prvng_platform/crates/orchestrator/src/config.rs

//! Configuration loading and management for the orchestrator
//!
//! This module handles loading configuration from Nickel files with support
//! for:
//! - Nickel-based configuration (schema-driven)
//! - Environment-specific overrides
//! - CLI argument overrides

use std::collections::HashMap;
use std::path::{Path, PathBuf};

use anyhow::{Context, Result};
use platform_config::ConfigLoader;
use serde::{Deserialize, Serialize};

/// Complete orchestrator configuration
#[derive(Debug, Clone, Default, Serialize, Deserialize)]
pub struct OrchestratorConfig {
    pub orchestrator: OrchestratorSettings,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct OrchestratorSettings {
    pub workspace: WorkspaceConfig,
    pub server: ServerConfig,
    pub storage: StorageConfig,
    pub queue: QueueConfig,
    pub batch: BatchConfig,
    #[serde(default)]
    pub extensions: ExtensionsConfig,
    #[serde(default)]
    pub monitoring: Option<MonitoringConfig>,
    #[serde(default)]
    pub logging: Option<LoggingConfig>,
    #[serde(default)]
    pub security: Option<SecurityConfig>,
    #[serde(default)]
    pub performance: Option<PerformanceConfig>,
    #[serde(default)]
    pub build: Option<DockerBuildConfig>,
}

/// Workspace configuration (from workspace.ncl schema)
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct WorkspaceConfig {
    pub name: String,
    pub path: String,
    #[serde(default = "default_true")]
    pub enabled: bool,
    #[serde(default)]
    pub multi_workspace: bool,
}

/// Server configuration
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ServerConfig {
    pub host: String,
    pub port: u16,
    #[serde(default)]
    pub workers: usize,
    #[serde(default)]
    pub keep_alive: u64,
    #[serde(default)]
    pub max_connections: usize,
}

impl Default for ServerConfig {
    fn default() -> Self {
        Self {
            host: "127.0.0.1".to_string(),
            port: 9090,
            workers: 4,
            keep_alive: 75,
            max_connections: 1000,
        }
    }
}

/// Storage configuration (from storage.ncl schema)
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct StorageConfig {
    pub backend: String,
    pub path: Option<String>,
    #[serde(default)]
    pub cache: Option<CacheStorageConfig>,
    #[serde(default)]
    pub compression: Option<CompressionConfig>,
    #[serde(default)]
    pub backup: Option<BackupConfig>,
    #[serde(default)]
    pub replication: Option<ReplicationConfig>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct CacheStorageConfig {
    #[serde(default = "default_true")]
    pub enabled: bool,
    pub cache_type: Option<String>,
    pub size: Option<u64>,
    pub eviction_policy: Option<String>,
    pub ttl: Option<u64>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct CompressionConfig {
    #[serde(default)]
    pub enabled: bool,
    pub algorithm: Option<String>,
    pub level: Option<u32>,
    pub min_size: Option<u64>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct BackupConfig {
    #[serde(default)]
    pub enabled: bool,
    pub interval: Option<u64>,
    pub path: Option<String>,
    pub max_backups: Option<usize>,
    #[serde(default)]
    pub incremental: bool,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ReplicationConfig {
    #[serde(default)]
    pub enabled: bool,
    pub replicas: Option<Vec<String>>,
    pub timeout: Option<u64>,
}

/// Queue configuration
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct QueueConfig {
    pub max_concurrent_tasks: usize,
    #[serde(default = "default_retry_attempts")]
    pub retry_attempts: u32,
    #[serde(default = "default_retry_delay")]
    pub retry_delay: u64, // milliseconds
    #[serde(default = "default_task_timeout")]
    pub task_timeout: u64, // milliseconds
    #[serde(default = "default_persist")]
    pub persist: bool,
    #[serde(default)]
    pub dead_letter_queue: Option<DeadLetterQueueConfig>,
    #[serde(default)]
    pub priority_queue: bool,
    #[serde(default)]
    pub metrics: bool,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct DeadLetterQueueConfig {
    pub enabled: bool,
    pub max_size: Option<usize>,
}

impl Default for QueueConfig {
    fn default() -> Self {
        Self {
            max_concurrent_tasks: 10,
            retry_attempts: 3,
            retry_delay: 5000,
            task_timeout: 3600000,
            persist: true,
            dead_letter_queue: None,
            priority_queue: false,
            metrics: false,
        }
    }
}

