) };
            dst.copy_from_slice(src);
        }

        let strings = strings.into_iter();
        let len = strings.clone().map(|s| s.len()).sum::<usize>();
        if len > 0 {
            Self(unsafe {
                ArcStr::init_with_unchecked(len, |buf| {
                    let mut offset = 0;
                    for s in strings {
                        let end = offset + s.len();
                        copy_bytes_uninit(&mut buf[offset..end], s.as_bytes());
                        offset = end;
                    }
                    debug_assert_eq!(offset, len);
                })
            })
        } else {
            // `ArcStr::init_with_unchecked()` panics for zero-length output
            // (see https://github.com/thomcc/arcstr/issues/67).
            Self(ArcStr::new())
        }
    }

    pub fn str(&self) -> &str {
        &self.0
    }

    pub fn len(&self) -> usize {
        self.0.len()
    }

    pub fn is_empty(&self) -> bool {
        self.0.is_empty()
    }
}

serialize_without_context!(SqlString);
deserialize_without_context!(SqlString);

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

impl From<ArcStr> for SqlString {
    fn from(value: ArcStr) -> Self {
        SqlString(value)
    }
}

impl From<String> for SqlString {
    fn from(value: String) -> Self {
        SqlString(StringRef::from(value))
    }
}

impl From<char> for SqlString {
    fn from(value: char) -> Self {
        // This avoids allocating and destroying a `String`.
        SqlString(StringRef::from(value.encode_utf8(&mut [0; 4])))
    }
}

impl From<&str> for SqlString {
    fn from(value: &str) -> Self {
        SqlString::from_ref(value)
    }
}

impl AsRef<ArcStr> for SqlString {
    fn as_ref(&self) -> &ArcStr {
        &self.0
    }
}

impl SizeOf for SqlString {
    fn size_of_children(&self, context: &mut Context) {
        self.0.size_of_children(context);
    }
}

impl rkyv::Archive for SqlString {
    type Archived = ArchivedString;
    type Resolver = StringResolver;

    #[inline]
    unsafe fn resolve(&self, pos: usize, resolver: Self::Resolver, out: *mut Self::Archived) {
        unsafe {
            ArchivedString::resolve_from_str(self.str(), pos, resolver, out);
        }
    }
}

impl<S: Fallible + ?Sized> rkyv::Serialize<S> for SqlString
where
    str: SerializeUnsized<S>,
{
    #[inline]
    fn serialize(&self, serializer: &mut S) -> Result<Self::Resolver, S::Error> {
        ArchivedString::serialize_from_str(self.str(), serializer)
    }
}

impl<D: Fallible + ?Sized> rkyv::Deserialize<SqlString, D> for ArchivedString
where
    str: DeserializeUnsized<str, D>,
{
    #[inline]
    fn deserialize(&self, _: &mut D) -> Result<SqlString, D::Error> {
        Ok(SqlString::from_ref(self.as_str()))
    }
}

#[doc(hidden)]
pub fn concat_s_s(left: SqlString, right: SqlString) -> SqlString {
    SqlString::from_concat(&[left.str(), right.str()])
}

some_polymorphic_function2!(concat, s, SqlString, s, SqlString, SqlString);

#[doc(hidden)]
fn substring3_impl(s: &str, left: i32, count: i32) -> &str {
    if count <= 0 {
        ""
    } else {
        // indexes in SQL start at 1
        let (start_char, char_count) = if left < 1 {
            // count applies from the start, even if start is negative
            (0_usize, max(count + left - 1, 0) as usize)
        } else {
            ((left - 1) as usize, count as usize)
        };

        let start_byte = byte_index(s, start_char);
        let end_byte = byte_index(&s[start_byte..], char_count) + start_byte;
        &s[start_byte..end_byte]
    }
}

/// Three-argument SUBSTRING in SQL.
///
/// (Calcite desugars SUBSTR into SUBSTRING.)
#[doc(hidden)]
pub fn substring3___(value: SqlString, left: i32, count: i32) -> SqlString {
    SqlString::maybe_reuse(substring3_impl(value.str(), left, count), &value)
}

some_function3!(substring3, SqlString, i32, i32, SqlString);

