surrealdb-core 3.2.0

use unicase::UniCase;

use super::unicode::is_identifier_continue;
use crate::syn::error::{SyntaxError, bail, syntax_error};
use crate::syn::lexer::keywords::KEYWORDS;
use crate::syn::lexer::{BytesReader, Lexer};
use crate::syn::token::{Span, Token, TokenKind};

const BRACKET_CHARACTERS: [u8; 3] = const {
	let mut b = [0; 3];
	if '⟨'.encode_utf8(&mut b).len() != 3 {
		panic!()
	}
	b
};
const BRACKET_START_CHARACTER: u8 = BRACKET_CHARACTERS[0];

impl Lexer<'_> {
	pub fn unescape_ident_span<'a>(
		str: &'a str,
		span: Span,
		buffer: &'a mut Vec<u8>,
	) -> Result<&'a str, SyntaxError> {
		let mut reader = BytesReader::new(str.as_bytes());
		match reader.next() {
			Some(b'`') => Self::unescape_backtick_span(reader, span, buffer),
			// Non utf-8 character could almost only be the bracket here
			Some(BRACKET_START_CHARACTER) => Self::unescape_bracket_span(reader, span, buffer),
			// This is an unescaped ident so there is no need to do anything.
			_ => Ok(str),
		}
	}

	fn unescape_backtick_span<'a>(
		mut reader: BytesReader,
		span: Span,
		buffer: &'a mut Vec<u8>,
	) -> Result<&'a str, SyntaxError> {
		buffer.clear();
		loop {
			// lexer ensures that backtick tokens end with `.
			let before = reader.offset();
			let x = reader.next().expect("lexer validated input");
			match x {
				b'\\' => {
					// Lexer already ensures there is a valid character after the \
					Self::lex_common_escape_sequence(&mut reader, span, before, buffer, false)?;
				}
				b'`' => break,
				x => {
					buffer.push(x);
				}
			}
		}

		// Safety: We clear the buffer, and then we only push either ascii characters, bytes from
		// an existing valid string or explicity utf-8 encoded bytes. Therefore it is safe to
		// assume the buffer is valid utf-8
		Ok(unsafe { std::str::from_utf8_unchecked(buffer) })
	}

	fn unescape_bracket_span<'a>(
		mut reader: BytesReader,
		span: Span,
		buffer: &'a mut Vec<u8>,
	) -> Result<&'a str, SyntaxError> {
		buffer.clear();
		assert_eq!(reader.complete_char(BRACKET_START_CHARACTER).expect("valid character"), '⟨');
		loop {
			// lexer ensures that backtick tokens end with `
			let before = reader.offset();
			let x = reader.next().expect("lexer validated input");
			match x {
				b'\\' => {
					// Lexer already ensures there is a valid character after the \
					Self::lex_common_escape_sequence(&mut reader, span, before, buffer, false)?;
				}
				x if !x.is_ascii() => {
					let c = reader.complete_char(x).expect("valid character");
					if c == '⟩' {
						break;
					} else {
						let mut char_buffer = [0u8; 4];
						buffer.extend_from_slice(c.encode_utf8(&mut char_buffer).as_bytes());
					}
				}
				x => {
					buffer.push(x);
				}
			}
		}

		// Safety: We clear the buffer, and then we only push either ascii characters, bytes from
		// an existing valid string or explicity utf-8 encoded bytes. Therefore it is safe to
		// assume the buffer is valid utf-8
		Ok(unsafe { std::str::from_utf8_unchecked(buffer) })
	}

	/// Lex a parameter in the form of `$[a-zA-Z0-9_]*`
	///
	/// # Lexer State
	/// Expected the lexer to have already eaten the param starting `$`
	pub(super) fn lex_param(&mut self) -> Token {
		loop {
			if let Some(x) = self.reader.peek()
				&& (x.is_ascii_alphanumeric() || x == b'_')
			{
				self.reader.next();
				continue;
			}
			return self.finish_token(TokenKind::Parameter);
		}
	}

	pub(super) fn lex_surrounded_param(&mut self, is_backtick: bool) -> Token {
		match self.lex_surrounded_ident_err(is_backtick) {
			Ok(_) => self.finish_token(TokenKind::Parameter),
			Err(e) => self.invalid_token(e),
		}
	}

	/// Lex an not surrounded identifier in the form of `[a-zA-Z0-9_]*`
	///
	/// The start byte should already a valid byte of the identifier.
	///
	/// When calling the caller should already know that the token can't be any
	/// other token covered by `[a-zA-Z0-9_]*`.
	pub(super) fn lex_ident_from_next_byte(&mut self, start: u8) -> Token {
		debug_assert!(matches!(start, b'a'..=b'z' | b'A'..=b'Z' | b'_'));
		self.lex_ident()
	}

	/// Lex a not surrounded identfier.
	///
	/// The scratch should contain only identifier valid chars.
	pub(super) fn lex_ident(&mut self) -> Token {
		loop {
			if let Some(x) = self.reader.peek()
				&& is_identifier_continue(x)
			{
				self.reader.next();
				continue;
			}

			let str = self.span_str(self.current_span());

			// When finished parsing the identifier, try to match it to an keyword.
			// If there is one, return it as the keyword. Original identifier can be
			// reconstructed from the token.
			if let Some(x) = KEYWORDS.get(&UniCase::ascii(str)).copied() {
				if x != TokenKind::Identifier {
					return self.finish_token(x);
				}
			} else if str == "NaN" {
				return self.finish_token(TokenKind::NaN);
			} else if str == "Infinity" {
				return self.finish_token(TokenKind::Infinity);
			}

			return self.finish_token(TokenKind::Identifier);
		}
	}

	/// Lex an ident which is surround by delimiters.
	pub(super) fn lex_surrounded_ident(&mut self, is_backtick: bool) -> Token {
		match self.lex_surrounded_ident_err(is_backtick) {
			Ok(_) => self.finish_token(TokenKind::Identifier),
			Err(e) => self.invalid_token(e),
		}
	}

	/// Lex an ident surrounded either by `⟨⟩` or `\`\``
	pub(super) fn lex_surrounded_ident_err(
		&mut self,
		is_backtick: bool,
	) -> Result<(), SyntaxError> {
		let start_span = self.current_span();
		loop {
			let Some(x) = self.reader.next() else {
				let end_char = if is_backtick {
					'`'
				} else {
					'⟩'
				};
				let error = syntax_error!("Unexpected end of file, expected identifier to end with `{end_char}`", @self.current_span());
				return Err(error);
			};
			match x {
				b'`' if is_backtick => {
					return Ok(());
				}
				b'\\' => {
					// Don't bother parsing escape sequences, just skip the next byte
					let Some(next) = self.reader.next() else {
						bail!("Unexpected end of file, expected identifier to end.", @start_span => "Identifier starting here.");
					};

					if !next.is_ascii() {
						self.reader.complete_char(next)?;
					}
				}
				BRACKET_START_CHARACTER if !is_backtick => {
					if self.reader.complete_char(BRACKET_START_CHARACTER)? == '⟩' {
						return Ok(());
					}
				}
				x => {
					if !x.is_ascii() {
						self.reader.complete_char(x)?;
					}
				}
			}
		}
	}
}