doot/crates/doot-core/src/state/store.rs

//! State persistence for doot.

use serde::{Deserialize, Serialize};
use std::collections::HashMap;
use std::path::{Path, PathBuf};
use thiserror::Error;

/// State storage errors.
#[derive(Error, Debug)]
pub enum StateError {
    #[error("io error: {0}")]
    IoError(#[from] std::io::Error),

    #[error("serialization error: {0}")]
    SerializationError(#[from] serde_json::Error),
}

/// Persistent doot state.
#[derive(Debug, Clone, Serialize, Deserialize, Default)]
pub struct State {
    pub version: u32,
    pub deployments: HashMap<String, DeploymentRecord>,
    pub packages: HashMap<String, PackageRecord>,
    pub snapshots: Vec<String>,
}

/// Deploy mode for a file.
#[derive(Debug, Clone, Copy, Serialize, Deserialize, PartialEq, Default)]
pub enum DeployMode {
    #[default]
    Copy,
    Link,
}

/// Record of a deployed file.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct DeploymentRecord {
    pub source: PathBuf,
    pub target: PathBuf,
    pub source_hash: String,
    pub target_hash: String,
    pub deployed_at: String,
    pub mode: DeployMode,
    /// Whether this file was deployed as a template (source != target content).
    #[serde(default)]
    pub template: bool,
}

/// Sync status after comparing current hashes with recorded state.
#[derive(Debug, Clone, Copy, PartialEq)]
pub enum SyncStatus {
    Synced,
    SourceChanged,
    TargetChanged,
    Conflict,
    NotDeployed,
    TargetMissing,
    SourceMissing,
}

/// Record of an installed package.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct PackageRecord {
    pub name: String,
    pub manager: String,
    pub installed_at: String,
}

/// Manages doot state persistence.
pub struct StateStore {
    path: PathBuf,
    state: State,
    dirty: bool,
}

impl StateStore {
    /// Loads or creates a state store at the given path.
    #[tracing::instrument(skip_all, fields(path = %path.display()))]
    pub fn new(path: &Path) -> Self {
        let state = if path.exists() {
            std::fs::read_to_string(path)
                .ok()
                .and_then(|s| serde_json::from_str(&s).ok())
                .unwrap_or_default()
        } else {
            State::default()
        };

        Self {
            path: path.to_path_buf(),
            state,
            dirty: false,
        }
    }

    /// Records a deployment with both source and target hashes.
    #[tracing::instrument(skip(self), fields(source = %source.display(), target = %target.display()))]
    pub fn record_deployment(&mut self, source: &Path, target: &Path, mode: DeployMode) {
        self.record_deployment_with_template(source, target, mode, false);
    }

    /// Records a deployment with template flag.
    #[tracing::instrument(skip(self), fields(source = %source.display(), target = %target.display()))]
    pub fn record_deployment_with_template(
        &mut self,
        source: &Path,
        target: &Path,
        mode: DeployMode,
        template: bool,
    ) {
        let source_hash = hash_path(source);
        let target_hash = hash_path(target);

        let record = DeploymentRecord {
            source: source.to_path_buf(),
            target: target.to_path_buf(),
            source_hash,
            target_hash,
            deployed_at: chrono_now(),
            mode,
            template,
        };

        self.state
            .deployments
            .insert(target.display().to_string(), record);
        self.dirty = true;
    }

    /// Checks sync status by comparing current hashes with recorded state.
    #[tracing::instrument(level = "trace", skip(self))]
    pub fn check_sync_status(&self, source: &Path, target: &Path) -> SyncStatus {
        self.check_sync_status_with_config(source, target, None, None)
    }

    /// Checks sync status, also detecting if template flag changed in config.
    #[tracing::instrument(level = "trace", skip(self))]
    pub fn check_sync_status_with_template(
        &self,
        source: &Path,
        target: &Path,
        current_template: Option<bool>,
    ) -> SyncStatus {
        self.check_sync_status_with_config(source, target, current_template, None)
    }