/// Returns the bytewise index in `s` corresponding to characterwise index
/// `char_index`.  If `char_index` is beyond the end of `s`, returns `s.len()`.
pub(crate) fn byte_index(s: &str, char_index: usize) -> usize {
    if char_index < s.len() {
        for (char_count, (byte_index, _)) in s.char_indices().enumerate() {
            if char_index == char_count {
                return byte_index;
            }
        }
    }
    s.len()
}

/// Returns the bytewise index in `s` corresponding to characterwise index
/// `char_index`, where character indexes count backward from the end of the
/// string, such that zero designates the last character in `s`.  If
/// `char_index` is greater than or equal to the number of characters in `s`,
/// returns `0`.
pub(crate) fn byte_index_rev(s: &str, char_index: usize) -> usize {
    if char_index < s.len() {
        for (char_count, (byte_index, _)) in s.char_indices().rev().enumerate() {
            if char_index == char_count {
                return byte_index;
            }
        }
    }
    0
}

fn substring2_impl(s: &str, left: i32) -> &str {
    // character indexes in SQL start at 1
    let char_index = left.max(1) - 1;
    let byte_index = byte_index(s, char_index as usize);
    &s[byte_index..]
}

/// Two-argument SUBSTRING in SQL.
///
/// (Calcite desugars SUBSTR into SUBSTRING.)
#[doc(hidden)]
pub fn substring2__(value: SqlString, left: i32) -> SqlString {
    let s = value.str();

    // character indexes in SQL start at 1
    let char_index = left.max(1) - 1;
    let byte_index = byte_index(s, char_index as usize);
    SqlString::maybe_reuse(&s[byte_index..], &value)
}

some_function2!(substring2, SqlString, i32, SqlString);

#[doc(hidden)]
pub fn trim_both_s_s(remove: SqlString, value: SqlString) -> SqlString {
    SqlString::maybe_reuse(
        value.str().trim_matches(|c| remove.str().contains(c)),
        &value,
    )
}

some_polymorphic_function2!(trim_both, s, SqlString, s, SqlString, SqlString);

#[doc(hidden)]
pub fn trim_leading_s_s(remove: SqlString, value: SqlString) -> SqlString {
    SqlString::maybe_reuse(
        value.str().trim_start_matches(|c| remove.str().contains(c)),
        &value,
    )
}

some_polymorphic_function2!(trim_leading, s, SqlString, s, SqlString, SqlString);

#[doc(hidden)]
pub fn trim_trailing_s_s(remove: SqlString, value: SqlString) -> SqlString {
    SqlString::maybe_reuse(
        value.str().trim_end_matches(|c| remove.str().contains(c)),
        &value,
    )
}

some_polymorphic_function2!(trim_trailing, s, SqlString, s, SqlString, SqlString);

#[doc(hidden)]
pub fn like2__(value: SqlString, pattern: SqlString) -> bool {
    Like::<false>::like(value.str(), pattern.str()).unwrap()
}

some_function2!(like2, SqlString, SqlString, bool);

#[doc(hidden)]
pub fn rlike__(value: SqlString, re: ParsedRegex) -> bool {
    match re {
        ParsedRegex::Invalid(r, e) => {
            panic!("Invalid regular expression in RLIKE '{}': {}", r, e);
        }
        ParsedRegex::Regex(re) => re.is_match(value.str()),
    }
}

some_function2!(rlike, SqlString, ParsedRegex, bool);

#[doc(hidden)]
pub fn like3___(value: SqlString, pattern: SqlString, escape: SqlString) -> bool {
    let escaped = pattern.str().escape(escape.str()).unwrap();
    Like::<true>::like(value.str(), &escaped).unwrap()
}

some_function3!(like3, SqlString, SqlString, SqlString, bool);

