5895308192

Committed 17 Aug 2023 06:16PM UTC coverage: 87.228% (-0.02%) from 87.246%

Build # 5895308192

Build Type

push

github

Committed by

web-flow

Commit Message

Fix performance of Sabre rust<->Python boundary (#10652) (#10659)

* Fix performance of Sabre rust<->Python boundary

In #10366 the SabreLayout and SabreSwap passes were refactored to
support nested control flow blocks. As part of that refactor a new
struct SabreResult was created to store the nested results for each
block. This new class however resulted in PyO3 cloning the full swap map
on every access. Since we have at least 1 access per gate in the circuit
(and another one for each swap) this extra clone() adds a lot of extra
overhead for deep circuits. In extreme cases this regression could be
quite extreme. To address this the result format is changed to be a
tuple (as it was originally), while this is less ergonomic the advantage
it provides is that for nested objects it moves the rust object to the
pyo3 interface so we avoid a copy as Python owns the object on the
return. However, for control flow blocks we're still using the
SabreResult class as it simplifies the internal implementation (which is
why #10366 introduced the struct). The potential impact of this is
mitigated by switching to only a single clone per control flow block,
by only accessing the SabreResult object's attribute on each recursive
call. However, if it is an issue in the future we can work on flattening
the nested structure before returning it to python to avoid any clones.

Fixes #10650

* Simplify recursive call logic in _apply_sabre_result

This commit simplifies the logic in the recursive call logic in
_apply_sabre_result to always use a tuple so we avoid an isinstance
check.

Co-authored-by: Kevin Hartman <kevin@hart.mn>

---------

Co-authored-by: Kevin Hartman <kevin@hart.mn>
(cherry picked from commit e6c431e30)

Co-authored-by: Matthew Treinish <mtreinish@kortar.org>

Run Details

25 of 25 new or added lines in 3 files covered. (100.0%)

74091 of 84939 relevant lines covered (87.23%)

60892.39 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

91.14

/crates/qasm2/src/lex.rs

// This code is part of Qiskit.
//
// (C) Copyright IBM 2023
//
// This code is licensed under the Apache License, Version 2.0. You may
// obtain a copy of this license in the LICENSE.txt file in the root directory
// of this source tree or at http://www.apache.org/licenses/LICENSE-2.0.
//
// Any modifications or derivative works of this code must retain this
// copyright notice, and modified files need to carry a notice indicating
// that they have been altered from the originals.

//! The lexing logic for OpenQASM 2, responsible for turning a sequence of bytes into a
//! lexed [TokenStream] for consumption by the parsing machinery.  The general strategy here is
//! quite simple; for all the symbol-like tokens, the lexer can use a very simple single-byte
//! lookahead to determine what token it needs to emit.  For keywords and identifiers, we just read
//! the identifier in completely, then produce the right token once we see the end of the
//! identifier characters.
//!
//! We effectively use a custom lexing mode to handle the version information after the `OPENQASM`
//! keyword; the spec technically says that any real number is valid, but in reality that leads to
//! weirdness like `200.0e-2` being a valid version specifier.  We do things with a custom
//! context-dependent match after seeing an `OPENQASM` token, to avoid clashes with the general
//! real-number tokenisation.

use hashbrown::HashMap;
use pyo3::prelude::PyResult;

use std::path::Path;

use crate::error::{message_generic, Position, QASM2ParseError};

/// Tokenised version information data.  This is more structured than the real number suggested by
/// the specification.
#[derive(Clone, Debug)]
pub struct Version {
    pub major: usize,
    pub minor: Option<usize>,
}

/// The context that is necessary to fully extract the information from a [Token].  This owns, for
/// example, the text of each token (where a token does not have a static text representation),
/// from which the other properties can later be derived.  This struct is effectively entirely
/// opaque outside this module; the associated functions on [Token] take this context object,
/// however, and extract the information from it.
#[derive(Clone, Debug)]
pub struct TokenContext {
    text: Vec<String>,
    lookup: HashMap<Vec<u8>, usize>,
}

impl TokenContext {
    /// Create a new context for tokens.  Nothing is heap-allocated until required.
    pub fn new() -> Self {
        TokenContext {
            text: vec![],
            lookup: HashMap::new(),
        }
    }

