@@ -0,0 +1,350 @@
+// This file is part of the uutils diffutils package.
+//
+// For the full copyright and license information, please view the LICENSE-*
+// files that was distributed with this source code.
+
+//! Translation of POSIX basic regular expressions (the `grep` syntax that
+//! options such as `-F` take) into the syntax of the `regex` crate.
+//!
+//! In a basic regular expression, `\(`, `\)`, `\{`, `\}`, `\|`, `\+` and `\?`
+//! are operators while their unescaped forms are ordinary characters, `*` is
+//! literal where it has nothing to repeat, `^` only anchors at the start of
+//! the expression or of a group or alternative, and `$` only anchors at their
+//! end.
+
+use regex::bytes::Regex;
+
+/// A compiled basic regular expression. Two of them are equal when their
+/// patterns are.
+#[derive(Clone, Debug)]
+pub struct Bre {
+    pattern: String,
+    regex: Regex,
+}
+
+impl Bre {
+    /// Compiles `pattern`, or returns an error message describing what is
+    /// wrong with it.
+    pub fn new(pattern: &str) -> Result<Self, String> {
+        let regex = Regex::new(&translate(pattern)?).map_err(|error| match error {
+            regex::Error::CompiledTooBig(_) => TOO_BIG.to_string(),
+            _ => "Invalid regular expression".to_string(),
+        })?;
+        Ok(Self {
+            pattern: pattern.to_string(),
+            regex,
+        })
+    }
+
+    pub fn is_match(&self, text: &[u8]) -> bool {
+        self.regex.is_match(text)
+    }
+}
+
+impl PartialEq for Bre {
+    fn eq(&self, other: &Self) -> bool {
+        self.pattern == other.pattern
+    }
+}
+
+impl Eq for Bre {}
+
+const TOO_BIG: &str = "Regular expression too big";
+/// The largest count of an interval, as in `a\{2,100\}`.
+const MAX_REPETITIONS: u32 = 0x7fff;
+
+/// Translates a basic regular expression into `regex` crate syntax.
+fn translate(pattern: &str) -> Result<String, String> {
+    let chars: Vec<char> = pattern.chars().collect();
+    let mut translated = String::with_capacity(pattern.len() * 2);
+    // Whether the previous token opens an expression, a group or an
+    // alternative, where `^` anchors and repetition operators are literal.
+    let mut at_expression_start = true;
+    let mut open_groups = 0_usize;
+    let mut index = 0;
+    while index < chars.len() {
+        let c = chars[index];
+        index += 1;
+        let mut opens_expression = false;
+        match c {
+            '\\' => {
+                let Some(&escaped) = chars.get(index) else {
+                    return Err("Trailing backslash".to_string());
+                };
+                index += 1;
+                match escaped {
+                    '(' => {
+                        open_groups += 1;
+                        translated.push('(');
+                        opens_expression = true;
+                    }
+                    ')' => {
+                        open_groups = open_groups
+                            .checked_sub(1)
+                            .ok_or_else(|| r"Unmatched ) or \)".to_string())?;
+                        translated.push(')');
+                    }
+                    '|' => {
+                        translated.push('|');
+                        opens_expression = true;
+                    }
+                    '{' | '}' | '+' | '?' if at_expression_start => {
+                        push_literal(&mut translated, escaped);
+                    }
+                    '{' => index = translate_interval(&chars, index, &mut translated)?,
+                    '+' | '?' => translated.push(escaped),
+                    '<' | '>' | 'b' | 'B' | 'w' | 'W' | 's' | 'S' => {
+                        translated.push('\\');
+                        translated.push(escaped);
+                    }
+                    '`' => translated.push_str(r"\A"),
+                    '\'' => translated.push_str(r"\z"),
+                    '1'..='9' => return Err("Invalid back reference".to_string()),
+                    _ => push_literal(&mut translated, escaped),
+                }
+            }
+            '^' if at_expression_start => {
+                translated.push('^');
+                opens_expression = true;
+            }
+            '$' if ends_expression(&chars[index..]) => translated.push('$'),
+            '*' if at_expression_start => translated.push_str(r"\*"),
+            '.' | '*' => translated.push(c),
+            '[' => index = translate_bracket(&chars, index, &mut translated)?,
+            _ => push_literal(&mut translated, c),
+        }
+        at_expression_start = opens_expression;
+    }
+    if open_groups > 0 {
+        return Err(r"Unmatched ( or \(".to_string());
+    }
+    Ok(translated)
+}
+
+/// Returns whether a `$` followed by `rest` ends an expression, a group or an
+/// alternative.
