xbp 10.43.0

//! Project-local ignore rules for XBP discovery, versioning, and worktree-watch.
//!
//! Sources (merged, later sources override earlier for the same pattern text):
//! 1. Built-in skip directories used by service discovery
//! 2. `.xbpignore` at the project root (optional)
//! 3. `.xbp/.xbpignore` (canonical project file)
//! 4. Extra patterns passed by the caller from `.xbp/xbp.yaml`:
//!    - service discovery / version targets → `ignore_paths`
//!      (aliases: `ignored_paths`, `xbpignore`)
//!    - worktree-watch only → `watch_ignore_paths` (alias: `watch_ignore`)
//! 5. Global `~/.xbp/.xbp-worktree-ignore` (worktree-watch only; seeded by XBP)
//!
//! `ignore_paths` never silences worktree-watch; use `watch_ignore_paths` for that.
//!
//! Pattern syntax is a practical gitignore subset:
//! - blank lines and `#` comments are ignored
//! - `!pattern` negates a previous match
//! - trailing `/` matches directories only
//! - leading `/` anchors to the project root
//! - `*` matches within a path segment; `**` matches across segments
//! - unanchored patterns match as a path suffix or any path segment (e.g. `node_modules`)

use std::fs;
use std::path::{Path, PathBuf};

use super::strip_utf8_bom;

/// Filename for machine-wide worktree-watch ignore rules under the global XBP root.
pub const GLOBAL_WORKTREE_IGNORE_FILENAME: &str = ".xbp-worktree-ignore";

/// Directory **names** always skipped for service discovery walks and treated as
/// built-in worktree-watch forbidden folders (any path segment).
///
/// Includes tooling/cache/output trees that must never become services or activity.
pub const DEFAULT_NOISE_DIR_NAMES: &[&str] = &[
    // VCS / editor / agent tooling
    ".git",
    ".github",
    ".hg",
    ".svn",
    ".agents",
    ".cargo",
    ".codex",
    ".cursor",
    ".idea",
    ".postman",
    // Framework / monorepo caches
    ".next",
    ".nx",
    ".open-next",
    ".turbo",
    ".vercel",
    ".xbp",
    ".dist",
    // Dependency / build output
    "node_modules",
    "target",
    "target-publish",
    "dist",
    "build",
    "out",
    "coverage",
    "tmp",
    "vendor",
    "venv",
    ".venv",
    "__pycache__",
    // Local agent / CLI session noise
    "terminals",
    // Agent / MCP workspaces
    "mcps",
    "agent-tools",
];

/// Alias used by service discovery (`WalkDir` skip + project ignore seed).
pub const DEFAULT_DISCOVERY_SKIP_DIRS: &[&str] = DEFAULT_NOISE_DIR_NAMES;

/// Default gitignore-style patterns always merged into `~/.xbp/.xbp-worktree-ignore`.
/// Kept in sync with [`DEFAULT_NOISE_DIR_NAMES`] plus anchored root forms.
pub const DEFAULT_GLOBAL_WORKTREE_IGNORE_PATTERNS: &[&str] = &[
    // Tooling
    ".agents/",
    ".cargo/",
    ".codex/",
    ".cursor/",
    ".git/",
    ".github/",
    ".idea/",
    ".postman/",
    ".xbp/",
    // Framework caches
    ".next/",
    ".nx/",
    ".open-next/",
    ".turbo/",
    ".vercel/",
    ".dist/",
    // Build / deps
    "target/",
    "target-publish/",
    "/target-publish/",
    "dist/",
    "build/",
    "out/",
    "coverage/",
    "node_modules/",
    "tmp/",
    "vendor/",
    "venv/",
    ".venv/",
    "__pycache__/",
    // Local agent / CLI session noise
    "terminals/",
    "/terminals/",
    // Agent workspaces (also anchored)
    "mcps/",
    "/mcps/",
    "agent-tools/",
    "/agent-tools/",
];

