whi 0.6.4

use std::path::PathBuf;

pub struct PathSearcher {
    dirs: Vec<PathBuf>,
    canon_dirs: std::cell::RefCell<Vec<Option<PathBuf>>>,
}

/// Validate a `PATH` entry for suspicious or malicious content
fn validate_path_entry(path: &str) -> Result<(), String> {
    // Check for null bytes
    if path.contains('\0') {
        return Err("PATH entry contains null byte".to_string());
    }

    // Check for control characters (except tab which is valid)
    for ch in path.chars() {
        if ch.is_control() && ch != '\t' {
            return Err(format!("PATH entry contains control character: {ch:?}"));
        }
    }

    Ok(())
}

/// Warn about potentially dangerous `PATH` entries
fn warn_suspicious_path(path: &str) {
    // Warn about shell metacharacters that could be dangerous
    const DANGEROUS_CHARS: &[char] = &['$', '`', ';', '&', '|', '<', '>', '(', ')', '{', '}'];

    for &ch in DANGEROUS_CHARS {
        if path.contains(ch) {
            eprintln!("Warning: PATH entry contains shell metacharacter '{ch}': {path}");
            return;
        }
    }

    // Warn about relative paths (but don't reject)
    if !path.starts_with('/') && !path.is_empty() && path != "." {
        eprintln!("Warning: Relative PATH entry detected: {path}");
    }
}

impl PathSearcher {
    #[must_use]
    pub fn new(path_var: &str) -> Self {
        let mut has_empty = false;

        let dirs: Vec<PathBuf> = path_var
            .split(':')
            .filter_map(|s| {
                // Check for empty components
                if s.is_empty() {
                    has_empty = true;
                    return None; // Skip empty components instead of treating as "."
                }

                // Validate entry
                if let Err(e) = validate_path_entry(s) {
                    eprintln!("Warning: Skipping invalid PATH entry: {e}");
                    return None;
                }

                // Warn about suspicious entries
                warn_suspicious_path(s);

                Some(PathBuf::from(s))
            })
            .collect();

        if has_empty {
            eprintln!("Warning: Empty PATH component(s) detected and skipped. Empty components can be a security risk.");
        }

        let canon_dirs = std::cell::RefCell::new(vec![None; dirs.len()]);
        PathSearcher { dirs, canon_dirs }
    }

    #[must_use]
    pub fn dirs(&self) -> &[PathBuf] {
        &self.dirs
    }

    /// Get the canonical path for a directory at the given index (0-based)
    fn canonicalize_index(&self, idx: usize) -> Option<PathBuf> {
        let mut cache = self.canon_dirs.borrow_mut();
        if cache[idx].is_none() {
            cache[idx] = std::fs::canonicalize(&self.dirs[idx]).ok();
        }
        cache[idx].clone()
    }

    /// Check if a path already exists in `PATH`
    #[must_use]
    pub fn contains(&self, path: &std::path::Path) -> bool {
        self.find_path_index(path).is_some()
    }

    /// Insert a path at the given position (1-based), mutating this `PathSearcher`
    /// Returns Err if position is invalid
    pub fn insert_at(&mut self, path: &std::path::Path, position: usize) -> Result<(), String> {
        if position == 0 {
            return Err("Position must be >= 1".to_string());
        }

        let path_buf = path.to_path_buf();

        // Convert to 0-based index, but cap at the end of the list
        let insert_idx = (position - 1).min(self.dirs.len());

        self.dirs.insert(insert_idx, path_buf);

        // Update cache size
        self.canon_dirs.borrow_mut().insert(insert_idx, None);

        Ok(())
    }

    pub fn move_entry(&self, from: usize, to: usize) -> Result<String, String> {
        let len = self.dirs.len();

        // Validate indices (1-based)
        if from == 0 || to == 0 {
            return Err(format!(
                "Invalid index: indices must be >= 1 (got from={from}, to={to})"
            ));
        }
        if from > len {
            return Err(format!(
                "Index {from} out of bounds (PATH has {len} entries)"
            ));
        }
        if to > len {
            return Err(format!("Index {to} out of bounds (PATH has {len} entries)"));
        }

        // Convert to 0-based
        let from_idx = from - 1;
        let to_idx = to - 1;

        // Create new ordering
        let mut new_dirs = self.dirs.clone();
        let item = new_dirs.remove(from_idx);
        new_dirs.insert(to_idx, item);

        // Return new PATH string
        Ok(new_dirs
            .iter()
            .map(|d| d.display().to_string())
            .collect::<Vec<_>>()
            .join(":"))
    }

    pub fn swap_entries(&self, idx1: usize, idx2: usize) -> Result<String, String> {
        let len = self.dirs.len();

        // Validate indices (1-based)
        if idx1 == 0 || idx2 == 0 {
            return Err(format!(
                "Invalid index: indices must be >= 1 (got idx1={idx1}, idx2={idx2})"
            ));
        }
        if idx1 > len {
            return Err(format!(
                "Index {idx1} out of bounds (PATH has {len} entries)"
            ));
        }
        if idx2 > len {
            return Err(format!(
                "Index {idx2} out of bounds (PATH has {len} entries)"
            ));
        }

        // Convert to 0-based
        let idx1_0 = idx1 - 1;
        let idx2_0 = idx2 - 1;

        // Create new ordering with swapped entries
        let mut new_dirs = self.dirs.clone();
        new_dirs.swap(idx1_0, idx2_0);

        // Return new PATH string
        Ok(new_dirs
            .iter()
            .map(|d| d.display().to_string())
            .collect::<Vec<_>>()
            .join(":"))
    }

    #[must_use]
    pub fn clean_duplicates(&self) -> (String, Vec<usize>) {
        let mut seen = std::collections::HashSet::new();
        let mut cleaned = Vec::new();
        let mut removed_indices = Vec::new();

        for (idx, dir) in self.dirs.iter().enumerate() {
            let dir_str = dir.display().to_string();
            if seen.insert(dir_str.clone()) {
                cleaned.push(dir_str);
            } else {
                // Duplicate found - track 1-based index
                removed_indices.push(idx + 1);
            }
        }

        (cleaned.join(":"), removed_indices)
    }

    pub fn delete_entry(&self, idx: usize) -> Result<String, String> {
        let len = self.dirs.len();

        // Validate index (1-based)
        if idx == 0 {
            return Err(format!("Invalid index: {idx} (must be >= 1)"));
        }
        if idx > len {
            return Err(format!(
                "Index {idx} out of bounds (PATH has {len} entries)"
            ));
        }

        // Convert to 0-based
        let idx_0 = idx - 1;

        // Create new PATH without this entry
        let mut new_dirs = self.dirs.clone();
        new_dirs.remove(idx_0);

        // Return new PATH string
        Ok(new_dirs
            .iter()
            .map(|d| d.display().to_string())
            .collect::<Vec<_>>()
            .join(":"))
    }

    pub fn delete_entries(&self, indices: &[usize]) -> Result<String, String> {
        let len = self.dirs.len();

        // Validate all indices (1-based)
        for &idx in indices {
            if idx == 0 {
                return Err(format!("Invalid index: {idx} (indices must be >= 1)"));
            }
            if idx > len {
                return Err(format!(
                    "Index {idx} out of bounds (PATH has {len} entries)"
                ));
            }
        }

        // Sort indices in descending order to delete from highest to lowest
        // This avoids index shifting issues
        let mut sorted_indices: Vec<usize> = indices.to_vec();
        sorted_indices.sort_unstable_by(|a, b| b.cmp(a));

        // Remove duplicates
        sorted_indices.dedup();

        // Create new PATH without these entries
        let mut new_dirs = self.dirs.clone();
        for &idx in &sorted_indices {
            let idx_0 = idx - 1; // Convert to 0-based
            new_dirs.remove(idx_0);
        }

        // Return new PATH string
        Ok(new_dirs
            .iter()
            .map(|d| d.display().to_string())
            .collect::<Vec<_>>()
            .join(":"))
    }

    /// Add a new directory to `PATH` if not already present at the beginning
    /// Returns the new `PATH` string and the index where it was added (1-based)
    pub fn add_path(&self, path: &std::path::Path) -> Result<(String, usize), String> {
        match self.add_path_at_position(path, 1) {
            Ok(new_path) => Ok((new_path, 1)),
            Err(e) => Err(e),
        }
    }

    /// Add a new directory to `PATH` at a specific position if not already present
    /// Returns the new `PATH` string (1-based position)
    pub fn add_path_at_position(
        &self,
        path: &std::path::Path,
        position: usize,
    ) -> Result<String, String> {
        let path_buf = path.to_path_buf();