    /// Intern the given `ascii_text` of a [Token], and return an index into the [TokenContext].
    /// This will not store strings that are already present in the context; instead, the previous
    /// index is transparently returned.
    fn index(&mut self, ascii_text: &[u8]) -> usize {
        match self.lookup.get(ascii_text) {
            Some(index) => *index,
            None => {
                let index = self.text.len();
                self.lookup.insert(ascii_text.to_vec(), index);
                self.text
                    .push(std::str::from_utf8(ascii_text).unwrap().to_owned());
                index
            }
        }
    }
}

// Clippy complains without this.
impl Default for TokenContext {
    fn default() -> Self {
        Self::new()
    }
}

/// An enumeration of the different types of [Token] that can be created during lexing.  This is
/// deliberately not a data enum, to make various abstract `expect` (and so on) methods more
/// ergonomic to use; one does not need to completely define the pattern match each time, but can
/// simply pass the type identifier.  This also saves memory, since the static variants do not need
/// to be aligned to include the space necessary for text pointers that would be in the non-static
/// forms, and allows strings to be shared between many tokens (using the [TokenContext] store).
#[derive(PartialEq, Eq, Clone, Copy, Debug)]
pub enum TokenType {
    // Keywords
    OpenQASM,
    Barrier,
    Cos,
    Creg,
    Exp,
    Gate,
    If,
    Include,
    Ln,
    Measure,
    Opaque,
    Qreg,
    Reset,
    Sin,
    Sqrt,
    Tan,
    Pi,
    // Symbols
    Plus,
    Minus,
    Arrow,
    Asterisk,
    Equals,
    Slash,
    Caret,
    Semicolon,
    Comma,
    LParen,
    RParen,
    LBracket,
    RBracket,
    LBrace,
    RBrace,
    // Content
    Id,
    Real,
    Integer,
    Filename,
    Version,
}

impl TokenType {
    pub fn variable_text(&self) -> bool {
        match self {
            TokenType::OpenQASM
            | TokenType::Barrier
            | TokenType::Cos
            | TokenType::Creg
            | TokenType::Exp
            | TokenType::Gate
            | TokenType::If
            | TokenType::Include
            | TokenType::Ln
            | TokenType::Measure
            | TokenType::Opaque
            | TokenType::Qreg
            | TokenType::Reset
            | TokenType::Sin
            | TokenType::Sqrt
            | TokenType::Tan
            | TokenType::Pi
            | TokenType::Plus
            | TokenType::Minus
            | TokenType::Arrow
            | TokenType::Asterisk
            | TokenType::Equals
            | TokenType::Slash
            | TokenType::Caret
            | TokenType::Semicolon
            | TokenType::Comma
            | TokenType::LParen
            | TokenType::RParen
            | TokenType::LBracket
            | TokenType::RBracket
            | TokenType::LBrace
            | TokenType::RBrace => false,
            TokenType::Id
            | TokenType::Real
            | TokenType::Integer
            | TokenType::Filename
            | TokenType::Version => true,
        }
    }

    /// Get a static description of the token type.  This is useful for producing messages when the
    /// full token context isn't available, or isn't important.
    pub fn describe(&self) -> &'static str {
        match self {
            TokenType::OpenQASM => "OPENQASM",
            TokenType::Barrier => "barrier",
            TokenType::Cos => "cos",
            TokenType::Creg => "creg",
            TokenType::Exp => "exp",
            TokenType::Gate => "gate",
            TokenType::If => "if",
            TokenType::Include => "include",
            TokenType::Ln => "ln",
            TokenType::Measure => "measure",
            TokenType::Opaque => "opaque",
            TokenType::Qreg => "qreg",
            TokenType::Reset => "reset",
            TokenType::Sin => "sin",
            TokenType::Sqrt => "sqrt",
            TokenType::Tan => "tan",
            TokenType::Pi => "pi",
            TokenType::Plus => "+",
            TokenType::Minus => "-",
            TokenType::Arrow => "->",
            TokenType::Asterisk => "*",
            TokenType::Equals => "==",
            TokenType::Slash => "/",
            TokenType::Caret => "^",
            TokenType::Semicolon => ";",
            TokenType::Comma => ",",
            TokenType::LParen => "(",
            TokenType::RParen => ")",
            TokenType::LBracket => "[",
            TokenType::RBracket => "]",
            TokenType::LBrace => "{",
            TokenType::RBrace => "}",
            TokenType::Id => "an identifier",
            TokenType::Real => "a real number",
            TokenType::Integer => "an integer",
            TokenType::Filename => "a filename string",
            TokenType::Version => "a '<major>.<minor>' version",
        }
    }
}