#[doc(hidden)]
pub fn ilike2__(value: SqlString, pattern: SqlString) -> bool {
    // Convert both the value and the pattern to lowercase for case-insensitive comparison
    Like::<false>::like(
        value.str().to_lowercase().as_str(),
        pattern.str().to_lowercase().as_str(),
    )
    .unwrap()
}

some_function2!(ilike2, SqlString, SqlString, bool);

#[doc(hidden)]
pub fn position__(needle: SqlString, haystack: SqlString) -> i32 {
    let s = haystack.str();
    let pos = s.find(needle.str());
    match pos {
        None => 0,
        Some(i) => (s[..i].chars().count() + 1) as i32,
    }
}

some_function2!(position, SqlString, SqlString, i32);

#[doc(hidden)]
pub fn char_length_(value: SqlString) -> i32 {
    value.str().chars().count() as i32
}

some_function1!(char_length, SqlString, i32);

#[doc(hidden)]
pub fn char_length_ref(value: &str) -> i32 {
    value.chars().count() as i32
}

#[doc(hidden)]
pub fn ascii_(value: SqlString) -> i32 {
    if value.str().is_empty() {
        0
    } else {
        value.str().chars().next().unwrap() as u32 as i32
    }
}

some_function1!(ascii, SqlString, i32);

#[doc(hidden)]
pub fn chr_(code: i32) -> SqlString {
    if code < 0 {
        SqlString::default()
    } else {
        let c = char::from_u32(code as u32);
        match c {
            None => SqlString::default(),
            Some(v) => SqlString::from(v),
        }
    }
}

some_function1!(chr, i32, SqlString);

#[doc(hidden)]
pub fn repeat__(value: SqlString, count: i32) -> SqlString {
    if count <= 0 {
        SqlString::default()
    } else {
        SqlString(ArcStr::repeat(value.str(), count as usize))
    }
}

some_function2!(repeat, SqlString, i32, SqlString);

#[doc(hidden)]
pub fn overlay3___(source: SqlString, replacement: SqlString, position: i32) -> SqlString {
    let len = char_length_ref(replacement.str());
    overlay4____(source, replacement, position, len)
}

some_function3!(overlay3, SqlString, SqlString, i32, SqlString);

#[doc(hidden)]
pub fn overlay4____(
    source: SqlString,
    replacement: SqlString,
    position: i32,
    remove: i32,
) -> SqlString {
    let mut remove = remove;
    if remove < 0 {
        remove = 0;
    }
    if position <= 0 {
        source
    } else if position > char_length_ref(source.str()) {
        concat_s_s(source, replacement)
    } else {
        let start = substring3_impl(source.str(), 1, position - 1);
        let middle = replacement.str();
        let end = substring2_impl(source.str(), position + remove);
        SqlString::from_concat(&[start, middle, end])
    }
}

some_function4!(overlay4, SqlString, SqlString, i32, i32, SqlString);

#[doc(hidden)]
pub fn lower_(source: SqlString) -> SqlString {
    SqlString::from(source.str().to_lowercase())
}

some_function1!(lower, SqlString, SqlString);

#[doc(hidden)]
pub fn upper_(source: SqlString) -> SqlString {
    SqlString::from(source.str().to_uppercase())
}

some_function1!(upper, SqlString, SqlString);

#[doc(hidden)]
pub fn initcap_(source: SqlString) -> SqlString {
    let mut result = String::with_capacity(source.str().len());
    let mut capitalize_next = true;
    for c in source.str().chars() {
        if c.is_alphanumeric() {
            if capitalize_next {
                for r in c.to_uppercase() {
                    result.push(r);
                }
                capitalize_next = false;
            } else {
                for r in c.to_lowercase() {
                    result.push(r);
                }
                capitalize_next = false;
            }
        } else {
            capitalize_next = true;
            result.push(c);
        }
    }
    SqlString::from(result)
}

some_function1!(initcap, SqlString, SqlString);

