ruff/crates/ruff_formatter/src/lib.rs

//! Infrastructure for code formatting
//!
//! This module defines [FormatElement], an IR to format code documents and provides a mean to print
//! such a document to a string. Objects that know how to format themselves implement the [Format] trait.
//!
//! ## Formatting Traits
//!
//! * [Format]: Implemented by objects that can be formatted.
//! * [FormatRule]: Rule that knows how to format an object of another type. Necessary in the situation where
//!  it's necessary to implement [Format] on an object from another crate. This module defines the
//!  [FormatRefWithRule] and [FormatOwnedWithRule] structs to pass an item with its corresponding rule.
//! * [FormatWithRule] implemented by objects that know how to format another type. Useful for implementing
//!  some reusable formatting logic inside of this module if the type itself doesn't implement [Format]
//!
//! ## Formatting Macros
//!
//! This crate defines two macros to construct the IR. These are inspired by Rust's `fmt` macros
//! * [`format!`]: Formats a formatable object
//! * [`format_args!`]: Concatenates a sequence of Format objects.
//! * [`write!`]: Writes a sequence of formatable objects into an output buffer.

#![allow(clippy::pedantic, unsafe_code)]
#![deny(rustdoc::broken_intra_doc_links)]

mod arguments;
mod buffer;
mod builders;
pub mod diagnostics;
pub mod format_element;
mod format_extensions;
pub mod formatter;
pub mod group_id;
pub mod macros;
pub mod prelude;
pub mod printer;
mod source_code;

use crate::formatter::Formatter;
use crate::group_id::UniqueGroupIdBuilder;
use crate::prelude::TagKind;
use std::fmt::Debug;

use crate::format_element::document::Document;
use crate::printer::{Printer, PrinterOptions};
pub use arguments::{Argument, Arguments};
pub use buffer::{
    Buffer, BufferExtensions, BufferSnapshot, Inspect, PreambleBuffer, RemoveSoftLinesBuffer,
    VecBuffer,
};
pub use builders::BestFitting;
pub use source_code::{SourceCode, SourceCodeSlice};

pub use crate::diagnostics::{ActualStart, FormatError, InvalidDocumentError, PrintError};
pub use format_element::{normalize_newlines, FormatElement, LINE_TERMINATORS};
pub use group_id::GroupId;
use ruff_text_size::{TextRange, TextSize};
use std::num::ParseIntError;
use std::str::FromStr;

#[derive(Debug, Eq, PartialEq, Clone, Copy, Hash)]
#[cfg_attr(
    feature = "serde",
    derive(serde::Serialize, serde::Deserialize, schemars::JsonSchema)
)]
#[derive(Default)]
pub enum IndentStyle {
    /// Tab
    #[default]
    Tab,
    /// Space, with its quantity
    Space(u8),
}

impl IndentStyle {
    pub const DEFAULT_SPACES: u8 = 2;

    /// Returns `true` if this is an [IndentStyle::Tab].
    pub const fn is_tab(&self) -> bool {
        matches!(self, IndentStyle::Tab)
    }

    /// Returns `true` if this is an [IndentStyle::Space].
    pub const fn is_space(&self) -> bool {
        matches!(self, IndentStyle::Space(_))
    }
}

impl FromStr for IndentStyle {
    type Err = &'static str;

    fn from_str(s: &str) -> Result<Self, Self::Err> {
        match s {
            "tab" | "Tabs" => Ok(Self::Tab),
            "space" | "Spaces" => Ok(Self::Space(IndentStyle::DEFAULT_SPACES)),
            // TODO: replace this error with a diagnostic
            _ => Err("Value not supported for IndentStyle"),
        }
    }
}

impl std::fmt::Display for IndentStyle {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            IndentStyle::Tab => std::write!(f, "Tab"),
            IndentStyle::Space(size) => std::write!(f, "Spaces, size: {size}"),
        }
    }
}

/// Validated value for the `line_width` formatter options
///
/// The allowed range of values is 1..=320
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
#[cfg_attr(
    feature = "serde",
    derive(serde::Serialize, serde::Deserialize, schemars::JsonSchema)
)]
pub struct LineWidth(u16);

impl LineWidth {
    /// Maximum allowed value for a valid [LineWidth]
    pub const MAX: u16 = 320;