+fn ends_expression(rest: &[char]) -> bool {
+    matches!(rest, [] | ['\\', ')' | '|', ..])
+}
+
+/// Translates the interval whose content starts at `index` (just after the
+/// `\{`), and returns the index following its closing `\}`.
+fn translate_interval(
+    chars: &[char],
+    index: usize,
+    translated: &mut String,
+) -> Result<usize, String> {
+    let end = (index..chars.len().saturating_sub(1))
+        .find(|&end| chars[end] == '\\' && chars[end + 1] == '}')
+        .ok_or_else(|| r"Unmatched \{".to_string())?;
+    let content: String = chars[index..end].iter().collect();
+    let invalid = || r"Invalid content of \{\}".to_string();
+    let count = |digits: &str| -> Result<Option<u32>, String> {
+        if digits.is_empty() {
+            return Ok(None);
+        }
+        if !digits.bytes().all(|b| b.is_ascii_digit()) {
+            return Err(invalid());
+        }
+        match digits.parse::<u32>() {
+            Ok(count) if count <= MAX_REPETITIONS => Ok(Some(count)),
+            _ => Err(TOO_BIG.to_string()),
+        }
+    };
+    match content.split_once(',') {
+        None => {
+            let exact = count(&content)?.ok_or_else(invalid)?;
+            translated.push_str(&format!("{{{exact}}}"));
+        }
+        Some((min, max)) => {
+            let min = count(min)?.unwrap_or(0);
+            match count(max)? {
+                Some(max) if max < min => return Err(invalid()),
+                Some(max) => translated.push_str(&format!("{{{min},{max}}}")),
+                None => translated.push_str(&format!("{{{min},}}")),
+            }
+        }
+    }
+    Ok(end + 2)
+}
+
+/// Characters with a meaning in `regex` crate syntax, inside or outside of
+/// bracket expressions.
+const META_CHARACTERS: &str = r"\.+*?()|[]{}^$#&-~";
+
+/// The names of the character classes of bracket expressions, as in
+/// `[[:alpha:]]`.
+const CLASS_NAMES: [&str; 12] = [
+    "alnum", "alpha", "blank", "cntrl", "digit", "graph", "lower", "print", "punct", "space",
+    "upper", "xdigit",
+];
+
+fn push_literal(translated: &mut String, c: char) {
+    if META_CHARACTERS.contains(c) {
+        translated.push('\\');
+    }
+    translated.push(c);
+}
+
+/// Translates the bracket expression whose content starts at `index` (just
+/// after the `[`), and returns the index following its closing `]`.
+fn translate_bracket(
+    chars: &[char],
+    mut index: usize,
+    translated: &mut String,
+) -> Result<usize, String> {
+    const UNMATCHED: &str = "Unmatched [, [^, [:, [., or [=";
+    translated.push('[');
+    if chars.get(index) == Some(&'^') {
+        translated.push('^');
+        index += 1;
+    }
+    let content_start = index;
+    // The last single character member, which can start a range.
+    let mut range_start = None;
+    loop {
+        let Some(&c) = chars.get(index) else {
+            return Err(UNMATCHED.to_string());
+        };
+        // A `]` first in the list is an ordinary member.
+        if c == ']' && index > content_start {
+            translated.push(']');
+            return Ok(index + 1);
+        }
+        if c == '[' && matches!(chars.get(index + 1), Some(':' | '.' | '=')) {
+            let delimiter = chars[index + 1];
+            let end = (index + 2..chars.len().saturating_sub(1))
+                .find(|&end| chars[end] == delimiter && chars[end + 1] == ']')
+                .ok_or_else(|| UNMATCHED.to_string())?;
+            let name: String = chars[index + 2..end].iter().collect();
+            if delimiter == ':' {
+                // A character class such as `[:alpha:]`, kept verbatim.
+                if !CLASS_NAMES.contains(&name.as_str()) {
+                    return Err("Invalid character class name".to_string());
+                }
+                translated.extend(&chars[index..end + 2]);
+                range_start = None;
+            } else {
+                // A collating element or equivalence class such as `[.a.]`
+                // or `[=a=]`, which only stands for a single character here.
+                let mut name_chars = name.chars();
+                let (Some(member), None) = (name_chars.next(), name_chars.next()) else {
+                    return Err("Invalid collation character".to_string());
+                };
+                push_literal(translated, member);
+                range_start = Some(member);
+            }
+            index = end + 2;
+            continue;
+        }
+        let is_range = c == '-'
+            && index > content_start
+            && chars.get(index + 1).is_some_and(|&next| next != ']');
+        if is_range {
+            if let (Some(start), Some(&end)) = (range_start, chars.get(index + 1)) {
+                if end < start {
+                    return Err("Invalid range end".to_string());
+                }
+            }
+            translated.push('-');
+            range_start = None;
+        } else {
+            // Backslashes are ordinary members of a bracket expression, and
+            // the `regex` crate gives meaning to `[`, `&&`, `--` and `~~`.
+            push_literal(translated, c);
+            range_start = Some(c);
+        }
+        index += 1;
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use pretty_assertions::assert_eq;
+
+    fn matches(pattern: &str, text: &str) -> bool {
+        Bre::new(pattern).unwrap().is_match(text.as_bytes())
+    }
+
+    #[test]
+    fn groups_and_alternatives() {
+        assert!(matches(r"^ *\(sub\|fn\) ", "  fn main"));
+        assert!(matches(r"^ *\(sub\|fn\) ", "sub x"));
+        assert!(!matches(r"^ *\(sub\|fn\) ", "  let x"));
+        // Unescaped operators of extended expressions are ordinary characters.
+        assert!(matches(r"f(x)|y+?", "f(x)|y+?"));
+        assert!(!matches(r"f(x)|y+?", "fx"));
+    }
+
+    #[test]
+    fn intervals_and_repetitions() {
+        assert!(matches(r"^ab\{2,3\}c$", "abbc"));
+        assert!(!matches(r"^ab\{2,3\}c$", "abc"));
+        assert!(matches(r"^x\+$", "xxx"));
+        assert!(matches(r"^y\?z$", "z"));
+        // `*` with nothing to repeat is literal.
+        assert!(matches(r"*q", "a*q"));
+        assert!(matches(r"^*q", "*q"));
+        assert!(matches(r"\(*\)", "*"));
+        assert!(matches(r"to*k", "tk"));
+    }
+
+    #[test]
+    fn anchors_only_at_the_ends() {
+        assert!(matches(r"a^b", "a^b"));
+        assert!(matches(r"a$b", "a$b"));
+        assert!(matches(r"end$", "the end"));
+        assert!(!matches(r"end$", "endless"));
+        assert!(matches(r"\(x$\|^y\)", "yes"));
+    }
+
+    #[test]
+    fn bracket_expressions() {
+        assert!(matches(r"^[[:alpha:]$_]", "$var"));
+        assert!(matches(r"^[[:alpha:]$_]", "_x"));
+        assert!(!matches(r"^[[:alpha:]$_]", " x"));
+        assert!(matches(r"[]x]", "]"));
+        assert!(matches(r"[^]x]", "y"));
+        assert!(matches(r"[a\]", "\\"));
+        assert!(matches(r"[a-c]", "b"));
+        assert!(matches(r"[a-]", "-"));
+        assert!(matches(r"[&&~~[]", "["));
+    }
+
+    #[test]
+    fn errors() {
+        for (pattern, message) in [
+            ("ab\\", "Trailing backslash"),
+            ("[ab", "Unmatched [, [^, [:, [., or [="),
+            ("[[:alpha:]", "Unmatched [, [^, [:, [., or [="),
+            (r"\(a\)\1", "Invalid back reference"),
+            (r"x\(y", r"Unmatched ( or \("),
+            (r"x\)y", r"Unmatched ) or \)"),
+            (r"k\{4", r"Unmatched \{"),
+            (r"k\{\}", r"Invalid content of \{\}"),
+            (r"k\{2,z\}", r"Invalid content of \{\}"),
+            (r"k\{5,2\}", r"Invalid content of \{\}"),
+            (r"k\{40000\}", "Regular expression too big"),
+            ("[[:vowel:]]", "Invalid character class name"),
+            ("[[.ab.]]", "Invalid collation character"),
+            ("[m-c]", "Invalid range end"),
+        ] {
+            assert_eq!(
+                Bre::new(pattern).err().as_deref(),
+                Some(message),
+                "{pattern}"
+            );
+        }
+    }
+
+    #[test]
+    fn literal_operators_with_nothing_to_repeat() {
+        assert!(matches(r"\{3\}", "{3}"));
+        assert!(matches(r"\(\+x\)", "+x"));
+        assert!(matches(r"q\{,2\}r", "r"));
+        assert!(matches(r"^q\{2,\}$", "qqqq"));
+        assert!(matches(r"[[.-.][=k=]]", "-"));
+        assert!(matches(r"[[.-.][=k=]]", "k"));
+    }
+}