#[doc(hidden)]
pub fn initcap_spaces_(source: SqlString) -> SqlString {
    let mut result = String::with_capacity(source.str().len());
    let mut capitalize_next = true;
    for c in source.str().chars() {
        if !c.is_whitespace() {
            if capitalize_next {
                for r in c.to_uppercase() {
                    result.push(r);
                }
                capitalize_next = false;
            } else {
                for r in c.to_lowercase() {
                    result.push(r);
                }
                capitalize_next = false;
            }
        } else {
            capitalize_next = true;
            result.push(c);
        }
    }
    SqlString::from(result)
}

some_function1!(initcap_spaces, SqlString, SqlString);

#[doc(hidden)]
pub fn replace___(haystack: SqlString, needle: SqlString, replacement: SqlString) -> SqlString {
    SqlString::from(haystack.str().replace(needle.str(), replacement.str()))
}

some_function3!(replace, SqlString, SqlString, SqlString, SqlString);

#[doc(hidden)]
pub fn left_s_i32(value: SqlString, n_chars: i32) -> SqlString {
    if n_chars <= 0 {
        return SqlString::new();
    }

    let str = value.str();
    let n_bytes = byte_index(str, n_chars as usize);
    if n_bytes < str.len() {
        SqlString::from_ref(&str[..n_bytes])
    } else {
        value
    }
}

some_polymorphic_function2!(left, s, SqlString, i32, i32, SqlString);

#[doc(hidden)]
pub fn right_s_i32(value: SqlString, n_chars: i32) -> SqlString {
    if n_chars <= 0 {
        return SqlString::new();
    }
    let str = value.str();
    let byte_index = byte_index_rev(str, (n_chars - 1) as usize);
    if byte_index == 0 {
        value
    } else {
        SqlString::from_ref(&str[byte_index..])
    }
}

some_polymorphic_function2!(right, s, SqlString, i32, i32, SqlString);

#[doc(hidden)]
pub fn split2__(source: SqlString, separators: SqlString) -> Array<SqlString> {
    if separators.str().is_empty() {
        return Arc::new(vec![source]);
    }
    if source.str().is_empty() {
        return Arc::new(vec![]);
    }
    source
        .str()
        .split(separators.str())
        .map(SqlString::from)
        .collect::<Vec<SqlString>>()
        .into()
}

some_function2!(split2, SqlString, SqlString, Array<SqlString>);

#[doc(hidden)]
pub fn split1_(source: SqlString) -> Array<SqlString> {
    split2__(source, SqlString(arcstr::literal!(",")))
}

some_function1!(split1, SqlString, Array<SqlString>);

#[doc(hidden)]
pub fn reverse_(s: SqlString) -> SqlString {
    let r: String = s.str().chars().rev().collect();
    SqlString::from(r)
}

some_function1!(reverse, SqlString, SqlString);

#[doc(hidden)]
pub fn split_part___(s: SqlString, delimiter: SqlString, n: i32) -> SqlString {
    let parts: Array<SqlString> = split2__(s, delimiter);
    let part_count = parts.len() as i32;

    // Handle negative indices
    let n = if n < 0 { part_count + n + 1 } else { n };

    if n <= 0 || n > part_count {
        return SqlString::new();
    }

    parts[(n - 1) as usize].clone()
}

some_function3!(split_part, SqlString, SqlString, i32, SqlString);

#[doc(hidden)]
pub fn array_to_string2_vec__(value: Array<SqlString>, separator: SqlString) -> SqlString {
    SqlString::from_concat_iterator(Itertools::intersperse(
        value.iter().map(|s| s.str()),
        separator.str(),
    ))
}

some_function2!(array_to_string2_vec, Array<SqlString>, SqlString, SqlString);

#[doc(hidden)]
pub fn array_to_string2Nvec__(value: Array<Option<SqlString>>, separator: SqlString) -> SqlString {
    SqlString::from_concat_iterator(Itertools::intersperse(
        value.iter().flatten().map(|s| s.str()),
        separator.str(),
    ))
}

some_function2!(
    array_to_string2Nvec,
    Array<Option<SqlString>>,
    SqlString,
    SqlString
);