        // Check if already exists
        if let Some(_idx) = self.find_path_index(&path_buf) {
            // Already exists - just return current PATH
            return Ok(self.to_path_string());
        }

        // Validate position (1-based)
        if position == 0 {
            return Err("Position must be >= 1".to_string());
        }

        let mut new_dirs = self.dirs.clone();

        // Convert to 0-based index, but cap at the end of the list
        let insert_idx = (position - 1).min(new_dirs.len());

        new_dirs.insert(insert_idx, path_buf);

        let new_path = new_dirs
            .iter()
            .map(|d| d.display().to_string())
            .collect::<Vec<_>>()
            .join(":");

        Ok(new_path)
    }

    /// Find the index of an exact path match (1-based)
    #[must_use]
    pub fn find_path_index(&self, path: &std::path::Path) -> Option<usize> {
        use std::fs;

        // Try to canonicalize the search path if it exists
        let canonical_search = fs::canonicalize(path).unwrap_or_else(|_| path.to_path_buf());

        for (idx, dir) in self.dirs.iter().enumerate() {
            // Compare both as-is and canonicalized
            if dir == path || dir == &canonical_search {
                return Some(idx + 1); // Return 1-based index
            }

            // Use cached canonicalized dir
            if let Some(canonical_dir) = self.canonicalize_index(idx) {
                if canonical_dir == canonical_search {
                    return Some(idx + 1);
                }
            }
        }

        None
    }

    /// Find all indices matching a fuzzy pattern
    #[must_use]
    pub fn find_fuzzy_indices(
        &self,
        pattern: &str,
        executable_name: Option<&str>,
    ) -> Vec<(usize, &PathBuf)> {
        use crate::path_resolver::FuzzyMatcher;

        let matcher = FuzzyMatcher::new(pattern);
        let mut fuzzy_results = Vec::new();

        for (idx, dir) in self.dirs.iter().enumerate() {
            if matcher.matches(dir) {
                // If executable specified, check it exists
                if let Some(name) = executable_name {
                    let exec_path = dir.join(name);
                    if !exec_path.exists() {
                        continue;
                    }
                }

                fuzzy_results.push((idx + 1, dir)); // 1-based index
            }
        }

        // Sort by match quality (shorter paths first)
        fuzzy_results.sort_by_key(|(_, path)| path.as_os_str().len());

        fuzzy_results
    }

    /// Delete a `PATH` entry by exact path match
    #[allow(dead_code)]
    pub fn delete_by_path(&self, path: &std::path::Path) -> Result<String, String> {
        if let Some(idx) = self.find_path_index(path) {
            self.delete_entry(idx)
        } else {
            Err(format!("Path not found in PATH: {}", path.display()))
        }
    }

    /// Check if an executable exists in a directory
    #[must_use]
    pub fn has_executable(&self, dir: &std::path::Path, name: &str) -> bool {
        use crate::executor::ExecutableCheck;

        let exec_path = dir.join(name);
        exec_path.exists() && ExecutableCheck::new(&exec_path).is_executable()
    }

    /// Convert current dirs to `PATH` string
    #[must_use]
    pub fn to_path_string(&self) -> String {
        self.dirs
            .iter()
            .map(|d| d.display().to_string())
            .collect::<Vec<_>>()
            .join(":")
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_move_entry_forward() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.move_entry(5, 2).unwrap();
        assert_eq!(result, "/a:/e:/b:/c:/d");
    }

    #[test]
    fn test_move_entry_backward() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.move_entry(2, 4).unwrap();
        assert_eq!(result, "/a:/c:/d:/b:/e");
    }

    #[test]
    fn test_move_entry_to_first() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.move_entry(4, 1).unwrap();
        assert_eq!(result, "/d:/a:/b:/c:/e");
    }

    #[test]
    fn test_move_entry_to_last() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.move_entry(2, 5).unwrap();
        assert_eq!(result, "/a:/c:/d:/e:/b");
    }

    #[test]
    fn test_move_entry_same_position() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.move_entry(3, 3).unwrap();
        assert_eq!(result, "/a:/b:/c:/d:/e");
    }