/// A representation of a token, including its type, span information and pointer to where its text
/// is stored in the context object.  These are relatively lightweight objects (though of course
/// not as light as the single type information).
#[derive(Clone, Copy, Debug)]
pub struct Token {
    pub ttype: TokenType,
    // The `line` and `col` refer only to the start of the token.  There are no tokens that span
    // more than one line (we don't tokenise comments), but the ending column offset can be
    // calculated by asking the associated `TokenContext` for the text associated with this token,
    // and inspecting the length of the returned value.
    pub line: usize,
    pub col: usize,
    // Index into the TokenContext object, to retrieve the text that makes up the token.  We don't
    // resolve this into a value during lexing; that comes with annoying typing issues or storage
    // wastage.  Instead, we only convert the text into a value type when asked to by calling a
    // relevant method on the token.
    index: usize,
}

impl Token {
    /// Get a reference to the string that was seen to generate this token.
    pub fn text<'a>(&self, context: &'a TokenContext) -> &'a str {
        match self.ttype {
            TokenType::Id
            | TokenType::Real
            | TokenType::Integer
            | TokenType::Filename
            | TokenType::Version => &context.text[self.index],
            _ => self.ttype.describe(),
        }
    }

    /// If the token is an identifier, this method can be called to get an owned string containing
    /// the text of the identifier.  Panics if the token is not an identifier.
    pub fn id(&self, context: &TokenContext) -> String {
        if self.ttype != TokenType::Id {
            panic!()
        }
        (&context.text[self.index]).into()
    }

    /// If the token is a real number, this method can be called to evaluate its value.  Panics if
    /// the token is not a real number.
    pub fn real(&self, context: &TokenContext) -> f64 {
        if self.ttype != TokenType::Real {
            panic!()
        }
        context.text[self.index].parse().unwrap()
    }

    /// If the token is an integer (by type, not just by value), this method can be called to
    /// evaluate its value.  Panics if the token is not an integer type.
    pub fn int(&self, context: &TokenContext) -> usize {
        if self.ttype != TokenType::Integer {
            panic!()
        }
        context.text[self.index].parse().unwrap()
    }

    /// If the token is a filename path, this method can be called to get a (regular) string
    /// representing it.  Panics if the token type was not a filename.
    pub fn filename(&self, context: &TokenContext) -> String {
        if self.ttype != TokenType::Filename {
            panic!()
        }
        let out = &context.text[self.index];
        // String slicing is fine to assume bytes here, because the characters we're slicing out
        // must both be the ASCII '"', which is a single-byte UTF-8 character.
        out[1..out.len() - 1].into()
    }

    /// If the token is a version-information token, this method can be called to evaluate the
    /// version information.  Panics if the token was not of the correct type.
    pub fn version(&self, context: &TokenContext) -> Version {
        if self.ttype != TokenType::Version {
            panic!()
        }
        // Everything in the version token is a valid ASCII character, so must be a one-byte token.
        let text = &context.text[self.index];
        match text.chars().position(|c| c == '.') {
            Some(pos) => Version {
                major: text[0..pos].parse().unwrap(),
                minor: Some(text[pos + 1..text.len()].parse().unwrap()),
            },
            None => Version {
                major: text.parse().unwrap(),
                minor: None,
            },
        }
    }
}

