predictor/internal/service/service.go

package service

import (
	"context"
	"os"
	"path/filepath"
	"sort"
	"strings"
	"sync"
	"time"

	"predictor-refactored/internal/dataset"
	"predictor-refactored/internal/downloader"
	"predictor-refactored/internal/elevation"

	"go.uber.org/zap"
)

// Service orchestrates the dataset lifecycle and provides prediction capabilities.
type Service struct {
	mu       sync.RWMutex
	ds       *dataset.File
	elev     *elevation.Dataset
	cfg      *downloader.Config
	dl       *downloader.Downloader
	log      *zap.Logger
	updating sync.Mutex // prevents concurrent downloads
}

// New creates a new Service.
func New(cfg *downloader.Config, log *zap.Logger) *Service {
	return &Service{
		cfg: cfg,
		dl:  downloader.NewDownloader(cfg, log),
		log: log,
	}
}

// LoadElevation loads the ruaumoko-compatible elevation dataset from path.
// If the file doesn't exist, elevation termination is disabled (falls back to sea level).
func (s *Service) LoadElevation(path string) {
	ds, err := elevation.Open(path)
	if err != nil {
		s.log.Warn("elevation dataset not available, using sea-level termination",
			zap.String("path", path), zap.Error(err))
		return
	}
	s.elev = ds
	s.log.Info("elevation dataset loaded", zap.String("path", path))
}

// Elevation returns the elevation dataset (may be nil).
func (s *Service) Elevation() *elevation.Dataset {
	return s.elev
}

// Ready returns true if the service has a loaded dataset.
func (s *Service) Ready() bool {
	s.mu.RLock()
	defer s.mu.RUnlock()
	return s.ds != nil
}

// DatasetTime returns the forecast time of the currently loaded dataset.
func (s *Service) DatasetTime() (time.Time, bool) {
	s.mu.RLock()
	defer s.mu.RUnlock()
	if s.ds == nil {
		return time.Time{}, false
	}
	return s.ds.DSTime, true
}

// Dataset returns the current dataset for reading.
func (s *Service) Dataset() *dataset.File {
	s.mu.RLock()
	defer s.mu.RUnlock()
	return s.ds
}

// Update checks for and downloads new forecast data if needed.
func (s *Service) Update(ctx context.Context) error {
	if !s.updating.TryLock() {
		s.log.Info("update already in progress, skipping")
		return nil
	}
	defer s.updating.Unlock()

	// Check if current dataset is still fresh
	if dsTime, ok := s.DatasetTime(); ok {
		if time.Since(dsTime) < s.cfg.DatasetTTL {
			s.log.Info("dataset still fresh, skipping update",
				zap.Time("dataset_time", dsTime),
				zap.Duration("age", time.Since(dsTime)))
			return nil
		}
	}

	// Try loading an existing dataset from disk first
	if err := s.loadExistingDataset(); err == nil {
		return nil
	}

	// Find latest available model run
	run, err := s.dl.FindLatestRun(ctx)
	if err != nil {
		return err
	}

	// Download and assemble
	path, err := s.dl.Download(ctx, run)
	if err != nil {
		return err
	}

	// Open the new dataset
	ds, err := dataset.Open(path, run)
	if err != nil {
		return err
	}

	// Swap in the new dataset
	s.setDataset(ds)
	s.log.Info("dataset loaded", zap.Time("run", run), zap.String("path", path))

	// Clean old datasets
	s.cleanOldDatasets(path)

	return nil
}

// loadExistingDataset tries to find and load an existing dataset from the data directory.
func (s *Service) loadExistingDataset() error {
	entries, err := os.ReadDir(s.cfg.DataDir)
	if err != nil {
		return err
	}

	// Collect valid dataset files (name is YYYYMMDDHH, no extension, correct size)
	type candidate struct {
		name string
		path string
		run  time.Time
	}
	var candidates []candidate

	for _, e := range entries {
		if e.IsDir() || strings.Contains(e.Name(), ".") {
			continue
		}
		if len(e.Name()) != 10 {
			continue
		}

		run, err := time.Parse("2006010215", e.Name())
		if err != nil {
			continue
		}

		path := filepath.Join(s.cfg.DataDir, e.Name())
		info, err := os.Stat(path)
		if err != nil || info.Size() != dataset.DatasetSize {
			continue
		}

		if time.Since(run) > s.cfg.DatasetTTL {
			continue
		}

		candidates = append(candidates, candidate{name: e.Name(), path: path, run: run})
	}

	if len(candidates) == 0 {
		return os.ErrNotExist
	}

	// Pick the newest
	sort.Slice(candidates, func(i, j int) bool {
		return candidates[i].run.After(candidates[j].run)
	})

	best := candidates[0]
	ds, err := dataset.Open(best.path, best.run)
	if err != nil {
		return err
	}

	s.setDataset(ds)
	s.log.Info("loaded existing dataset",
		zap.Time("run", best.run),
		zap.String("path", best.path))
	return nil
}

// setDataset swaps the current dataset with a new one, closing the old one.
func (s *Service) setDataset(ds *dataset.File) {
	s.mu.Lock()
	old := s.ds
	s.ds = ds
	s.mu.Unlock()

	if old != nil {
		if err := old.Close(); err != nil {
			s.log.Error("failed to close old dataset", zap.Error(err))
		}
	}
}

// cleanOldDatasets removes dataset files other than the one at keepPath.
func (s *Service) cleanOldDatasets(keepPath string) {
	entries, err := os.ReadDir(s.cfg.DataDir)
	if err != nil {
		return
	}

	for _, e := range entries {
		if e.IsDir() {
			continue
		}
		path := filepath.Join(s.cfg.DataDir, e.Name())
		if path == keepPath {
			continue
		}
		// Remove old datasets and temp files
		if len(e.Name()) == 10 || strings.HasSuffix(e.Name(), ".downloading") {
			if err := os.Remove(path); err != nil {
				s.log.Warn("failed to remove old file", zap.String("path", path), zap.Error(err))
			} else {
				s.log.Info("removed old dataset", zap.String("path", path))
			}
		}
	}
}

// Close releases all resources.
func (s *Service) Close() error {
	s.mu.Lock()
	defer s.mu.Unlock()
	if s.ds != nil {
		err := s.ds.Close()
		s.ds = nil
		return err
	}
	return nil
}