    #[test]
    fn test_move_entry_zero_index() {
        let searcher = PathSearcher::new("/a:/b:/c");
        let result = searcher.move_entry(0, 2);
        assert!(result.is_err());
        let err = result.unwrap_err();
        assert!(err.contains("must be >= 1"));
        assert!(err.contains("0"));
    }

    #[test]
    fn test_move_entry_out_of_bounds() {
        let searcher = PathSearcher::new("/a:/b:/c");
        let result = searcher.move_entry(1, 5);
        assert!(result.is_err());
        assert!(result.unwrap_err().contains("out of bounds"));
    }

    #[test]
    fn test_swap_entries_basic() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.swap_entries(2, 4).unwrap();
        assert_eq!(result, "/a:/d:/c:/b:/e");
    }

    #[test]
    fn test_swap_entries_same_index() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.swap_entries(3, 3).unwrap();
        assert_eq!(result, "/a:/b:/c:/d:/e");
    }

    #[test]
    fn test_swap_entries_first_and_last() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.swap_entries(1, 5).unwrap();
        assert_eq!(result, "/e:/b:/c:/d:/a");
    }

    #[test]
    fn test_swap_entries_zero_index() {
        let searcher = PathSearcher::new("/a:/b:/c");
        let result = searcher.swap_entries(0, 2);
        assert!(result.is_err());
        let err = result.unwrap_err();
        assert!(err.contains("must be >= 1"));
        assert!(err.contains("0"));
    }

    #[test]
    fn test_swap_entries_out_of_bounds() {
        let searcher = PathSearcher::new("/a:/b:/c");
        let result = searcher.swap_entries(2, 5);
        assert!(result.is_err());
        assert!(result.unwrap_err().contains("out of bounds"));
    }

    #[test]
    fn test_clean_no_duplicates() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let (result, removed) = searcher.clean_duplicates();
        assert_eq!(result, "/a:/b:/c:/d:/e");
        assert!(removed.is_empty());
    }

    #[test]
    fn test_clean_with_duplicates() {
        let searcher = PathSearcher::new("/a:/b:/c:/b:/d:/a");
        let (result, removed) = searcher.clean_duplicates();
        assert_eq!(result, "/a:/b:/c:/d");
        assert_eq!(removed, vec![4, 6]); // /b at idx 4, /a at idx 6
    }

    #[test]
    fn test_clean_all_same() {
        let searcher = PathSearcher::new("/a:/a:/a");
        let (result, removed) = searcher.clean_duplicates();
        assert_eq!(result, "/a");
        assert_eq!(removed, vec![2, 3]);
    }

    #[test]
    fn test_clean_consecutive_duplicates() {
        let searcher = PathSearcher::new("/a:/a:/b:/b:/c");
        let (result, removed) = searcher.clean_duplicates();
        assert_eq!(result, "/a:/b:/c");
        assert_eq!(removed, vec![2, 4]);
    }

    #[test]
    fn test_clean_empty() {
        let searcher = PathSearcher::new("");
        let (result, removed) = searcher.clean_duplicates();
        assert_eq!(result, "");
        assert!(removed.is_empty());
    }

    #[test]
    fn test_clean_matches_delete() {
        // Verify that clean and delete produce identical results
        let path = "/a:/b:/c:/b:/d:/a:/e:/c";
        let searcher = PathSearcher::new(path);

        // Get clean result and removed indices
        let (clean_result, removed) = searcher.clean_duplicates();

        // Apply delete with the same indices
        let delete_result = searcher.delete_entries(&removed).unwrap();

        // Results must be identical
        assert_eq!(clean_result, delete_result);
        assert_eq!(removed, vec![4, 6, 8]); // /b at 4, /a at 6, /c at 8
    }

    #[test]
    fn test_delete_first() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.delete_entry(1).unwrap();
        assert_eq!(result, "/b:/c:/d:/e");
    }

    #[test]
    fn test_delete_middle() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.delete_entry(3).unwrap();
        assert_eq!(result, "/a:/b:/d:/e");
    }

    #[test]
    fn test_delete_last() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.delete_entry(5).unwrap();
        assert_eq!(result, "/a:/b:/c:/d");
    }

    #[test]
    fn test_delete_only_entry() {
        let searcher = PathSearcher::new("/a");
        let result = searcher.delete_entry(1).unwrap();
        assert_eq!(result, "");
    }