#[doc(hidden)]
pub fn array_to_string3_vec___(
    value: Array<SqlString>,
    separator: SqlString,
    _null_value: SqlString,
) -> SqlString {
    array_to_string2_vec__(value, separator)
}

some_function3!(
    array_to_string3_vec,
    Array<SqlString>,
    SqlString,
    SqlString,
    SqlString
);

#[doc(hidden)]
pub fn array_to_string3Nvec___(
    value: Array<Option<SqlString>>,
    separator: SqlString,
    null_value: SqlString,
) -> SqlString {
    SqlString::from_concat_iterator(Itertools::intersperse(
        value
            .iter()
            .map(|s| s.as_ref().unwrap_or(&null_value).str()),
        separator.str(),
    ))
}

some_function3!(
    array_to_string3Nvec,
    Array<Option<SqlString>>,
    SqlString,
    SqlString,
    SqlString
);

#[doc(hidden)]
pub fn writelog<T: std::fmt::Display>(format: SqlString, argument: T) -> T {
    let format_arg = format!("{}", argument);
    let formatted = format.str().replace("%%", &format_arg);
    print!("{}", formatted);
    argument
}

#[doc(hidden)]
pub fn parse_json_s(value: SqlString) -> Variant {
    let var = serde_json::from_str::<Variant>(value.str());
    match var {
        Ok(v) => v,
        Err(_) => Variant::SqlNull,
    }
}

#[doc(hidden)]
pub fn parse_json_nullN(_value: Option<()>) -> Option<Variant> {
    None
}

some_polymorphic_function1!(parse_json, s, SqlString, Variant);

#[doc(hidden)]
pub fn to_json_V(value: Variant) -> Option<SqlString> {
    match value.to_json_string() {
        Ok(s) => Some(SqlString::from(s)),
        _ => None,
    }
}

some_polymorphic_null_function1!(to_json, V, Variant, SqlString);

#[doc(hidden)]
pub fn to_json_nullN(_value: Option<()>) -> Option<SqlString> {
    None
}

#[doc(hidden)]
#[derive(Debug, Clone)]
pub enum ParsedRegex {
    Regex(Regex),
    // The expression and the error produced while parsing
    Invalid(SqlString, String),
}

#[doc(hidden)]
pub fn make_regex_(re: SqlString) -> ParsedRegex {
    match Regex::new(re.str()) {
        Ok(r) => ParsedRegex::Regex(r),
        Err(e) => {
            warn!(
                "failed to parse '{}' as a regular expression: {}",
                re,
                e.to_string()
            );
            ParsedRegex::Invalid(re, e.to_string())
        }
    }
}

some_function1!(make_regex, SqlString, ParsedRegex);

#[doc(hidden)]
pub fn regexp_replace3___(str: SqlString, re: ParsedRegex, repl: SqlString) -> SqlString {
    match re {
        ParsedRegex::Invalid(r, e) => {
            panic!("Invalid regular expression in REPLACE '{}': {}", r, e);
        }
        ParsedRegex::Regex(re) => {
            SqlString::maybe_reuse(re.replace_all(str.str(), repl.str()).as_ref(), &str)
        }
    }
}

some_function3!(
    regexp_replace3,
    SqlString,
    ParsedRegex,
    SqlString,
    SqlString
);

#[doc(hidden)]
pub fn regexp_replace2__(str: SqlString, re: ParsedRegex) -> SqlString {
    regexp_replace3___(str, re, SqlString::new())
}

some_function2!(regexp_replace2, SqlString, ParsedRegex, SqlString);

#[doc(hidden)]
pub fn concat_ws___(sep: SqlString, left: SqlString, right: SqlString) -> SqlString {
    SqlString::from_concat(&[left.str(), sep.str(), right.str()])
}

#[doc(hidden)]
pub fn concat_wsN__(
    sep: Option<SqlString>,
    left: SqlString,
    right: SqlString,
) -> Option<SqlString> {
    let sep = sep?;
    Some(concat_ws___(sep, left, right))
}

#[doc(hidden)]
pub fn concat_ws_N_(sep: SqlString, left: Option<SqlString>, right: SqlString) -> SqlString {
    match left {
        None => right,
        Some(left) => concat_ws___(sep, left, right),
    }
}

#[doc(hidden)]
pub fn concat_ws_NN(
    sep: SqlString,
    left: Option<SqlString>,
    right: Option<SqlString>,
) -> SqlString {
    match (left, right) {
        (None, None) => SqlString::new(),
        (None, Some(right)) => right,
        (Some(left), None) => left,
        (Some(left), Some(right)) => concat_ws___(sep, left, right),
    }
}

#[doc(hidden)]
pub fn concat_wsNN_(
    sep: Option<SqlString>,
    left: Option<SqlString>,
    right: SqlString,
) -> Option<SqlString> {
    let sep = sep?;
    Some(concat_ws_N_(sep, left, right))
}

#[doc(hidden)]
pub fn concat_ws__N(sep: SqlString, left: SqlString, right: Option<SqlString>) -> SqlString {
    match right {
        None => left,
        Some(right) => concat_ws___(sep, left, right),
    }
}

#[doc(hidden)]
pub fn concat_wsN_N(
    sep: Option<SqlString>,
    left: SqlString,
    right: Option<SqlString>,
) -> Option<SqlString> {
    let sep = sep?;
    Some(concat_ws__N(sep, left, right))
}

#[doc(hidden)]
pub fn concat_wsNNN(
    sep: Option<SqlString>,
    left: Option<SqlString>,
    right: Option<SqlString>,
) -> Option<SqlString> {
    let sep = sep?;
    Some(concat_ws_NN(sep, left, right))
}

#[doc(hidden)]
#[cfg(test)]
mod tests {
    use std::sync::Arc;

    use crate::{
        SqlString, array_to_string2_vec__, array_to_string2Nvec__, array_to_string3Nvec___,
        byte_index, byte_index_rev, left_s_i32, right_s_i32, substring2__, substring3___,
    };

    use dbsp::storage::file::to_bytes;
    use rkyv::from_bytes;
    use size_of::SizeOf;

    #[test]
    fn rkyv_serialize_deserialize_sqlstring() {
        let s = SqlString::from_ref("abc1✅");
        let archived = to_bytes(&s).unwrap();
        let deserialized: SqlString = from_bytes(&archived).unwrap();
        assert_eq!(s.str(), deserialized.str());
    }

    #[test]
    fn sizeof_sqlstring() {
        let s = SqlString::from_ref("abcdefghijklmnopqrstuvwxyz");
        let total_size = SizeOf::size_of(&s);

        // The exact size may depend on the architecture's pointer size.
        dbg!(&total_size);
        assert!(total_size.total_bytes() > 26);
        assert!(total_size.shared_bytes() >= 26);
    }

    #[test]
    fn test_byte_index() {
        for (s, byte_offsets) in [
            // ASCII letters plus 2-byte e with acute accent.
            ("abcdéfg", vec![0, 1, 2, 3, 4, 6, 7, 8]),
            // Japanese katakana, 3 bytes each.
            ("アイウエオ", vec![0, 3, 6, 9, 12, 15]),
        ] {
            for (char_offset, byte_offset) in byte_offsets.iter().copied().enumerate() {
                assert_eq!(byte_index(s, char_offset), byte_offset);
                assert_eq!(
                    left_s_i32(SqlString::from_ref(s), char_offset as i32),
                    SqlString::from_ref(&s[..byte_offset])
                );
                assert_eq!(
                    substring2__(SqlString::from_ref(s), (char_offset + 1) as i32),
                    SqlString::from_ref(&s[byte_offset..])
                );
                for (char_count, end_byte_offset) in
                    byte_offsets[char_offset..].iter().copied().enumerate()
                {
                    assert_eq!(
                        substring3___(
                            SqlString::from_ref(s),
                            (char_offset + 1) as i32,
                            char_count as i32
                        ),
                        SqlString::from_ref(&s[byte_offset..end_byte_offset])
                    );
                }
            }
            for (char_offset, byte_offset) in byte_offsets.iter().copied().rev().skip(1).enumerate()
            {
                assert_eq!(byte_index_rev(s, char_offset), byte_offset);
                assert_eq!(
                    right_s_i32(SqlString::from_ref(s), (char_offset + 1) as i32),
                    SqlString::from_ref(&s[byte_offset..])
                )
            }
            assert_eq!(left_s_i32(SqlString::from_ref(s), 0), SqlString::new());
            assert_eq!(right_s_i32(SqlString::from_ref(s), 0), SqlString::new());
        }
    }