/// Batch workflow configuration
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct BatchConfig {
    #[serde(default = "default_parallel_limit")]
    pub parallel_limit: usize,
    #[serde(default = "default_operation_timeout")]
    pub operation_timeout_minutes: u64,
    #[serde(default)]
    pub checkpointing: Option<CheckpointingConfig>,
    #[serde(default)]
    pub rollback: Option<RollbackConfig>,
    #[serde(default)]
    pub metrics: bool,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct CheckpointingConfig {
    pub enabled: bool,
    pub interval: Option<u64>,
    pub max_checkpoints: Option<usize>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct RollbackConfig {
    pub enabled: bool,
    pub strategy: Option<String>,
    pub max_rollback_depth: Option<usize>,
}

impl Default for BatchConfig {
    fn default() -> Self {
        Self {
            parallel_limit: 5,
            operation_timeout_minutes: 2,
            checkpointing: None,
            rollback: None,
            metrics: false,
        }
    }
}

/// Extensions configuration
#[derive(Debug, Clone, Default, Serialize, Deserialize)]
pub struct ExtensionsConfig {
    #[serde(default)]
    pub auto_load: bool,
    pub oci_registry_url: Option<String>,
    pub oci_namespace: Option<String>,
    pub discovery_interval: Option<u64>,
    pub max_concurrent: Option<usize>,
    pub timeout: Option<u64>,
    #[serde(default = "default_true")]
    pub sandbox: bool,
}

/// Monitoring configuration (from monitoring.ncl schema)
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct MonitoringConfig {
    pub enabled: bool,
    #[serde(default)]
    pub metrics: Option<MetricsConfig>,
    #[serde(default)]
    pub health_check: Option<HealthCheckConfig>,
    #[serde(default)]
    pub tracing: Option<TracingConfig>,
    #[serde(default)]
    pub alerting: Option<AlertingConfig>,
    #[serde(default)]
    pub resources: Option<ResourcesMonitoringConfig>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct MetricsConfig {
    #[serde(default)]
    pub enabled: bool,
    pub interval: Option<u64>,
    pub exporters: Option<Vec<String>>,
    pub prometheus_path: Option<String>,
    pub retention_days: Option<u64>,
    pub buffer_size: Option<usize>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct HealthCheckConfig {
    #[serde(default)]
    pub enabled: bool,
    pub check_type: Option<String>,
    pub interval: Option<u64>,
    pub timeout: Option<u64>,
    pub unhealthy_threshold: Option<usize>,
    pub healthy_threshold: Option<usize>,
    pub endpoint: Option<String>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct TracingConfig {
    #[serde(default)]
    pub enabled: bool,
    pub sample_rate: Option<f32>,
    pub backend: Option<String>,
    pub endpoint: Option<String>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct AlertingConfig {
    #[serde(default)]
    pub enabled: bool,
    pub rules_path: Option<String>,
    pub channels: Option<Vec<String>>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ResourcesMonitoringConfig {
    #[serde(default)]
    pub cpu: bool,
    #[serde(default)]
    pub memory: bool,
    #[serde(default)]
    pub disk: bool,
    #[serde(default)]
    pub network: bool,
    pub alert_threshold: Option<f64>,
}

impl Default for MonitoringConfig {
    fn default() -> Self {
        Self {
            enabled: true,
            metrics: Some(MetricsConfig {
                enabled: true,
                interval: Some(60),
                exporters: None,
                prometheus_path: None,
                retention_days: None,
                buffer_size: None,
            }),
            health_check: Some(HealthCheckConfig {
                enabled: true,
                check_type: None,
                interval: Some(30),
                timeout: None,
                unhealthy_threshold: None,
                healthy_threshold: None,
                endpoint: None,
            }),
            tracing: None,
            alerting: None,
            resources: Some(ResourcesMonitoringConfig {
                cpu: false,
                memory: false,
                disk: false,
                network: false,
                alert_threshold: None,
            }),
        }
    }
}