    /// Return the numeric value for this [LineWidth]
    pub fn value(&self) -> u16 {
        self.0
    }
}

impl Default for LineWidth {
    fn default() -> Self {
        Self(80)
    }
}

/// Error type returned when parsing a [LineWidth] from a string fails
pub enum ParseLineWidthError {
    /// The string could not be parsed as a valid [u16]
    ParseError(ParseIntError),
    /// The [u16] value of the string is not a valid [LineWidth]
    TryFromIntError(LineWidthFromIntError),
}

impl Debug for ParseLineWidthError {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        std::fmt::Display::fmt(self, f)
    }
}

impl std::fmt::Display for ParseLineWidthError {
    fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            ParseLineWidthError::ParseError(err) => std::fmt::Display::fmt(err, fmt),
            ParseLineWidthError::TryFromIntError(err) => std::fmt::Display::fmt(err, fmt),
        }
    }
}

impl FromStr for LineWidth {
    type Err = ParseLineWidthError;

    fn from_str(s: &str) -> Result<Self, Self::Err> {
        let value = u16::from_str(s).map_err(ParseLineWidthError::ParseError)?;
        let value = Self::try_from(value).map_err(ParseLineWidthError::TryFromIntError)?;
        Ok(value)
    }
}

/// Error type returned when converting a u16 to a [LineWidth] fails
#[derive(Clone, Copy, Debug)]
pub struct LineWidthFromIntError(pub u16);

impl TryFrom<u16> for LineWidth {
    type Error = LineWidthFromIntError;

    fn try_from(value: u16) -> Result<Self, Self::Error> {
        if value > 0 && value <= Self::MAX {
            Ok(Self(value))
        } else {
            Err(LineWidthFromIntError(value))
        }
    }
}

impl std::fmt::Display for LineWidthFromIntError {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        writeln!(
            f,
            "The line width exceeds the maximum value ({})",
            LineWidth::MAX
        )
    }
}

impl From<LineWidth> for u16 {
    fn from(value: LineWidth) -> Self {
        value.0
    }
}

/// Context object storing data relevant when formatting an object.
pub trait FormatContext {
    type Options: FormatOptions;

    /// Returns the formatting options
    fn options(&self) -> &Self::Options;

    /// Returns the source code from the document that gets formatted.
    fn source_code(&self) -> SourceCode;
}

/// Options customizing how the source code should be formatted.
pub trait FormatOptions {
    /// The indent style.
    fn indent_style(&self) -> IndentStyle;

    /// What's the max width of a line. Defaults to 80.
    fn line_width(&self) -> LineWidth;

    /// Derives the print options from the these format options
    fn as_print_options(&self) -> PrinterOptions;
}

#[derive(Debug, Default, Eq, PartialEq)]
pub struct SimpleFormatContext {
    options: SimpleFormatOptions,
    source_code: String,
}

impl SimpleFormatContext {
    pub fn new(options: SimpleFormatOptions) -> Self {
        Self {
            options,
            source_code: String::new(),
        }
    }

    pub fn with_source_code(mut self, code: &str) -> Self {
        self.source_code = String::from(code);
        self
    }
}

impl FormatContext for SimpleFormatContext {
    type Options = SimpleFormatOptions;

    fn options(&self) -> &Self::Options {
        &self.options
    }

    fn source_code(&self) -> SourceCode {
        SourceCode::new(&self.source_code)
    }
}

#[derive(Debug, Default, Eq, PartialEq, Clone)]
pub struct SimpleFormatOptions {
    pub indent_style: IndentStyle,
    pub line_width: LineWidth,
}

impl FormatOptions for SimpleFormatOptions {
    fn indent_style(&self) -> IndentStyle {
        self.indent_style
    }

    fn line_width(&self) -> LineWidth {
        self.line_width
    }

    fn as_print_options(&self) -> PrinterOptions {
        PrinterOptions::default()
            .with_indent(self.indent_style)
            .with_print_width(self.line_width.into())
    }
}

/// Lightweight sourcemap marker between source and output tokens
#[derive(Debug, Copy, Clone, Eq, PartialEq)]
#[cfg_attr(
    feature = "serde",
    derive(serde::Serialize, serde::Deserialize, schemars::JsonSchema)
)]
pub struct SourceMarker {
    /// Position of the marker in the original source
    pub source: TextSize,
    /// Position of the marker in the output code
    pub dest: TextSize,
}

#[derive(Debug, Clone, Eq, PartialEq)]
pub struct Formatted<Context> {
    document: Document,
    context: Context,
}

impl<Context> Formatted<Context> {
    pub fn new(document: Document, context: Context) -> Self {
        Self { document, context }
    }

    /// Returns the context used during formatting.
    pub fn context(&self) -> &Context {
        &self.context
    }

    /// Returns the formatted document.
    pub fn document(&self) -> &Document {
        &self.document
    }

    /// Consumes `self` and returns the formatted document.
    pub fn into_document(self) -> Document {
        self.document
    }
}

impl<Context> Formatted<Context>
where
    Context: FormatContext,
{
    pub fn print(&self) -> PrintResult<Printed> {
        let source_code = self.context.source_code();
        let print_options = self.context.options().as_print_options();
        let printed = Printer::new(source_code, print_options).print(&self.document)?;

        Ok(printed)
    }

    pub fn print_with_indent(&self, indent: u16) -> PrintResult<Printed> {
        let source_code = self.context.source_code();
        let print_options = self.context.options().as_print_options();
        let printed =
            Printer::new(source_code, print_options).print_with_indent(&self.document, indent)?;

        Ok(printed)
    }
}
pub type PrintResult<T> = Result<T, PrintError>;

#[derive(Debug, Clone, Eq, PartialEq)]
#[cfg_attr(
    feature = "serde",
    derive(serde::Serialize, serde::Deserialize, schemars::JsonSchema)
)]
pub struct Printed {
    code: String,
    range: Option<TextRange>,
    sourcemap: Vec<SourceMarker>,
    verbatim_ranges: Vec<TextRange>,
}

impl Printed {
    pub fn new(
        code: String,
        range: Option<TextRange>,
        sourcemap: Vec<SourceMarker>,
        verbatim_source: Vec<TextRange>,
    ) -> Self {
        Self {
            code,
            range,
            sourcemap,
            verbatim_ranges: verbatim_source,
        }
    }

    /// Construct an empty formatter result
    pub fn new_empty() -> Self {
        Self {
            code: String::new(),
            range: None,
            sourcemap: Vec::new(),
            verbatim_ranges: Vec::new(),
        }
    }

    /// Range of the input source file covered by this formatted code,
    /// or None if the entire file is covered in this instance
    pub fn range(&self) -> Option<TextRange> {
        self.range
    }

    /// Returns a list of [SourceMarker] mapping byte positions
    /// in the output string to the input source code.
    /// It's not guaranteed that the markers are sorted by source position.
    pub fn sourcemap(&self) -> &[SourceMarker] {
        &self.sourcemap
    }

    /// Returns a list of [SourceMarker] mapping byte positions
    /// in the output string to the input source code, consuming the result
    pub fn into_sourcemap(self) -> Vec<SourceMarker> {
        self.sourcemap
    }

    /// Takes the list of [SourceMarker] mapping byte positions in the output string
    /// to the input source code.
    pub fn take_sourcemap(&mut self) -> Vec<SourceMarker> {
        std::mem::take(&mut self.sourcemap)
    }

    /// Access the resulting code, borrowing the result
    pub fn as_code(&self) -> &str {
        &self.code
    }

    /// Access the resulting code, consuming the result
    pub fn into_code(self) -> String {
        self.code
    }

    /// The text in the formatted code that has been formatted as verbatim.
    pub fn verbatim(&self) -> impl Iterator<Item = (TextRange, &str)> {
        self.verbatim_ranges
            .iter()
            .map(|range| (*range, &self.code[*range]))
    }

    /// Ranges of the formatted code that have been formatted as verbatim.
    pub fn verbatim_ranges(&self) -> &[TextRange] {
        &self.verbatim_ranges
    }

    /// Takes the ranges of nodes that have been formatted as verbatim, replacing them with an empty list.
    pub fn take_verbatim_ranges(&mut self) -> Vec<TextRange> {
        std::mem::take(&mut self.verbatim_ranges)
    }
}

/// Public return type of the formatter
pub type FormatResult<F> = Result<F, FormatError>;

/// Formatting trait for types that can create a formatted representation. The `ruff_formatter` equivalent
/// to [std::fmt::Display].
///
/// ## Example
/// Implementing `Format` for a custom struct
///
/// ```
/// use ruff_formatter::{format, write, IndentStyle, LineWidth};
/// use ruff_formatter::prelude::*;
/// use ruff_text_size::TextSize;
///
/// struct Paragraph(String);
///
/// impl Format<SimpleFormatContext> for Paragraph {
///     fn fmt(&self, f: &mut Formatter<SimpleFormatContext>) -> FormatResult<()> {
///         write!(f, [
///             hard_line_break(),
///             dynamic_text(&self.0, None),
///             hard_line_break(),
///         ])
///     }
/// }
///
/// # fn main() -> FormatResult<()> {
/// let paragraph = Paragraph(String::from("test"));
/// let formatted = format!(SimpleFormatContext::default(), [paragraph])?;
///
/// assert_eq!("test\n", formatted.print()?.as_code());
/// # Ok(())
/// # }
/// ```
pub trait Format<Context> {
    /// Formats the object using the given formatter.
    fn fmt(&self, f: &mut Formatter<Context>) -> FormatResult<()>;
}

impl<T, Context> Format<Context> for &T
where
    T: ?Sized + Format<Context>,
{
    #[inline(always)]
    fn fmt(&self, f: &mut Formatter<Context>) -> FormatResult<()> {
        Format::fmt(&**self, f)
    }
}

impl<T, Context> Format<Context> for &mut T
where
    T: ?Sized + Format<Context>,
{
    #[inline(always)]
    fn fmt(&self, f: &mut Formatter<Context>) -> FormatResult<()> {
        Format::fmt(&**self, f)
    }
}

impl<T, Context> Format<Context> for Option<T>
where
    T: Format<Context>,
{
    fn fmt(&self, f: &mut Formatter<Context>) -> FormatResult<()> {
        match self {
            Some(value) => value.fmt(f),
            None => Ok(()),
        }
    }
}

impl<Context> Format<Context> for () {
    #[inline]
    fn fmt(&self, _: &mut Formatter<Context>) -> FormatResult<()> {
        // Intentionally left empty
        Ok(())
    }
}

/// Rule that knows how to format an object of type `T`.
///
/// Implementing [Format] on the object itself is preferred over implementing [FormatRule] but
/// this isn't possible inside of a dependent crate for external type.
///
/// For example, the `ruff_js_formatter` crate isn't able to implement [Format] on `JsIfStatement`
/// because both the [Format] trait and `JsIfStatement` are external types (Rust's orphan rule).
///
/// That's why the `ruff_js_formatter` crate must define a new-type that implements the formatting
/// of `JsIfStatement`.
pub trait FormatRule<T> {
    type Context;

    fn fmt(&self, item: &T, f: &mut Formatter<Self::Context>) -> FormatResult<()>;
}

/// Rule that supports customizing how it formats an object of type `T`.
pub trait FormatRuleWithOptions<T>: FormatRule<T> {
    type Options;

    /// Returns a new rule that uses the given options to format an object.
    fn with_options(self, options: Self::Options) -> Self;
}

/// Trait for an object that formats an object with a specified rule.
///
/// Gives access to the underlying item.
///
/// Useful in situation where a type itself doesn't implement [Format] (e.g. because of Rust's orphan rule)
/// but you want to implement some common formatting logic.
///
/// ## Examples
///
/// This can be useful if you want to format a `SyntaxNode` inside ruff_formatter.. `SyntaxNode` doesn't implement [Format]
/// itself but the language specific crate implements `AsFormat` and `IntoFormat` for it and the returned [Format]
/// implement [FormatWithRule].
///
/// ```ignore
/// use ruff_formatter::prelude::*;
/// use ruff_formatter::{format, Formatted, FormatWithRule};
/// use ruff_rowan::{Language, SyntaxNode};
/// fn format_node<L: Language, F: FormatWithRule<SimpleFormatContext, Item=SyntaxNode<L>>>(node: F) -> FormatResult<Formatted<SimpleFormatContext>> {
///     let formatted = format!(SimpleFormatContext::default(), [node]);
///     let syntax = node.item();
///     // Do something with syntax
///     formatted;
/// }
/// ```
pub trait FormatWithRule<Context>: Format<Context> {
    type Item;

    /// Returns the associated item
    fn item(&self) -> &Self::Item;
}

/// Formats the referenced `item` with the specified rule.
#[derive(Debug, Copy, Clone)]
pub struct FormatRefWithRule<'a, T, R>
where
    R: FormatRule<T>,
{
    item: &'a T,
    rule: R,
}

impl<'a, T, R> FormatRefWithRule<'a, T, R>
where
    R: FormatRule<T>,
{
    pub fn new(item: &'a T, rule: R) -> Self {
        Self { item, rule }
    }
}

impl<T, R, O> FormatRefWithRule<'_, T, R>
where
    R: FormatRuleWithOptions<T, Options = O>,
{
    pub fn with_options(mut self, options: O) -> Self {
        self.rule = self.rule.with_options(options);
        self
    }
}

impl<T, R> FormatWithRule<R::Context> for FormatRefWithRule<'_, T, R>
where
    R: FormatRule<T>,
{
    type Item = T;

    fn item(&self) -> &Self::Item {
        self.item
    }
}

impl<T, R> Format<R::Context> for FormatRefWithRule<'_, T, R>
where
    R: FormatRule<T>,
{
    #[inline(always)]
    fn fmt(&self, f: &mut Formatter<R::Context>) -> FormatResult<()> {
        self.rule.fmt(self.item, f)
    }
}

/// Formats the `item` with the specified rule.
#[derive(Debug, Clone)]
pub struct FormatOwnedWithRule<T, R>
where
    R: FormatRule<T>,
{
    item: T,
    rule: R,
}

impl<T, R> FormatOwnedWithRule<T, R>
where
    R: FormatRule<T>,
{
    pub fn new(item: T, rule: R) -> Self {
        Self { item, rule }
    }

    pub fn with_item(mut self, item: T) -> Self {
        self.item = item;
        self
    }

    pub fn into_item(self) -> T {
        self.item
    }
}

impl<T, R> Format<R::Context> for FormatOwnedWithRule<T, R>
where
    R: FormatRule<T>,
{
    #[inline(always)]
    fn fmt(&self, f: &mut Formatter<R::Context>) -> FormatResult<()> {
        self.rule.fmt(&self.item, f)
    }
}

impl<T, R, O> FormatOwnedWithRule<T, R>
where
    R: FormatRuleWithOptions<T, Options = O>,
{
    pub fn with_options(mut self, options: O) -> Self {
        self.rule = self.rule.with_options(options);
        self
    }
}

impl<T, R> FormatWithRule<R::Context> for FormatOwnedWithRule<T, R>
where
    R: FormatRule<T>,
{
    type Item = T;

    fn item(&self) -> &Self::Item {
        &self.item
    }
}

/// The `write` function takes a target buffer and an `Arguments` struct that can be precompiled with the `format_args!` macro.
///
/// The arguments will be formatted in-order into the output buffer provided.
///
/// # Examples
///
/// ```
/// use ruff_formatter::prelude::*;
/// use ruff_formatter::{VecBuffer, format_args, FormatState, write, Formatted};
///
/// # fn main() -> FormatResult<()> {
/// let mut state = FormatState::new(SimpleFormatContext::default());
/// let mut buffer = VecBuffer::new(&mut state);
///
/// write!(&mut buffer, [format_args!(text("Hello World"))])?;
///
/// let formatted = Formatted::new(Document::from(buffer.into_vec()), SimpleFormatContext::default());
///
/// assert_eq!("Hello World", formatted.print()?.as_code());
/// # Ok(())
/// # }
/// ```
///
/// Please note that using [`write!`] might be preferable. Example:
///
/// ```
/// use ruff_formatter::prelude::*;
/// use ruff_formatter::{VecBuffer, format_args, FormatState, write, Formatted};
///
/// # fn main() -> FormatResult<()> {
/// let mut state = FormatState::new(SimpleFormatContext::default());
/// let mut buffer = VecBuffer::new(&mut state);
///
/// write!(&mut buffer, [text("Hello World")])?;
///
/// let formatted = Formatted::new(Document::from(buffer.into_vec()), SimpleFormatContext::default());
///
/// assert_eq!("Hello World", formatted.print()?.as_code());
/// # Ok(())
/// # }
/// ```
///
#[inline(always)]
pub fn write<Context>(
    output: &mut dyn Buffer<Context = Context>,
    args: Arguments<Context>,
) -> FormatResult<()> {
    let mut f = Formatter::new(output);

    f.write_fmt(args)
}

/// The `format` function takes an [`Arguments`] struct and returns the resulting formatting IR.
///
/// The [`Arguments`] instance can be created with the [`format_args!`].
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// use ruff_formatter::prelude::*;
/// use ruff_formatter::{format, format_args};
///
/// # fn main() -> FormatResult<()> {
/// let formatted = format!(SimpleFormatContext::default(), [&format_args!(text("test"))])?;
/// assert_eq!("test", formatted.print()?.as_code());
/// # Ok(())
/// # }
/// ```
///
/// Please note that using [`format!`] might be preferable. Example:
///
/// ```
/// use ruff_formatter::prelude::*;
/// use ruff_formatter::{format};
///
/// # fn main() -> FormatResult<()> {
/// let formatted = format!(SimpleFormatContext::default(), [text("test")])?;
/// assert_eq!("test", formatted.print()?.as_code());
/// # Ok(())
/// # }
/// ```
pub fn format<Context>(
    context: Context,
    arguments: Arguments<Context>,
) -> FormatResult<Formatted<Context>>
where
    Context: FormatContext,
{
    let mut state = FormatState::new(context);
    let mut buffer = VecBuffer::with_capacity(arguments.items().len(), &mut state);

    buffer.write_fmt(arguments)?;

    let mut document = Document::from(buffer.into_vec());
    document.propagate_expand();

    Ok(Formatted::new(document, state.into_context()))
}

/// This structure stores the state that is relevant for the formatting of the whole document.
///
/// This structure is different from [crate::Formatter] in that the formatting infrastructure
/// creates a new [crate::Formatter] for every [crate::write!] call, whereas this structure stays alive
/// for the whole process of formatting a root with [crate::format!].
pub struct FormatState<Context> {
    context: Context,

    group_id_builder: UniqueGroupIdBuilder,
}

impl<Context> std::fmt::Debug for FormatState<Context>
where
    Context: std::fmt::Debug,
{
    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
        f.debug_struct("FormatState")
            .field("context", &self.context)
            .finish()
    }
}

impl<Context> FormatState<Context> {
    /// Creates a new state with the given language specific context
    pub fn new(context: Context) -> Self {
        Self {
            context,
            group_id_builder: Default::default(),
        }
    }

    pub fn into_context(self) -> Context {
        self.context
    }

    /// Returns the context specifying how to format the current CST
    pub fn context(&self) -> &Context {
        &self.context
    }

    /// Returns a mutable reference to the context
    pub fn context_mut(&mut self) -> &mut Context {
        &mut self.context
    }

    /// Creates a new group id that is unique to this document. The passed debug name is used in the
    /// [std::fmt::Debug] of the document if this is a debug build.
    /// The name is unused for production builds and has no meaning on the equality of two group ids.
    pub fn group_id(&self, debug_name: &'static str) -> GroupId {
        self.group_id_builder.group_id(debug_name)
    }
}