    #[test]
    fn empty() {
        assert_eq!(
            SqlString::from_concat(&["", "", "", ""]),
            SqlString::from("")
        );
    }

    #[test]
    fn concat() {
        assert_eq!(
            SqlString::from_concat(&["a", "xyzzy", "✨", "", "c", "plop"]),
            SqlString::from("axyzzy✨cplop")
        );
    }

    #[test]
    fn array_to_string_vec() {
        assert_eq!(
            array_to_string2_vec__(
                Arc::new(vec![
                    SqlString::from("xyzzy"),
                    SqlString::from("quux"),
                    SqlString::from("カキクケコ"),
                    SqlString::from("foo"),
                ]),
                SqlString::from("_"),
            ),
            SqlString::from("xyzzy_quux_カキクケコ_foo")
        );
        assert_eq!(
            array_to_string2Nvec__(
                Arc::new(vec![
                    None,
                    Some(SqlString::from("xyzzy")),
                    Some(SqlString::from("quux")),
                    None,
                    Some(SqlString::from("foo")),
                    None,
                ]),
                SqlString::from("_"),
            ),
            SqlString::from("xyzzy_quux_foo")
        );
        assert_eq!(
            array_to_string3Nvec___(
                Arc::new(vec![
                    None,
                    Some(SqlString::from("Tokyo(東京)")),
                    Some(SqlString::from("quux")),
                    None,
                    Some(SqlString::from("foo")),
                    None,
                ]),
                SqlString::from("_"),
                SqlString::from("#"),
            ),
            SqlString::from("#_Tokyo(東京)_quux_#_foo_#")
        );
    }
}

#[doc(hidden)]
pub fn md5_s(source: SqlString) -> SqlString {
    let mut hasher = Md5::new();
    hasher.update(source.str());
    let result = hasher.finalize();
    SqlString::from(format!("{:x}", result))
}

some_polymorphic_function1!(md5, s, SqlString, SqlString);

#[doc(hidden)]
pub fn xxhash_s_i64(source: SqlString, seed: i64) -> i64 {
    let hash = xxh64(source.str().as_bytes(), seed as u64);
    hash as i64
}

some_polymorphic_function2!(xxhash, s, SqlString, i64, i64, i64);

#[doc(hidden)]
pub fn intern(s: Option<SqlString>) -> Option<InternedString> {
    s.map(|s| intern_string(&s))
}

#[doc(hidden)]
pub fn unintern(id: Option<InternedString>) -> Option<SqlString> {
    match id {
        None => None,
        Some(id) => match unintern_string(&id) {
            Some(s) => Some(s),
            None => panic!("Interning library has not returned a string for id='{id}'"),
        },
    }
}

#[cfg(test)]
mod test {
    use crate::SqlString;
    use rkyv::{option::ArchivedOption, string::ArchivedString};

    #[test]
    fn sql_string_size() {
        // This is the same, the None is optimized away
        assert_eq!(
            std::mem::size_of::<SqlString>(),
            std::mem::size_of::<Option<SqlString>>()
        );

        // This None isn't optimized away here, ideally it would be
        assert!(
            std::mem::size_of::<ArchivedString>()
                < std::mem::size_of::<ArchivedOption<ArchivedString>>()
        )
    }
}