/// Logging configuration (from logging.ncl schema)
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct LoggingConfig {
    pub level: String,
    pub format: String,
    #[serde(default)]
    pub outputs: Option<Vec<String>>,
    #[serde(default)]
    pub file: Option<FileLoggingConfig>,
    #[serde(default)]
    pub syslog: Option<SyslogConfig>,
    #[serde(default)]
    pub fields: Option<FieldsConfig>,
    #[serde(default)]
    pub sampling: Option<SamplingConfig>,
    #[serde(default)]
    pub modules: Option<serde_json::Value>,
    #[serde(default)]
    pub performance: Option<PerformanceLoggingConfig>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct FileLoggingConfig {
    pub path: Option<String>,
    pub max_size: Option<u64>,
    pub max_backups: Option<usize>,
    pub max_age: Option<u64>,
    #[serde(default)]
    pub compress: bool,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct SyslogConfig {
    pub address: Option<String>,
    pub facility: Option<String>,
    pub protocol: Option<String>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct FieldsConfig {
    #[serde(default = "default_true")]
    pub service_name: bool,
    #[serde(default = "default_true")]
    pub hostname: bool,
    #[serde(default = "default_true")]
    pub pid: bool,
    #[serde(default = "default_true")]
    pub timestamp: bool,
    #[serde(default)]
    pub caller: bool,
    #[serde(default)]
    pub stack_trace: bool,
    pub custom: Option<Vec<String>>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct SamplingConfig {
    #[serde(default)]
    pub enabled: bool,
    pub initial: Option<u64>,
    pub thereafter: Option<u64>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct PerformanceLoggingConfig {
    #[serde(default)]
    pub enabled: bool,
    pub slow_threshold: Option<u64>,
    #[serde(default)]
    pub memory_info: bool,
}

impl Default for LoggingConfig {
    fn default() -> Self {
        Self {
            level: "info".to_string(),
            format: "text".to_string(),
            outputs: None,
            file: None,
            syslog: None,
            fields: None,
            sampling: None,
            modules: None,
            performance: None,
        }
    }
}

/// Security configuration (from security.ncl schema)
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct SecurityConfig {
    #[serde(default)]
    pub tls: Option<TlsConfig>,
    #[serde(default)]
    pub rbac: Option<RbacConfig>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct TlsConfig {
    pub enabled: bool,
    pub cert_path: Option<String>,
    pub key_path: Option<String>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct RbacConfig {
    pub enabled: bool,
    pub policy_file: Option<String>,
}

/// Performance tuning configuration
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct PerformanceConfig {
    #[serde(default)]
    pub profiling: bool,
    #[serde(default)]
    pub cpu_affinity: bool,
    #[serde(default)]
    pub memory_limits: Option<MemoryLimitsConfig>,
}

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct MemoryLimitsConfig {
    pub max_heap_mb: Option<u64>,
    pub gc_threshold: Option<u64>,
}

/// Docker build configuration (from docker-build.ncl schema)
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct DockerBuildConfig {
    pub base_image: String,
    #[serde(default)]
    pub build_args: HashMap<String, String>,
}

// Serde default helpers
fn default_retry_attempts() -> u32 {
    3
}
fn default_retry_delay() -> u64 {
    5000
}
fn default_task_timeout() -> u64 {
    3600000
}
fn default_persist() -> bool {
    true
}
fn default_parallel_limit() -> usize {
    5
}
fn default_operation_timeout() -> u64 {
    2
}
fn default_true() -> bool {
    true
}

impl Default for OrchestratorSettings {
    fn default() -> Self {
        Self {
            workspace: WorkspaceConfig {
                name: "default".to_string(),
                path: "/tmp".to_string(),
                enabled: true,
                multi_workspace: false,
            },
            server: ServerConfig::default(),
            storage: StorageConfig {
                backend: "filesystem".to_string(),
                path: Some("/tmp/orchestrator".to_string()),
                cache: None,
                compression: None,
                backup: None,
                replication: None,
            },
            queue: QueueConfig::default(),
            batch: BatchConfig::default(),
            extensions: ExtensionsConfig::default(),
            monitoring: Some(MonitoringConfig::default()),
            logging: Some(LoggingConfig::default()),
            security: None,
            performance: None,
            build: None,
        }
    }
}

impl OrchestratorConfig {
    /// Load configuration from Nickel orchestrator.ncl
    pub fn load() -> Result<Self> {
        let config_json = platform_config::load_service_config_from_ncl("orchestrator")
            .context("Failed to load orchestrator configuration from Nickel")?;

        let config: OrchestratorConfig = serde_json::from_value(config_json)
            .context("Failed to deserialize orchestrator configuration")?;

        let mut config = config;
        Self::apply_env_overrides(&mut config)?;
        Ok(config)
    }