#[derive(Debug, Clone, PartialEq, Eq)]
struct IgnorePattern {
    raw: String,
    negated: bool,
    dir_only: bool,
    /// Pattern body after stripping `!` and trailing `/`, normalized to `/`.
    body: String,
    anchored: bool,
}

#[derive(Debug, Clone, Default, PartialEq, Eq)]
pub struct XbpIgnoreSet {
    patterns: Vec<IgnorePattern>,
}

impl XbpIgnoreSet {
    pub fn empty() -> Self {
        Self {
            patterns: Vec::new(),
        }
    }

    pub fn is_empty(&self) -> bool {
        self.patterns.is_empty()
    }

    pub fn patterns(&self) -> impl Iterator<Item = &str> {
        self.patterns.iter().map(|pattern| pattern.raw.as_str())
    }

    pub fn from_patterns<I, S>(patterns: I) -> Self
    where
        I: IntoIterator<Item = S>,
        S: AsRef<str>,
    {
        let mut set = Self::empty();
        set.extend_patterns(patterns);
        set
    }

    pub fn from_file_content(content: &str) -> Self {
        let mut set = Self::empty();
        set.extend_file_content(content);
        set
    }

    pub fn extend_patterns<I, S>(&mut self, patterns: I)
    where
        I: IntoIterator<Item = S>,
        S: AsRef<str>,
    {
        for pattern in patterns {
            if let Some(parsed) = parse_ignore_pattern(pattern.as_ref()) {
                self.patterns.push(parsed);
            }
        }
    }

    pub fn extend_file_content(&mut self, content: &str) {
        let stripped = strip_utf8_bom(content);
        for line in stripped.lines() {
            let trimmed = line.trim();
            if trimmed.is_empty() || trimmed.starts_with('#') {
                continue;
            }
            if let Some(parsed) = parse_ignore_pattern(trimmed) {
                self.patterns.push(parsed);
            }
        }
    }

    pub fn merge(&mut self, other: &XbpIgnoreSet) {
        self.patterns.extend(other.patterns.iter().cloned());
    }

    /// True when the relative path should be ignored.
    ///
    /// `is_dir` should be true when the path is known to be a directory (walk
    /// filtering, service roots). When unknown, pass `false` so file-oriented
    /// patterns still apply and directory-only patterns only match if the path
    /// looks like a directory prefix match via children is not required.
    pub fn is_ignored_relative(&self, relative: &str, is_dir: bool) -> bool {
        let normalized = normalize_relative_path(relative);
        if normalized.is_empty() || normalized == "." {
            return false;
        }

        let mut ignored = false;
        for pattern in &self.patterns {
            if pattern_matches(pattern, &normalized, is_dir) {
                ignored = !pattern.negated;
            }
        }
        ignored
    }

    pub fn is_ignored_path(&self, project_root: &Path, path: &Path) -> bool {
        let relative = match path.strip_prefix(project_root) {
            Ok(rel) => rel.to_string_lossy().replace('\\', "/"),
            Err(_) => path.to_string_lossy().replace('\\', "/"),
        };
        let is_dir = path.is_dir();
        self.is_ignored_relative(&relative, is_dir)
    }

    /// True when a directory entry should not be descended into during discovery.
    pub fn skips_dir_name(&self, dir_name: &str) -> bool {
        let name = dir_name.trim().trim_matches(|ch| ch == '/' || ch == '\\');
        if name.is_empty() || name == "." || name == ".." {
            return false;
        }
        // Bare folder names in patterns (and built-ins expressed as such).
        self.is_ignored_relative(name, true)
    }

    /// True when a service root directory (project-relative) is ignored.
    pub fn ignores_service_root(&self, root_directory: &str) -> bool {
        let normalized = normalize_relative_path(root_directory);
        if normalized.is_empty() || normalized == "." {
            return false;
        }
        self.is_ignored_relative(&normalized, true)
    }