/// The workhouse struct of the lexer.  This represents a peekable iterable object that is abstract
/// over some buffered reader.  The struct itself essentially represents the mutable state of the
/// lexer, with its main public associated functions being the iterable method [Self::next()] and
/// the [std::iter::Peekable]-like function [Self::peek()].
///
/// The stream exposes one public attributes directly: the [filename] that this stream comes from
/// (set to some placeholder value for streams that do not have a backing file).  The associated
/// `TokenContext` object is managed separately to the stream and is passed in each call to `next`;
/// this allows for multiple streams to operate on the same context, such as when a new stream
/// begins in order to handle an `include` statement.
pub struct TokenStream {
    /// The filename from which this stream is derived.  May be a placeholder if there is no
    /// backing file or other named resource.
    pub filename: std::ffi::OsString,
    strict: bool,
    source: Box<dyn std::io::BufRead + Send>,
    line_buffer: Vec<u8>,
    done: bool,
    line: usize,
    col: usize,
    try_version: bool,
    // This is a manual peekable structure (rather than using the `peekable` method of `Iterator`)
    // because we still want to be able to access the other members of the struct at the same time.
    peeked: Option<Option<Token>>,
}

impl TokenStream {
    /// Create and initialise a generic [TokenStream], given a source that implements
    /// [std::io::BufRead] and a filename (or resource path) that describes its source.
    fn new(
        source: Box<dyn std::io::BufRead + Send>,
        filename: std::ffi::OsString,
        strict: bool,
    ) -> Self {
        TokenStream {
            filename,
            strict,
            source,
            line_buffer: Vec::with_capacity(80),
            done: false,
            // The first line is numbered "1", and the first column is "0".  The counts are
            // initialised like this so the first call to `next_byte` can easily detect that it
            // needs to extract the next line.
            line: 0,
            col: 0,
            try_version: false,
            peeked: None,
        }
    }

    /// Create a [TokenStream] from a string containing the OpenQASM 2 program.
    pub fn from_string(string: String, strict: bool) -> Self {
        TokenStream::new(
            Box::new(std::io::Cursor::new(string)),
            "<input>".into(),
            strict,
        )
    }

    /// Create a [TokenStream] from a path containing the OpenQASM 2 program.
    pub fn from_path<P: AsRef<Path>>(path: P, strict: bool) -> Result<Self, std::io::Error> {
        let file = std::fs::File::open(path.as_ref())?;
        Ok(TokenStream::new(
            Box::new(std::io::BufReader::new(file)),
            Path::file_name(path.as_ref()).unwrap().into(),
            strict,
        ))
    }

    /// Read the next line into the managed buffer in the struct, updating the tracking information
    /// of the position, and the `done` state of the iterator.
    fn advance_line(&mut self) -> PyResult<usize> {
        if self.done {
            Ok(0)
        } else {
            self.line += 1;
            self.col = 0;
            self.line_buffer.clear();
            // We can assume that nobody's running this on ancient Mac software that uses only '\r'
            // as its linebreak character.
            match self.source.read_until(b'\n', &mut self.line_buffer) {
                Ok(count) => {
                    if count == 0 || self.line_buffer[count - 1] != b'\n' {
                        self.done = true;
                    }
                    Ok(count)
                }
                Err(err) => {
                    self.done = true;
                    Err(QASM2ParseError::new_err(message_generic(
                        Some(&Position::new(&self.filename, self.line, self.col)),
                        &format!("lexer failed to read stream: {}", err),
                    )))
                }
            }
        }
    }

    /// Get the next character in the stream.  This updates the line and column information for the
    /// current byte as well.
    fn next_byte(&mut self) -> PyResult<Option<u8>> {
        if self.col >= self.line_buffer.len() && self.advance_line()? == 0 {
            return Ok(None);
        }
        let out = self.line_buffer[self.col];
        self.col += 1;
        match out {
            b @ 0x80..=0xff => {
                self.done = true;
                Err(QASM2ParseError::new_err(message_generic(
                    Some(&Position::new(&self.filename, self.line, self.col)),
                    &format!("encountered a non-ASCII byte: {:02X?}", b),
                )))
            }
            b => Ok(Some(b)),
        }
    }