    /// Apply environment variable overrides to configuration
    /// Environment variables use format: ORCHESTRATOR_{SECTION}_{KEY}=value
    fn apply_env_overrides(config: &mut Self) -> Result<()> {
        // Server overrides
        if let Ok(host) = std::env::var("ORCHESTRATOR_SERVER_HOST") {
            config.orchestrator.server.host = host;
        }
        if let Ok(port) = std::env::var("ORCHESTRATOR_SERVER_PORT") {
            config.orchestrator.server.port = port
                .parse()
                .context("ORCHESTRATOR_SERVER_PORT must be a valid port number")?;
        }
        if let Ok(workers) = std::env::var("ORCHESTRATOR_SERVER_WORKERS") {
            config.orchestrator.server.workers = workers
                .parse()
                .context("ORCHESTRATOR_SERVER_WORKERS must be a valid number")?;
        }

        // Logging overrides
        if let Ok(level) = std::env::var("ORCHESTRATOR_LOG_LEVEL") {
            if let Some(ref mut logging) = config.orchestrator.logging {
                logging.level = level;
            }
        }

        Ok(())
    }

    /// Apply CLI argument overrides to the configuration
    pub fn apply_cli_overrides(&mut self, args: &crate::Args) {
        if let Some(port) = args.port {
            self.orchestrator.server.port = port;
        }
    }

    /// Get server configuration
    pub fn server(&self) -> &ServerConfig {
        &self.orchestrator.server
    }

    /// Get storage configuration
    pub fn storage(&self) -> &StorageConfig {
        &self.orchestrator.storage
    }

    /// Get queue configuration
    pub fn queue(&self) -> &QueueConfig {
        &self.orchestrator.queue
    }

    /// Get batch configuration
    pub fn batch(&self) -> &BatchConfig {
        &self.orchestrator.batch
    }

    /// Get extensions configuration
    pub fn extensions(&self) -> &ExtensionsConfig {
        &self.orchestrator.extensions
    }

    /// Get logging configuration
    pub fn logging(&self) -> Option<&LoggingConfig> {
        self.orchestrator.logging.as_ref()
    }

    /// Get monitoring configuration
    pub fn monitoring(&self) -> Option<&MonitoringConfig> {
        self.orchestrator.monitoring.as_ref()
    }
}

impl ConfigLoader for OrchestratorConfig {
    fn service_name() -> &'static str {
        "orchestrator"
    }

    fn load_from_hierarchy() -> std::result::Result<Self, Box<dyn std::error::Error + Send + Sync>>
    {
        let service = Self::service_name();

        if let Some(path) = platform_config::resolve_config_path(service) {
            return Self::from_path(&path);
        }

        Ok(Self::default())
    }

    fn apply_env_overrides(
        &mut self,
    ) -> std::result::Result<(), Box<dyn std::error::Error + Send + Sync>> {
        Self::apply_env_overrides(self).map_err(|e| {
            Box::new(std::io::Error::other(e.to_string()))
                as Box<dyn std::error::Error + Send + Sync>
        })
    }

    fn from_path<P: AsRef<Path>>(
        path: P,
    ) -> std::result::Result<Self, Box<dyn std::error::Error + Send + Sync>> {
        let path = path.as_ref();
        let json_value = platform_config::format::load_config(path).map_err(|e| {
            let err: Box<dyn std::error::Error + Send + Sync> = Box::new(e);
            err
        })?;

        serde_json::from_value(json_value).map_err(|e| {
            let err_msg = format!(
                "Failed to deserialize orchestrator config from {:?}: {}",
                path, e
            );
            Box::new(std::io::Error::new(
                std::io::ErrorKind::InvalidData,
                err_msg,
            )) as Box<dyn std::error::Error + Send + Sync>
        })
    }
}

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

    #[test]
    fn test_default_server_config() {
        let server = ServerConfig::default();
        assert_eq!(server.host, "127.0.0.1");
        assert_eq!(server.port, 9090);
        assert_eq!(server.workers, 4);
    }

    #[test]
    fn test_default_queue_config() {
        let queue = QueueConfig::default();
        assert_eq!(queue.max_concurrent_tasks, 10);
        assert_eq!(queue.retry_attempts, 3);
        assert_eq!(queue.retry_delay, 5000);
    }

    #[test]
    fn test_default_batch_config() {
        let batch = BatchConfig::default();
        assert_eq!(batch.parallel_limit, 5);
        assert_eq!(batch.operation_timeout_minutes, 2);
    }
}