    /// True when a version target path should be excluded from versioning.
    pub fn ignores_version_target(&self, target: &str) -> bool {
        let normalized = normalize_relative_path(target);
        if normalized.is_empty() {
            return false;
        }
        // Version manifests are files; also ignore when a parent directory is banned.
        if self.is_ignored_relative(&normalized, false) {
            return true;
        }
        // Parent path segments: packages/icons/package.json → packages/icons
        if let Some(parent) = Path::new(&normalized).parent() {
            let parent_rel = parent.to_string_lossy().replace('\\', "/");
            if !parent_rel.is_empty() && parent_rel != "." {
                return self.is_ignored_relative(&parent_rel, true);
            }
        }
        false
    }
}

/// Load ignore rules for a project root.
///
/// Order: optional extra patterns (e.g. from yaml) are applied after files so
/// config can refine file rules. Built-in discovery skip dirs are always first.
pub fn load_project_xbp_ignore(project_root: &Path, extra_patterns: &[String]) -> XbpIgnoreSet {
    let mut set = XbpIgnoreSet::from_patterns(
        DEFAULT_DISCOVERY_SKIP_DIRS
            .iter()
            .map(|name| (*name).to_string()),
    );

    for candidate in xbp_ignore_file_candidates(project_root) {
        if let Ok(content) = fs::read_to_string(&candidate) {
            set.extend_file_content(&content);
        }
    }

    if !extra_patterns.is_empty() {
        set.extend_patterns(extra_patterns.iter().map(String::as_str));
    }

    set
}

/// Canonical ignore file paths for a project (existence not required).
pub fn xbp_ignore_file_candidates(project_root: &Path) -> Vec<PathBuf> {
    vec![
        project_root.join(".xbpignore"),
        project_root.join(".xbp").join(".xbpignore"),
    ]
}

pub fn default_xbp_ignore_path(project_root: &Path) -> PathBuf {
    project_root.join(".xbp").join(".xbpignore")
}

/// Initial contents written when the global worktree-watch ignore file is first created.
pub fn default_global_worktree_ignore_content() -> String {
    let mut lines = vec![
        "# Global worktree-watch ignore patterns (gitignore-style syntax).".to_string(),
        "# Managed by XBP at ~/.xbp/.xbp-worktree-ignore (platform equivalent).".to_string(),
        "# Edit to add machine-wide paths excluded from worktree-watch activity.".to_string(),
        "# Built-in noise dirs (target, .next, .xbp, mcps, …) are also hard-coded in the CLI.".to_string(),
        String::new(),
    ];
    for pattern in DEFAULT_GLOBAL_WORKTREE_IGNORE_PATTERNS {
        lines.push((*pattern).to_string());
    }
    lines.push(String::new());
    lines.join("\n")
}

/// Load global worktree-watch ignore rules from the given file (missing file → empty set).
pub fn load_global_worktree_ignore_from_path(path: &Path) -> XbpIgnoreSet {
    match fs::read_to_string(path) {
        Ok(content) => XbpIgnoreSet::from_file_content(&content),
        Err(_) => XbpIgnoreSet::empty(),
    }
}

/// Ensure default MCP ignore patterns exist in the global worktree-watch ignore file.
pub fn sync_global_worktree_ignore_file(path: &Path) -> Result<(), String> {
    if !path.exists() {
        if let Some(parent) = path.parent() {
            fs::create_dir_all(parent).map_err(|error| {
                format!(
                    "Failed to create global worktree-ignore directory {}: {error}",
                    parent.display()
                )
            })?;
        }
        fs::write(path, default_global_worktree_ignore_content()).map_err(|error| {
            format!(
                "Failed to write global worktree-ignore file {}: {error}",
                path.display()
            )
        })?;
        return Ok(());
    }

    let content = fs::read_to_string(path).map_err(|error| {
        format!(
            "Failed to read global worktree-ignore file {}: {error}",
            path.display()
        )
    })?;

    let existing_patterns = content
        .lines()
        .map(str::trim)
        .filter(|line| !line.is_empty() && !line.starts_with('#'))
        .map(|line| line.trim_end_matches('/').to_string())
        .collect::<Vec<_>>();

    let missing = DEFAULT_GLOBAL_WORKTREE_IGNORE_PATTERNS
        .iter()
        .copied()
        .filter(|pattern| {
            let normalized = pattern.trim_end_matches('/');
            !existing_patterns.iter().any(|existing| {
                existing == normalized || existing == *pattern
            })
        })
        .collect::<Vec<_>>();

    if missing.is_empty() {
        return Ok(());
    }

    let mut updated = content;
    if !updated.ends_with('\n') {
        updated.push('\n');
    }
    updated.push_str("\n# Added by XBP (default worktree-watch exclusions)\n");
    for pattern in missing {
        updated.push_str(pattern);
        updated.push('\n');
    }

    fs::write(path, updated).map_err(|error| {
        format!(
            "Failed to update global worktree-ignore file {}: {error}",
            path.display()
        )
    })
}

fn parse_ignore_pattern(raw: &str) -> Option<IgnorePattern> {
    let trimmed = raw.trim();
    if trimmed.is_empty() || trimmed.starts_with('#') {
        return None;
    }

    let (negated, rest) = if let Some(rest) = trimmed.strip_prefix('!') {
        (true, rest.trim())
    } else {
        (false, trimmed)
    };
    if rest.is_empty() {
        return None;
    }

    let dir_only = rest.ends_with('/');
    let without_slash = rest.trim_end_matches(|ch| ch == '/' || ch == '\\');
    let normalized = without_slash.replace('\\', "/");
    let anchored = normalized.starts_with('/');
    let body = normalized
        .trim_start_matches('/')
        .trim_start_matches("./")
        .trim_matches('/')
        .to_string();
    if body.is_empty() {
        return None;
    }

    Some(IgnorePattern {
        raw: trimmed.to_string(),
        negated,
        dir_only,
        body,
        anchored,
    })
}

fn normalize_relative_path(raw: &str) -> String {
    raw.trim()
        .replace('\\', "/")
        .trim_start_matches("./")
        .trim_matches('/')
        .to_string()
}

fn pattern_matches(pattern: &IgnorePattern, path: &str, is_dir: bool) -> bool {
    if pattern.dir_only && !is_dir {
        // Directory-only patterns still match when the path is under that directory.
        // Example: `docs/` ignores `docs/readme.md`.
        let prefix = &pattern.body;
        if path == *prefix || path.starts_with(&format!("{prefix}/")) {
            // file under ignored dir → ignored
            return path != *prefix || is_dir;
        }
        // For exact file path that equals a dir pattern name without being a dir,
        // treat as no match unless path is under it.
        if path == *prefix {
            return false;
        }
        // Unanchored dir pattern: match if any ancestor segment path matches.
        return path_has_directory_match(path, pattern);
    }

    if pattern.anchored {
        return path_glob_match(&pattern.body, path)
            || (!pattern.dir_only && path_is_under_glob(&pattern.body, path));
    }

    // Unanchored: match full path, any suffix path, or a single path component.
    if path_glob_match(&pattern.body, path) {
        return true;
    }
    if path_is_under_glob(&pattern.body, path) {
        return true;
    }

    // Single-segment patterns match any path component (gitignore behavior).
    if !pattern.body.contains('/') && !pattern.body.contains("**") {
        return path
            .split('/')
            .any(|component| path_glob_match(&pattern.body, component));
    }

    // Multi-segment unanchored: match if any suffix matches the pattern.
    let segments: Vec<&str> = path.split('/').filter(|s| !s.is_empty()).collect();
    for start in 0..segments.len() {
        let suffix = segments[start..].join("/");
        if path_glob_match(&pattern.body, &suffix)
            || path_is_under_glob(&pattern.body, &suffix)
        {
            return true;
        }
    }

    false
}

fn path_has_directory_match(path: &str, pattern: &IgnorePattern) -> bool {
    if pattern.anchored {
        return path == pattern.body || path.starts_with(&format!("{}/", pattern.body));
    }
    if !pattern.body.contains('/') && !pattern.body.contains('*') {
        return path
            .split('/')
            .any(|component| component.eq_ignore_ascii_case(&pattern.body));
    }
    let segments: Vec<&str> = path.split('/').filter(|s| !s.is_empty()).collect();
    for start in 0..segments.len() {
        let suffix = segments[start..].join("/");
        if path_glob_match(&pattern.body, &suffix)
            || suffix.starts_with(&format!("{}/", pattern.body))
            || path_is_under_glob(&pattern.body, &suffix)
        {
            return true;
        }
    }
    false
}

fn path_is_under_glob(pattern: &str, path: &str) -> bool {
    // If pattern matches a parent of path, path is ignored.
    let segments: Vec<&str> = path.split('/').filter(|s| !s.is_empty()).collect();
    for end in 1..segments.len() {
        let parent = segments[..end].join("/");
        if path_glob_match(pattern, &parent) {
            return true;
        }
    }
    false
}

/// Glob match with `*` (segment-local) and `**` (cross-segment).
fn path_glob_match(pattern: &str, path: &str) -> bool {
    let pattern = pattern.trim_matches('/');
    let path = path.trim_matches('/');
    if pattern == "**" {
        return true;
    }
    glob_match_segments(
        &split_glob_segments(pattern),
        &path.split('/').filter(|s| !s.is_empty()).collect::<Vec<_>>(),
    )
}

fn split_glob_segments(pattern: &str) -> Vec<&str> {
    pattern.split('/').filter(|s| !s.is_empty()).collect()
}

fn glob_match_segments(pattern: &[&str], path: &[&str]) -> bool {
    fn rec(p: &[&str], s: &[&str]) -> bool {
        match (p.first().copied(), s.first().copied()) {
            (None, None) => true,
            (Some("**"), _) => {
                // `**` matches zero or more segments, then the rest of the pattern.
                if rec(&p[1..], s) {
                    return true;
                }
                // Consume one path segment and keep `**`.
                if s.is_empty() {
                    return false;
                }
                rec(p, &s[1..])
            }
            (Some(pat), Some(seg)) => {
                if segment_glob_match(pat, seg) {
                    rec(&p[1..], &s[1..])
                } else {
                    false
                }
            }
            (Some(_), None) | (None, Some(_)) => false,
        }
    }
    rec(pattern, path)
}

fn segment_glob_match(pattern: &str, segment: &str) -> bool {
    if pattern == "*" {
        return true;
    }
    if !pattern.contains('*') {
        return pattern == segment;
    }

    // Simple `*` wildcards within a single segment.
    let parts: Vec<&str> = pattern.split('*').collect();
    if parts.len() == 1 {
        return pattern == segment;
    }

    let mut rest = segment;
    for (index, part) in parts.iter().enumerate() {
        if part.is_empty() {
            continue;
        }
        if index == 0 {
            if !rest.starts_with(part) {
                return false;
            }
            rest = &rest[part.len()..];
            continue;
        }
        if index == parts.len() - 1 {
            return rest.ends_with(part);
        }
        if let Some(pos) = rest.find(part) {
            rest = &rest[pos + part.len()..];
        } else {
            return false;
        }
    }
    true
}

#[cfg(test)]
mod tests {
    use super::*;
    use std::fs;
    use std::time::{SystemTime, UNIX_EPOCH};

    fn temp_root(name: &str) -> PathBuf {
        let nanos = SystemTime::now()
            .duration_since(UNIX_EPOCH)
            .expect("time")
            .as_nanos();
        let path = std::env::temp_dir().join(format!("xbp-ignore-{name}-{nanos}"));
        let _ = fs::remove_dir_all(&path);
        fs::create_dir_all(path.join(".xbp")).expect("mkdir");
        path
    }