    /// Peek at the next byte in the stream without consuming it.  This still returns an error if
    /// the next byte isn't in the valid range for OpenQASM 2, or if the file/stream has failed to
    /// read into the buffer for some reason.
    fn peek_byte(&mut self) -> PyResult<Option<u8>> {
        if self.col >= self.line_buffer.len() && self.advance_line()? == 0 {
            return Ok(None);
        }
        match self.line_buffer[self.col] {
            b @ 0x80..=0xff => {
                self.done = true;
                Err(QASM2ParseError::new_err(message_generic(
                    Some(&Position::new(&self.filename, self.line, self.col)),
                    &format!("encountered a non-ASCII byte: {:02X?}", b),
                )))
            }
            b => Ok(Some(b)),
        }
    }

    /// Expect that the next byte is not a word continuation, providing a suitable error message if
    /// it is.
    fn expect_word_boundary(&mut self, after: &str, start_col: usize) -> PyResult<()> {
        match self.peek_byte()? {
            Some(c @ (b'a'..=b'z' | b'A'..=b'Z' | b'0'..=b'9' | b'_')) => {
                Err(QASM2ParseError::new_err(message_generic(
                    Some(&Position::new(&self.filename, self.line, start_col)),
                    &format!(
                        "expected a word boundary after {}, but saw '{}'",
                        after, c as char
                    ),
                )))
            }
            _ => Ok(()),
        }
    }

    /// Complete the lexing of a floating-point value from the position of maybe accepting an
    /// exponent.  The previous part of the token must be a valid stand-alone float, or the next
    /// byte must already have been peeked and known to be `b'e' | b'E'`.
    fn lex_float_exponent(&mut self, start_col: usize) -> PyResult<TokenType> {
        if !matches!(self.peek_byte()?, Some(b'e' | b'E')) {
            self.expect_word_boundary("a float", start_col)?;
            return Ok(TokenType::Real);
        }
        // Consume the rest of the exponent.
        self.next_byte()?;
        if let Some(b'+' | b'-') = self.peek_byte()? {
            self.next_byte()?;
        }
        // Exponents must have at least one digit in them.
        if !matches!(self.peek_byte()?, Some(b'0'..=b'9')) {
            return Err(QASM2ParseError::new_err(message_generic(
                Some(&Position::new(&self.filename, self.line, start_col)),
                "needed to see an integer exponent for this float",
            )));
        }
        while let Some(b'0'..=b'9') = self.peek_byte()? {
            self.next_byte()?;
        }
        self.expect_word_boundary("a float", start_col)?;
        Ok(TokenType::Real)
    }

    /// Lex a numeric token completely.  This can return a successful integer or a real number; the
    /// function distinguishes based on what it sees.  If `self.try_version`, this can also be a
    /// version identifier (will take precedence over either other type, if possible).
    fn lex_numeric(&mut self, start_col: usize) -> PyResult<TokenType> {
        let first = self.line_buffer[start_col];
        if first == b'.' {
            return match self.next_byte()? {
                // In the case of a float that begins with '.', we require at least one digit, so
                // just force consume it and then loop over the rest.
                Some(b'0'..=b'9') => {
                    while let Some(b'0'..=b'9') = self.peek_byte()? {
                        self.next_byte()?;
                    }
                    self.lex_float_exponent(start_col)
                }
                _ => Err(QASM2ParseError::new_err(message_generic(
                    Some(&Position::new(&self.filename, self.line, start_col)),
                    "expected a numeric fractional part after the bare decimal point",
                ))),
            };
        }
        while let Some(b'0'..=b'9') = self.peek_byte()? {
            self.next_byte()?;
        }
        match self.peek_byte()? {
            Some(b'.') => {
                self.next_byte()?;
                let mut has_fractional = false;
                while let Some(b'0'..=b'9') = self.peek_byte()? {
                    has_fractional = true;
                    self.next_byte()?;
                }
                if self.try_version
                    && has_fractional
                    && !matches!(self.peek_byte()?, Some(b'e' | b'E'))
                {
                    self.expect_word_boundary("a version identifier", start_col)?;
                    return Ok(TokenType::Version);
                }
                return self.lex_float_exponent(start_col);
            }
            // In this situation, what we've lexed so far is an integer (maybe with leading
            // zeroes), but it can still be a float if it's followed by an exponent.  This
            // particular path is not technically within the spec (so should be subject to `strict`
            // mode), but pragmatically that's more just a nuisance for OQ2 generators, since many
            // languages will happily spit out something like `5e-5` when formatting floats.
            Some(b'e' | b'E') => {
                return if self.strict {
                    Err(QASM2ParseError::new_err(message_generic(
                        Some(&Position::new(&self.filename, self.line, start_col)),
                        "[strict] all floats must include a decimal point",
                    )))
                } else {
                    self.lex_float_exponent(start_col)
                }
            }
            _ => (),
        }
        if first == b'0' && self.col - start_col > 1 {
            // Integers can't start with a leading zero unless they are only the single '0', but we
            // didn't see a decimal point.
            Err(QASM2ParseError::new_err(message_generic(
                Some(&Position::new(&self.filename, self.line, start_col)),
                "integers cannot have leading zeroes",
            )))
        } else if self.try_version {
            self.expect_word_boundary("a version identifier", start_col)?;
            Ok(TokenType::Version)
        } else {
            self.expect_word_boundary("an integer", start_col)?;
            Ok(TokenType::Integer)
        }
    }

