[rustc.git] / src / tools / clippy / clippy_utils / src / source.rs

//! Utils for extracting, inspecting or transforming source code

#![allow(clippy::module_name_repetitions)]

use rustc_errors::Applicability;
use rustc_hir::{Expr, ExprKind};
use rustc_lint::{LateContext, LintContext};
use rustc_session::Session;
use rustc_span::hygiene;
use rustc_span::source_map::{original_sp, SourceMap};
use rustc_span::{BytePos, Pos, Span, SpanData, SyntaxContext, DUMMY_SP};
use std::borrow::Cow;

/// Like `snippet_block`, but add braces if the expr is not an `ExprKind::Block`.
pub fn expr_block<T: LintContext>(
    cx: &T,
    expr: &Expr<'_>,
    outer: SyntaxContext,
    default: &str,
    indent_relative_to: Option<Span>,
    app: &mut Applicability,
) -> String {
    let (code, from_macro) = snippet_block_with_context(cx, expr.span, outer, default, indent_relative_to, app);
    if from_macro {
        format!("{{ {code} }}")
    } else if let ExprKind::Block(_, _) = expr.kind {
        format!("{code}")
    } else {
        format!("{{ {code} }}")
    }
}

/// Returns a new Span that extends the original Span to the first non-whitespace char of the first
/// line.
///
/// ```rust,ignore
///     let x = ();
/// //          ^^
/// // will be converted to
///     let x = ();
/// //  ^^^^^^^^^^
/// ```
pub fn first_line_of_span<T: LintContext>(cx: &T, span: Span) -> Span {
    first_char_in_first_line(cx, span).map_or(span, |first_char_pos| span.with_lo(first_char_pos))
}

fn first_char_in_first_line<T: LintContext>(cx: &T, span: Span) -> Option<BytePos> {
    let line_span = line_span(cx, span);
    snippet_opt(cx, line_span).and_then(|snip| {
        snip.find(|c: char| !c.is_whitespace())
            .map(|pos| line_span.lo() + BytePos::from_usize(pos))
    })
}

/// Extends the span to the beginning of the spans line, incl. whitespaces.
///
/// ```rust
///        let x = ();
/// //             ^^
/// // will be converted to
///        let x = ();
/// // ^^^^^^^^^^^^^^
/// ```
fn line_span<T: LintContext>(cx: &T, span: Span) -> Span {
    let span = original_sp(span, DUMMY_SP);
    let source_map_and_line = cx.sess().source_map().lookup_line(span.lo()).unwrap();
    let line_no = source_map_and_line.line;
    let line_start = source_map_and_line.sf.lines(|lines| lines[line_no]);
    span.with_lo(line_start)
}

/// Returns the indentation of the line of a span
///
/// ```rust,ignore
/// let x = ();
/// //      ^^ -- will return 0
///     let x = ();
/// //          ^^ -- will return 4
/// ```
pub fn indent_of<T: LintContext>(cx: &T, span: Span) -> Option<usize> {
    snippet_opt(cx, line_span(cx, span)).and_then(|snip| snip.find(|c: char| !c.is_whitespace()))
}

/// Gets a snippet of the indentation of the line of a span
pub fn snippet_indent<T: LintContext>(cx: &T, span: Span) -> Option<String> {
    snippet_opt(cx, line_span(cx, span)).map(|mut s| {
        let len = s.len() - s.trim_start().len();
        s.truncate(len);
        s
    })
}

// If the snippet is empty, it's an attribute that was inserted during macro
// expansion and we want to ignore those, because they could come from external
// sources that the user has no control over.
// For some reason these attributes don't have any expansion info on them, so
// we have to check it this way until there is a better way.
pub fn is_present_in_source<T: LintContext>(cx: &T, span: Span) -> bool {
    if let Some(snippet) = snippet_opt(cx, span) {
        if snippet.is_empty() {
            return false;
        }
    }
    true
}

/// Returns the position just before rarrow
///
/// ```rust,ignore
/// fn into(self) -> () {}
///              ^
/// // in case of unformatted code
/// fn into2(self)-> () {}
///               ^
/// fn into3(self)   -> () {}
///               ^
/// ```
pub fn position_before_rarrow(s: &str) -> Option<usize> {
    s.rfind("->").map(|rpos| {
        let mut rpos = rpos;
        let chars: Vec<char> = s.chars().collect();
        while rpos > 1 {
            if let Some(c) = chars.get(rpos - 1) {
                if c.is_whitespace() {
                    rpos -= 1;
                    continue;
                }
            }
            break;
        }
        rpos
    })
}

/// Reindent a multiline string with possibility of ignoring the first line.
#[expect(clippy::needless_pass_by_value)]
pub fn reindent_multiline(s: Cow<'_, str>, ignore_first: bool, indent: Option<usize>) -> Cow<'_, str> {
    let s_space = reindent_multiline_inner(&s, ignore_first, indent, ' ');
    let s_tab = reindent_multiline_inner(&s_space, ignore_first, indent, '\t');
    reindent_multiline_inner(&s_tab, ignore_first, indent, ' ').into()
}

fn reindent_multiline_inner(s: &str, ignore_first: bool, indent: Option<usize>, ch: char) -> String {
    let x = s
        .lines()
        .skip(usize::from(ignore_first))
        .filter_map(|l| {
            if l.is_empty() {
                None
            } else {
                // ignore empty lines
                Some(l.char_indices().find(|&(_, x)| x != ch).unwrap_or((l.len(), ch)).0)
            }
        })
        .min()
        .unwrap_or(0);
    let indent = indent.unwrap_or(0);
    s.lines()
        .enumerate()
        .map(|(i, l)| {
            if (ignore_first && i == 0) || l.is_empty() {
                l.to_owned()
            } else if x > indent {
                l.split_at(x - indent).1.to_owned()
            } else {
                " ".repeat(indent - x) + l
            }
        })
        .collect::<Vec<String>>()
        .join("\n")
}

/// Converts a span to a code snippet if available, otherwise returns the default.
///
/// This is useful if you want to provide suggestions for your lint or more generally, if you want
/// to convert a given `Span` to a `str`. To create suggestions consider using
/// [`snippet_with_applicability`] to ensure that the applicability stays correct.
///
/// # Example
/// ```rust,ignore
/// // Given two spans one for `value` and one for the `init` expression.
/// let value = Vec::new();
/// //  ^^^^^   ^^^^^^^^^^
/// //  span1   span2
///
/// // The snipped call would return the corresponding code snippet
/// snippet(cx, span1, "..") // -> "value"
/// snippet(cx, span2, "..") // -> "Vec::new()"
/// ```
pub fn snippet<'a, T: LintContext>(cx: &T, span: Span, default: &'a str) -> Cow<'a, str> {
    snippet_opt(cx, span).map_or_else(|| Cow::Borrowed(default), From::from)
}

/// Same as [`snippet`], but it adapts the applicability level by following rules:
///
/// - Applicability level `Unspecified` will never be changed.
/// - If the span is inside a macro, change the applicability level to `MaybeIncorrect`.
/// - If the default value is used and the applicability level is `MachineApplicable`, change it to
/// `HasPlaceholders`
pub fn snippet_with_applicability<'a, T: LintContext>(
    cx: &T,
    span: Span,
    default: &'a str,
    applicability: &mut Applicability,
) -> Cow<'a, str> {
    snippet_with_applicability_sess(cx.sess(), span, default, applicability)
}

fn snippet_with_applicability_sess<'a>(
    sess: &Session,
    span: Span,
    default: &'a str,
    applicability: &mut Applicability,
) -> Cow<'a, str> {
    if *applicability != Applicability::Unspecified && span.from_expansion() {
        *applicability = Applicability::MaybeIncorrect;
    }
    snippet_opt_sess(sess, span).map_or_else(
        || {
            if *applicability == Applicability::MachineApplicable {
                *applicability = Applicability::HasPlaceholders;
            }
            Cow::Borrowed(default)
        },
        From::from,
    )
}

/// Converts a span to a code snippet. Returns `None` if not available.
pub fn snippet_opt(cx: &impl LintContext, span: Span) -> Option<String> {
    snippet_opt_sess(cx.sess(), span)
}

fn snippet_opt_sess(sess: &Session, span: Span) -> Option<String> {
    sess.source_map().span_to_snippet(span).ok()
}

/// Converts a span (from a block) to a code snippet if available, otherwise use default.
///
/// This trims the code of indentation, except for the first line. Use it for blocks or block-like
/// things which need to be printed as such.
///
/// The `indent_relative_to` arg can be used, to provide a span, where the indentation of the
/// resulting snippet of the given span.
///
/// # Example
///
/// ```rust,ignore
/// snippet_block(cx, block.span, "..", None)
/// // where, `block` is the block of the if expr
///     if x {
///         y;
///     }
/// // will return the snippet
/// {
///     y;
/// }
/// ```
///
/// ```rust,ignore
/// snippet_block(cx, block.span, "..", Some(if_expr.span))
/// // where, `block` is the block of the if expr
///     if x {
///         y;
///     }
/// // will return the snippet
/// {
///         y;
///     } // aligned with `if`
/// ```
/// Note that the first line of the snippet always has 0 indentation.
pub fn snippet_block<'a, T: LintContext>(
    cx: &T,
    span: Span,
    default: &'a str,
    indent_relative_to: Option<Span>,
) -> Cow<'a, str> {
    let snip = snippet(cx, span, default);
    let indent = indent_relative_to.and_then(|s| indent_of(cx, s));
    reindent_multiline(snip, true, indent)
}

/// Same as `snippet_block`, but adapts the applicability level by the rules of
/// `snippet_with_applicability`.
pub fn snippet_block_with_applicability<'a>(
    cx: &impl LintContext,
    span: Span,
    default: &'a str,
    indent_relative_to: Option<Span>,
    applicability: &mut Applicability,
) -> Cow<'a, str> {
    let snip = snippet_with_applicability(cx, span, default, applicability);
    let indent = indent_relative_to.and_then(|s| indent_of(cx, s));
    reindent_multiline(snip, true, indent)
}

pub fn snippet_block_with_context<'a>(
    cx: &impl LintContext,
    span: Span,
    outer: SyntaxContext,
    default: &'a str,
    indent_relative_to: Option<Span>,
    app: &mut Applicability,
) -> (Cow<'a, str>, bool) {
    let (snip, from_macro) = snippet_with_context(cx, span, outer, default, app);
    let indent = indent_relative_to.and_then(|s| indent_of(cx, s));
    (reindent_multiline(snip, true, indent), from_macro)
}

/// Same as `snippet_with_applicability`, but first walks the span up to the given context. This
/// will result in the macro call, rather then the expansion, if the span is from a child context.
/// If the span is not from a child context, it will be used directly instead.
///
/// e.g. Given the expression `&vec![]`, getting a snippet from the span for `vec![]` as a HIR node
/// would result in `box []`. If given the context of the address of expression, this function will
/// correctly get a snippet of `vec![]`.
///
/// This will also return whether or not the snippet is a macro call.
pub fn snippet_with_context<'a>(
    cx: &impl LintContext,
    span: Span,
    outer: SyntaxContext,
    default: &'a str,
    applicability: &mut Applicability,
) -> (Cow<'a, str>, bool) {
    snippet_with_context_sess(cx.sess(), span, outer, default, applicability)
}

fn snippet_with_context_sess<'a>(
    sess: &Session,
    span: Span,
    outer: SyntaxContext,
    default: &'a str,
    applicability: &mut Applicability,
) -> (Cow<'a, str>, bool) {
    let (span, is_macro_call) = walk_span_to_context(span, outer).map_or_else(
        || {
            // The span is from a macro argument, and the outer context is the macro using the argument
            if *applicability != Applicability::Unspecified {
                *applicability = Applicability::MaybeIncorrect;
            }
            // TODO: get the argument span.
            (span, false)
        },
        |outer_span| (outer_span, span.ctxt() != outer),
    );

    (
        snippet_with_applicability_sess(sess, span, default, applicability),
        is_macro_call,
    )
}

/// Walks the span up to the target context, thereby returning the macro call site if the span is
/// inside a macro expansion, or the original span if it is not. Note this will return `None` in the
/// case of the span being in a macro expansion, but the target context is from expanding a macro
/// argument.
///
/// Given the following
///
/// ```rust,ignore
/// macro_rules! m { ($e:expr) => { f($e) }; }
/// g(m!(0))
/// ```
///
/// If called with a span of the call to `f` and a context of the call to `g` this will return a
/// span containing `m!(0)`. However, if called with a span of the literal `0` this will give a span
/// containing `0` as the context is the same as the outer context.
///
/// This will traverse through multiple macro calls. Given the following:
///
/// ```rust,ignore
/// macro_rules! m { ($e:expr) => { n!($e, 0) }; }
/// macro_rules! n { ($e:expr, $f:expr) => { f($e, $f) }; }
/// g(m!(0))
/// ```
///
/// If called with a span of the call to `f` and a context of the call to `g` this will return a
/// span containing `m!(0)`.
pub fn walk_span_to_context(span: Span, outer: SyntaxContext) -> Option<Span> {
    let outer_span = hygiene::walk_chain(span, outer);
    (outer_span.ctxt() == outer).then_some(outer_span)
}

/// Removes block comments from the given `Vec` of lines.
///
/// # Examples
///
/// ```rust,ignore
/// without_block_comments(vec!["/*", "foo", "*/"]);
/// // => vec![]
///
/// without_block_comments(vec!["bar", "/*", "foo", "*/"]);
/// // => vec!["bar"]
/// ```
pub fn without_block_comments(lines: Vec<&str>) -> Vec<&str> {
    let mut without = vec![];

    let mut nest_level = 0;

    for line in lines {
        if line.contains("/*") {
            nest_level += 1;
            continue;
        } else if line.contains("*/") {
            nest_level -= 1;
            continue;
        }

        if nest_level == 0 {
            without.push(line);
        }
    }

    without
}

/// Trims the whitespace from the start and the end of the span.
pub fn trim_span(sm: &SourceMap, span: Span) -> Span {
    let data = span.data();
    let sf: &_ = &sm.lookup_source_file(data.lo);
    let Some(src) = sf.src.as_deref() else {
        return span;
    };
    let Some(snip) = &src.get((data.lo - sf.start_pos).to_usize()..(data.hi - sf.start_pos).to_usize()) else {
        return span;
    };
    let trim_start = snip.len() - snip.trim_start().len();
    let trim_end = snip.len() - snip.trim_end().len();
    SpanData {
        lo: data.lo + BytePos::from_usize(trim_start),
        hi: data.hi - BytePos::from_usize(trim_end),
        ctxt: data.ctxt,
        parent: data.parent,
    }
    .span()
}

/// Expand a span to include a preceding comma
/// ```rust,ignore
/// writeln!(o, "")   ->   writeln!(o, "")
///             ^^                   ^^^^
/// ```
pub fn expand_past_previous_comma(cx: &LateContext<'_>, span: Span) -> Span {
    let extended = cx.sess().source_map().span_extend_to_prev_char(span, ',', true);
    extended.with_lo(extended.lo() - BytePos(1))
}

#[cfg(test)]
mod test {
    use super::{reindent_multiline, without_block_comments};

    #[test]
    fn test_reindent_multiline_single_line() {
        assert_eq!("", reindent_multiline("".into(), false, None));
        assert_eq!("...", reindent_multiline("...".into(), false, None));
        assert_eq!("...", reindent_multiline("    ...".into(), false, None));
        assert_eq!("...", reindent_multiline("\t...".into(), false, None));
        assert_eq!("...", reindent_multiline("\t\t...".into(), false, None));
    }

    #[test]
    #[rustfmt::skip]
    fn test_reindent_multiline_block() {
        assert_eq!("\
    if x {
        y
    } else {
        z
    }", reindent_multiline("    if x {
            y
        } else {
            z
        }".into(), false, None));
        assert_eq!("\
    if x {
    \ty
    } else {
    \tz
    }", reindent_multiline("    if x {
        \ty
        } else {
        \tz
        }".into(), false, None));
    }

    #[test]
    #[rustfmt::skip]
    fn test_reindent_multiline_empty_line() {
        assert_eq!("\
    if x {
        y

    } else {
        z
    }", reindent_multiline("    if x {
            y

        } else {
            z
        }".into(), false, None));
    }

    #[test]
    #[rustfmt::skip]
    fn test_reindent_multiline_lines_deeper() {
        assert_eq!("\
        if x {
            y
        } else {
            z
        }", reindent_multiline("\
    if x {
        y
    } else {
        z
    }".into(), true, Some(8)));
    }

    #[test]
    fn test_without_block_comments_lines_without_block_comments() {
        let result = without_block_comments(vec!["/*", "", "*/"]);
        println!("result: {result:?}");
        assert!(result.is_empty());

        let result = without_block_comments(vec!["", "/*", "", "*/", "#[crate_type = \"lib\"]", "/*", "", "*/", ""]);
        assert_eq!(result, vec!["", "#[crate_type = \"lib\"]", ""]);

        let result = without_block_comments(vec!["/* rust", "", "*/"]);
        assert!(result.is_empty());

        let result = without_block_comments(vec!["/* one-line comment */"]);
        assert!(result.is_empty());

        let result = without_block_comments(vec!["/* nested", "/* multi-line", "comment", "*/", "test", "*/"]);
        assert!(result.is_empty());

        let result = without_block_comments(vec!["/* nested /* inline /* comment */ test */ */"]);
        assert!(result.is_empty());

        let result = without_block_comments(vec!["foo", "bar", "baz"]);
        assert_eq!(result, vec!["foo", "bar", "baz"]);
    }
}