    #[test]
    fn matches_unanchored_folder_anywhere() {
        let set = XbpIgnoreSet::from_patterns(["node_modules"]);
        assert!(set.is_ignored_relative("node_modules", true));
        assert!(set.is_ignored_relative("apps/web/node_modules", true));
        assert!(set.is_ignored_relative("apps/web/node_modules/pkg/index.js", false));
        assert!(!set.is_ignored_relative("apps/web/src/main.ts", false));
    }

    #[test]
    fn matches_anchored_and_directory_only_patterns() {
        let set = XbpIgnoreSet::from_patterns(["/docs/", "packages/icons"]);
        assert!(set.is_ignored_relative("docs", true));
        assert!(set.is_ignored_relative("docs/guide.md", false));
        assert!(!set.is_ignored_relative("apps/docs/guide.md", false));
        assert!(set.is_ignored_relative("packages/icons", true));
        assert!(set.is_ignored_relative("packages/icons/package.json", false));
        assert!(!set.is_ignored_relative("packages/web/package.json", false));
    }

    #[test]
    fn supports_negation() {
        let set = XbpIgnoreSet::from_patterns(["secrets/", "!secrets/public.env"]);
        assert!(set.is_ignored_relative("secrets/private.key", false));
        assert!(!set.is_ignored_relative("secrets/public.env", false));
    }

    #[test]
    fn loads_dot_xbp_ignore_and_yaml_patterns() {
        let root = temp_root("load");
        fs::write(
            root.join(".xbp").join(".xbpignore"),
            "# project ignores\npackages/icons/\n*.snap\n",
        )
        .expect("write ignore");
        fs::write(root.join(".xbpignore"), "legacy-root-ignore/\n").expect("write root ignore");

        let set = load_project_xbp_ignore(
            &root,
            &["e2e/".to_string(), "packages/icons/README.md".to_string()],
        );

        assert!(set.ignores_service_root("packages/icons"));
        assert!(set.ignores_version_target("packages/icons/package.json"));
        assert!(set.is_ignored_relative("legacy-root-ignore/foo", false));
        assert!(set.is_ignored_relative("e2e/test.spec.ts", false));
        assert!(set.is_ignored_relative("apps/web/__snapshots__/a.snap", false));
        // built-in
        assert!(set.skips_dir_name("node_modules"));
        assert!(set.skips_dir_name("target"));

        let _ = fs::remove_dir_all(root);
    }

    #[test]
    fn global_mcps_patterns_ignore_nested_paths() {
        let set = XbpIgnoreSet::from_patterns(DEFAULT_GLOBAL_WORKTREE_IGNORE_PATTERNS);
        assert!(set.is_ignored_relative("mcps", true));
        assert!(set.is_ignored_relative("mcps/github/tools", false));
        assert!(set.is_ignored_relative("apps/api/mcps/local", true));
        assert!(set.is_ignored_relative("mcps/catalog.json", false));
        assert!(!set.is_ignored_relative("apps/web/src/main.ts", false));
    }

    #[test]
    fn sync_global_worktree_ignore_seeds_defaults() {
        let root = temp_root("global-watch-ignore");
        let path = root.join(GLOBAL_WORKTREE_IGNORE_FILENAME);
        sync_global_worktree_ignore_file(&path).expect("seed");
        let content = fs::read_to_string(&path).expect("read");
        assert!(content.contains("mcps/"));
        let set = load_global_worktree_ignore_from_path(&path);
        assert!(set.is_ignored_relative("packages/foo/mcps/bar.json", false));
        let _ = fs::remove_dir_all(root);
    }

    #[test]
    fn glob_double_star_and_star_segment() {
        let set = XbpIgnoreSet::from_patterns(["**/dist", "apps/*/tmp"]);
        assert!(set.is_ignored_relative("packages/foo/dist/index.js", false));
        assert!(set.is_ignored_relative("dist", true));
        assert!(set.is_ignored_relative("apps/web/tmp", true));
        assert!(set.is_ignored_relative("apps/web/tmp/cache", false));
        assert!(!set.is_ignored_relative("apps/web/src/tmp-file.ts", false));
    }
}