    /// Lex a text-like token into a complete token.  This can return any of the keyword-like
    /// tokens (e.g. [TokenType::Pi]), or a [TokenType::Id] if the token is not a built-in keyword.
    fn lex_textlike(&mut self, start_col: usize) -> PyResult<TokenType> {
        let first = self.line_buffer[start_col];
        while let Some(b'a'..=b'z' | b'A'..=b'Z' | b'0'..=b'9' | b'_') = self.peek_byte()? {
            self.next_byte()?;
        }
        // No need to expect the word boundary after this, because it's the same check as above.
        let text = &self.line_buffer[start_col..self.col];
        if let b'A'..=b'Z' = first {
            match text {
                b"OPENQASM" => Ok(TokenType::OpenQASM),
                b"U" | b"CX" => Ok(TokenType::Id),
                _ => Err(QASM2ParseError::new_err(message_generic(
                        Some(&Position::new(&self.filename, self.line, start_col)),
                        "identifiers cannot start with capital letters except for the builtins 'U' and 'CX'"))),
            }
        } else {
            match text {
                b"barrier" => Ok(TokenType::Barrier),
                b"cos" => Ok(TokenType::Cos),
                b"creg" => Ok(TokenType::Creg),
                b"exp" => Ok(TokenType::Exp),
                b"gate" => Ok(TokenType::Gate),
                b"if" => Ok(TokenType::If),
                b"include" => Ok(TokenType::Include),
                b"ln" => Ok(TokenType::Ln),
                b"measure" => Ok(TokenType::Measure),
                b"opaque" => Ok(TokenType::Opaque),
                b"qreg" => Ok(TokenType::Qreg),
                b"reset" => Ok(TokenType::Reset),
                b"sin" => Ok(TokenType::Sin),
                b"sqrt" => Ok(TokenType::Sqrt),
                b"tan" => Ok(TokenType::Tan),
                b"pi" => Ok(TokenType::Pi),
                _ => Ok(TokenType::Id),
            }
        }
    }

    /// Lex a filename token completely.  This is always triggered by seeing a `b'"'` byte in the
    /// input stream.
    fn lex_filename(&mut self, terminator: u8, start_col: usize) -> PyResult<TokenType> {
        loop {
            match self.next_byte()? {
                None => {
                    return Err(QASM2ParseError::new_err(message_generic(
                        Some(&Position::new(&self.filename, self.line, start_col)),
                        "unexpected end-of-file while lexing string literal",
                    )))
                }
                Some(b'\n' | b'\r') => {
                    return Err(QASM2ParseError::new_err(message_generic(
                        Some(&Position::new(&self.filename, self.line, start_col)),
                        "unexpected line break while lexing string literal",
                    )))
                }
                Some(c) if c == terminator => {
                    return Ok(TokenType::Filename);
                }
                Some(_) => (),
            }
        }
    }