    #[test]
    fn test_delete_zero_index() {
        let searcher = PathSearcher::new("/a:/b:/c");
        let result = searcher.delete_entry(0);
        assert!(result.is_err());
        let err = result.unwrap_err();
        assert!(err.contains("must be >= 1"));
        assert!(err.contains("0"));
    }

    #[test]
    fn test_delete_out_of_bounds() {
        let searcher = PathSearcher::new("/a:/b:/c");
        let result = searcher.delete_entry(5);
        assert!(result.is_err());
        assert!(result.unwrap_err().contains("out of bounds"));
    }

    #[test]
    fn test_delete_entries_multiple() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.delete_entries(&[2, 4]).unwrap();
        assert_eq!(result, "/a:/c:/e");
    }

    #[test]
    fn test_delete_entries_unordered() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.delete_entries(&[5, 2, 3]).unwrap();
        assert_eq!(result, "/a:/d");
    }

    #[test]
    fn test_delete_entries_with_duplicates() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.delete_entries(&[2, 2, 4, 4]).unwrap();
        assert_eq!(result, "/a:/c:/e");
    }

    #[test]
    fn test_delete_entries_all() {
        let searcher = PathSearcher::new("/a:/b:/c");
        let result = searcher.delete_entries(&[1, 2, 3]).unwrap();
        assert_eq!(result, "");
    }

    #[test]
    fn test_delete_entries_zero_index() {
        let searcher = PathSearcher::new("/a:/b:/c");
        let result = searcher.delete_entries(&[1, 0, 3]);
        assert!(result.is_err());
        let err = result.unwrap_err();
        assert!(err.contains("must be >= 1"));
        assert!(err.contains("0"));
    }

    #[test]
    fn test_delete_entries_out_of_bounds() {
        let searcher = PathSearcher::new("/a:/b:/c");
        let result = searcher.delete_entries(&[1, 5, 2]);
        assert!(result.is_err());
        assert!(result.unwrap_err().contains("out of bounds"));
    }

    #[test]
    fn test_delete_entries_single() {
        let searcher = PathSearcher::new("/a:/b:/c:/d:/e");
        let result = searcher.delete_entries(&[3]).unwrap();
        assert_eq!(result, "/a:/b:/d:/e");
    }

    // Security tests

    #[test]
    fn test_path_validation_null_byte() {
        let result = validate_path_entry("hello\0world");
        assert!(result.is_err());
        assert!(result.unwrap_err().contains("null byte"));
    }

    #[test]
    fn test_path_validation_control_chars() {
        let result = validate_path_entry("hello\x01world");
        assert!(result.is_err());
        assert!(result.unwrap_err().contains("control character"));
    }

    #[test]
    fn test_path_validation_tab_allowed() {
        // Tab is a valid character in paths
        let result = validate_path_entry("hello\tworld");
        assert!(result.is_ok());
    }

    #[test]
    fn test_path_validation_newline_rejected() {
        let result = validate_path_entry("hello\nworld");
        assert!(result.is_err());
    }

    #[test]
    fn test_empty_path_components_skipped() {
        // Empty components should be skipped, not treated as "."
        let searcher = PathSearcher::new("/a::/b");
        let dirs = searcher.dirs();
        assert_eq!(dirs.len(), 2);
        assert_eq!(dirs[0].to_str().unwrap(), "/a");
        assert_eq!(dirs[1].to_str().unwrap(), "/b");
    }

    #[test]
    fn test_malicious_path_filtered() {
        // Path with null byte should be filtered out
        let searcher = PathSearcher::new("/good:/bad\0path:/alsogood");
        let dirs = searcher.dirs();
        // Only /good and /alsogood should remain
        assert_eq!(dirs.len(), 2);
        assert_eq!(dirs[0].to_str().unwrap(), "/good");
        assert_eq!(dirs[1].to_str().unwrap(), "/alsogood");
    }

    #[test]
    fn test_error_messages_include_values() {
        let searcher = PathSearcher::new("/a:/b:/c");

        // Test zero index error includes the value
        let err = searcher.move_entry(0, 2).unwrap_err();
        assert!(err.contains("0"));
        assert!(err.contains("must be >= 1"));

        // Test out of bounds error includes the value
        let err = searcher.move_entry(5, 2).unwrap_err();
        assert!(err.contains("5"));
        assert!(err.contains("out of bounds"));
        assert!(err.contains("3 entries"));
    }
}