    /// Checks sync status, also detecting if config flags changed.
    #[tracing::instrument(level = "trace", skip(self))]
    pub fn check_sync_status_with_config(
        &self,
        source: &Path,
        target: &Path,
        current_template: Option<bool>,
        current_mode: Option<DeployMode>,
    ) -> SyncStatus {
        let Some(record) = self.get_deployment(target) else {
            return SyncStatus::NotDeployed;
        };

        // If template flag changed in config, force re-deploy
        if let Some(is_template) = current_template
            && is_template != record.template
        {
            return SyncStatus::SourceChanged;
        }

        // If deploy mode changed in config, force re-deploy
        if let Some(mode) = current_mode
            && mode != record.mode
        {
            return SyncStatus::SourceChanged;
        }

        if !source.exists() {
            return SyncStatus::SourceMissing;
        }
        if !target.exists() {
            return SyncStatus::TargetMissing;
        }

        // If stored hashes are empty (legacy record), treat as needing re-sync
        if record.source_hash.is_empty() || record.target_hash.is_empty() {
            // For templates, we can't compare source to target directly
            if record.template {
                return SyncStatus::SourceChanged;
            }

            let current_source_hash = hash_path(source);
            let current_target_hash = hash_path(target);

            // If source and target currently match, consider it synced
            if current_source_hash == current_target_hash {
                return SyncStatus::Synced;
            }
            // Otherwise, treat as source changed (needs re-deploy)
            return SyncStatus::SourceChanged;
        }

        let current_source_hash = hash_path(source);
        let current_target_hash = hash_path(target);
        tracing::trace!(source_hash = %current_source_hash, target_hash = %current_target_hash, "computed hashes");

        let source_changed = current_source_hash != record.source_hash;
        let target_changed = current_target_hash != record.target_hash;

        match (source_changed, target_changed) {
            (false, false) => SyncStatus::Synced,
            (true, false) => SyncStatus::SourceChanged,
            (false, true) => SyncStatus::TargetChanged,
            (true, true) => SyncStatus::Conflict,
        }
    }

    /// Records a package installation.
    #[tracing::instrument(skip(self))]
    pub fn record_package(&mut self, name: &str, manager: &str) {
        let record = PackageRecord {
            name: name.to_string(),
            manager: manager.to_string(),
            installed_at: chrono_now(),
        };

        self.state.packages.insert(name.to_string(), record);
        self.dirty = true;
    }

    /// Gets a deployment record by target path.
    pub fn get_deployment(&self, target: &Path) -> Option<&DeploymentRecord> {
        self.state.deployments.get(&target.display().to_string())
    }

    /// Returns all deployment records.
    pub fn get_all_deployments(&self) -> &HashMap<String, DeploymentRecord> {
        &self.state.deployments
    }

    /// Removes a deployment record.
    pub fn remove_deployment(&mut self, target: &Path) {
        self.state.deployments.remove(&target.display().to_string());
        self.dirty = true;
    }

    /// Records a snapshot name.
    pub fn add_snapshot(&mut self, name: &str) {
        self.state.snapshots.push(name.to_string());
        self.dirty = true;
    }

    /// Returns all snapshot names.
    pub fn get_snapshots(&self) -> &[String] {
        &self.state.snapshots
    }

    /// Saves state to disk if dirty.
    #[tracing::instrument(skip(self))]
    pub fn save(&mut self) -> Result<(), StateError> {
        if !self.dirty {
            return Ok(());
        }

        if let Some(parent) = self.path.parent() {
            std::fs::create_dir_all(parent)?;
        }

        let json = serde_json::to_string_pretty(&self.state)?;
        std::fs::write(&self.path, json)?;
        self.dirty = false;

        Ok(())
    }

    /// Checks if a target is deployed.
    pub fn is_deployed(&self, target: &Path) -> bool {
        self.state
            .deployments
            .contains_key(&target.display().to_string())
    }

    /// Checks if source has changed since deployment.
    pub fn has_changed(&self, source: &Path, target: &Path) -> bool {
        matches!(
            self.check_sync_status(source, target),
            SyncStatus::SourceChanged | SyncStatus::Conflict | SyncStatus::NotDeployed
        )
    }

    /// Records a directory deployment by tracking each file individually.
    #[tracing::instrument(skip(self))]
    pub fn record_directory_deployment(
        &mut self,
        source_dir: &Path,
        target_dir: &Path,
        mode: DeployMode,
    ) {
        let mut files = Vec::new();
        collect_files(source_dir, &mut files);

        for source_file in files {
            if let Ok(relative) = source_file.strip_prefix(source_dir) {
                let target_file = target_dir.join(relative);
                self.record_deployment(&source_file, &target_file, mode);
            }
        }
    }

    /// Returns files that have changed in a directory.
    /// Returns (source_path, target_path, status) for each changed file.
    #[tracing::instrument(skip(self), fields(source_dir = %source_dir.display(), target_dir = %target_dir.display()))]
    pub fn get_changed_files_in_dir(
        &self,
        source_dir: &Path,
        target_dir: &Path,
    ) -> Vec<(PathBuf, PathBuf, SyncStatus)> {
        let mut changed = Vec::new();

        // Check files in source directory
        let mut source_files = Vec::new();
        collect_files(source_dir, &mut source_files);

        for source_file in source_files {
            if let Ok(relative) = source_file.strip_prefix(source_dir) {
                let target_file = target_dir.join(relative);
                let status = self.check_sync_status(&source_file, &target_file);

                if status != SyncStatus::Synced {
                    changed.push((source_file, target_file, status));
                }
            }
        }

        // Check for files that exist in target but not in source (deleted from source)
        let mut target_files = Vec::new();
        if target_dir.exists() {
            collect_files(target_dir, &mut target_files);
        }

        for target_file in target_files {
            if let Ok(relative) = target_file.strip_prefix(target_dir) {
                let source_file = source_dir.join(relative);
                if !source_file.exists() {
                    // Only mark as SourceMissing if we previously tracked this file
                    // Files that were never in source (e.g., fish_variables) should be ignored
                    if self.get_deployment(&target_file).is_some() {
                        changed.push((source_file, target_file, SyncStatus::SourceMissing));
                    }
                    // Otherwise, ignore - it's an untracked file created in target
                }
            }
        }

        changed
    }

    /// Removes all deployment records for files within a directory.
    #[tracing::instrument(skip(self))]
    pub fn remove_directory_deployment(&mut self, target_dir: &Path) {
        let target_prefix = target_dir.display().to_string();
        let to_remove: Vec<String> = self
            .state
            .deployments
            .keys()
            .filter(|k| k.starts_with(&target_prefix))
            .cloned()
            .collect();

        for key in to_remove {
            self.state.deployments.remove(&key);
        }
        self.dirty = true;
    }
}

fn hash_path(path: &Path) -> String {
    if !path.exists() {
        return String::new();
    }

    if path.is_file() {
        std::fs::read(path)
            .map(|content| blake3::hash(&content).to_hex().to_string())
            .unwrap_or_default()
    } else if path.is_dir() {
        hash_directory(path)
    } else {
        String::new()
    }
}

fn hash_directory(dir: &Path) -> String {
    let mut hasher = blake3::Hasher::new();
    let mut entries = Vec::new();

    // Collect all file paths recursively
    collect_files(dir, &mut entries);

    // Sort for deterministic hashing
    entries.sort();

    for file_path in entries {
        // Include the relative path in the hash to detect renames
        if let Ok(relative) = file_path.strip_prefix(dir) {
            hasher.update(relative.to_string_lossy().as_bytes());
        }

        // Hash the file content
        if let Ok(content) = std::fs::read(&file_path) {
            hasher.update(&content);
        }
    }

    hasher.finalize().to_hex().to_string()
}

fn collect_files(dir: &Path, files: &mut Vec<PathBuf>) {
    if let Ok(entries) = std::fs::read_dir(dir) {
        for entry in entries.flatten() {
            let path = entry.path();
            if path.is_dir() {
                collect_files(&path, files);
            } else if path.is_file() {
                files.push(path);
            }
        }
    }
}

fn chrono_now() -> String {
    use std::time::{SystemTime, UNIX_EPOCH};
    let secs = SystemTime::now()
        .duration_since(UNIX_EPOCH)
        .unwrap()
        .as_secs();
    format!("{}", secs)
}