    /// The actual core of the iterator.  Read from the stream (ignoring preceding whitespace)
    /// until a complete [Token] has been constructed, or the end of the iterator is reached.  This
    /// returns `Some` for all tokens, including the error token, and only returns `None` if there
    /// are no more tokens left to take.
    fn next_inner(&mut self, context: &mut TokenContext) -> PyResult<Option<Token>> {
        // Consume preceding whitespace.  Beware that this can still exhaust the underlying stream,
        // or scan through an invalid token in the encoding.
        loop {
            match self.peek_byte()? {
                Some(b' ' | b'\t' | b'\r' | b'\n') => {
                    self.next_byte()?;
                }
                None => return Ok(None),
                _ => break,
            }
        }
        let start_col = self.col;
        // The whitespace loop (or [Self::try_lex_version]) has already peeked the next token, so
        // we know it's going to be the `Some` variant.
        let ttype = match self.next_byte()?.unwrap() {
            b'+' => TokenType::Plus,
            b'*' => TokenType::Asterisk,
            b'^' => TokenType::Caret,
            b';' => TokenType::Semicolon,
            b',' => TokenType::Comma,
            b'(' => TokenType::LParen,
            b')' => TokenType::RParen,
            b'[' => TokenType::LBracket,
            b']' => TokenType::RBracket,
            b'{' => TokenType::LBrace,
            b'}' => TokenType::RBrace,
            b'/' => {
                if let Some(b'/') = self.peek_byte()? {
                    self.advance_line()?;
                    return self.next(context);
                } else {
                    TokenType::Slash
                }
            }
            b'-' => {
                if let Ok(Some(b'>')) = self.peek_byte() {
                    self.col += 1;
                    TokenType::Arrow
                } else {
                    TokenType::Minus
                }
            }
            b'=' => {
                if let Ok(Some(b'=')) = self.peek_byte() {
                    self.col += 1;
                    TokenType::Equals
                } else {
                    return Err(QASM2ParseError::new_err(
                        "single equals '=' is never valid".to_owned(),
                    ));
                }
            }
            b'0'..=b'9' | b'.' => self.lex_numeric(start_col)?,
            b'a'..=b'z' | b'A'..=b'Z' => self.lex_textlike(start_col)?,
            c @ (b'"' | b'\'') => {
                if self.strict && c != b'"' {
                    return Err(QASM2ParseError::new_err(message_generic(
                        Some(&Position::new(&self.filename, self.line, start_col)),
                        "[strict] paths must be in double quotes (\"\")",
                    )));
                } else {
                    self.lex_filename(c, start_col)?
                }
            }
            c => {
                return Err(QASM2ParseError::new_err(message_generic(
                    Some(&Position::new(&self.filename, self.line, start_col)),
                    &format!(
                        "encountered '{}', which doesn't match any valid tokens",
                        // Non-ASCII bytes should already have been rejected by `next_byte()`.
                        c as char,
                    ),
                )));
            }
        };
        self.try_version = ttype == TokenType::OpenQASM;
        Ok(Some(Token {
            ttype,
            line: self.line,
            col: start_col,
            index: if ttype.variable_text() {
                context.index(&self.line_buffer[start_col..self.col])
            } else {
                usize::MAX
            },
        }))
    }

    /// Get an optional reference to the next token in the iterator stream without consuming it.
    /// This is a direct analogue of the same method on the [std::iter::Peekable] struct, except it
    /// is manually defined here to avoid hiding the rest of the public fields of the [TokenStream]
    /// struct itself.
    pub fn peek(&mut self, context: &mut TokenContext) -> PyResult<Option<&Token>> {
        if self.peeked.is_none() {
            self.peeked = Some(self.next_inner(context)?);
        }
        Ok(self.peeked.as_ref().unwrap().as_ref())
    }

    pub fn next(&mut self, context: &mut TokenContext) -> PyResult<Option<Token>> {
        match self.peeked.take() {
            Some(token) => Ok(token),
            None => self.next_inner(context),
        }
    }
}

Qiskit / qiskit-terra / 5895308192

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous