zurg/internal/torrent/repair.go

package torrent

import (
	"context"
	"fmt"
	"math"
	"strings"
	"sync"
	"time"

	"github.com/debridmediamanager/zurg/internal/config"
	"github.com/debridmediamanager/zurg/pkg/realdebrid"
	mapset "github.com/deckarep/golang-set/v2"
	cmap "github.com/orcaman/concurrent-map/v2"
)

const (
	EXPIRED_LINK_TOLERANCE_HOURS = 24
)

func (t *TorrentManager) StartRepairJob() {
	if !t.Config.EnableRepair() {
		t.log.Debug("Repair is disabled, skipping repair job")
		return
	}
	t.repairTrigger = make(chan *Torrent)
	t.repairQueue = mapset.NewSet[*Torrent]()
	// there is 1 repair worker, with max 1 blocking task
	go func() {
		t.log.Debug("Starting periodic repair job")
		repairTicker := time.NewTicker(time.Duration(t.Config.GetRepairEveryMins()) * time.Minute)
		defer repairTicker.Stop()

		for {
			select {
			case <-repairTicker.C:
				t.invokeRepair(nil)
			case torrent := <-t.repairTrigger:
				// On-demand trigger with a specific torrent
				t.invokeRepair(torrent)
			case <-t.RepairKillSwitch:
				t.log.Info("Stopping periodic repair job")
				return
			}
		}
	}()
}

func (t *TorrentManager) invokeRepair(torrent *Torrent) {
	t.repairRunningMu.Lock()
	if t.repairRunning {
		t.repairRunningMu.Unlock()
		t.repairQueue.Add(torrent)
		// don't do anything if repair is already running
		return
	}

	t.repairRunning = true
	t.repairRunningMu.Unlock()

	// Execute the repair job
	t.repairAll(torrent)

	// After repair is done
	t.repairRunningMu.Lock()
	t.repairRunning = false
	t.repairRunningMu.Unlock()

	// before we let go, let's check repairQueue
	t.workerPool.Submit(func() {
		queuedTorrent, exists := t.repairQueue.Pop()
		if exists {
			t.TriggerRepair(queuedTorrent)
		}
	})
}

// TriggerRepair allows an on-demand repair to be initiated.
func (t *TorrentManager) TriggerRepair(torrent *Torrent) {
	t.repairTrigger <- torrent
}

func (t *TorrentManager) repairAll(torrent *Torrent) {
	// todo: a more elegant way to do this
	var haystack cmap.ConcurrentMap[string, *Torrent]
	if torrent == nil {
		haystack, _ = t.DirectoryMap.Get(INT_ALL)
		t.log.Debug("Periodic repair started; searching for broken torrents")
	} else {
		haystack = cmap.New[*Torrent]()
		haystack.Set("", torrent)
	}

	// collect all torrents that need to be repaired
	toRepair := mapset.NewSet[*Torrent]()

	var wg sync.WaitGroup
	haystack.IterCb(func(_ string, torrent *Torrent) {
		wg.Add(1)
		_ = t.workerPool.Submit(func() {
			defer wg.Done()
			if torrent.UnrepairableReason != "" {
				return
			}
			// check 1: for broken files
			brokenFileCount := 0
			torrent.SelectedFiles.IterCb(func(_ string, file *File) {
				if file.State.Is("broken_file") {
					brokenFileCount++
				}
			})
			if brokenFileCount > 0 {
				t.log.Debugf("Torrent %s has %d/%d broken files, adding to repair list", t.GetKey(torrent), brokenFileCount, torrent.SelectedFiles.Count())
				toRepair.Add(torrent)
				return
			}
			// check 2: for unassigned links (this means the torrent has started to deteriorate)
			unassignedCount := torrent.UnassignedLinks.Cardinality()
			if unassignedCount > 0 {
				t.log.Debugf("Torrent %s has %d unassigned links, adding to repair list", t.GetKey(torrent), unassignedCount)
				toRepair.Add(torrent)
				return
			}
		})
	})

	wg.Wait()

	toRepair.Each(func(torrent *Torrent) bool {
		wg.Add(1)
		t.Repair(torrent, &wg)
		return false
	})
	wg.Wait()

	t.log.Infof("Finished periodic repair sequence for %d broken torrent(s)", toRepair.Cardinality())
}

func (t *TorrentManager) Repair(torrent *Torrent, wg *sync.WaitGroup) {
	// blocks for approx 45 minutes if active torrents are full
	if !t.canCapacityHandle() {
		t.log.Error("Blocked for too long due to limit of active torrents, cannot continue with the repair")
		wg.Done()
		return
	}

	// assign to a worker
	_ = t.workerPool.Submit(func() {
		defer wg.Done()
		if err := torrent.State.Event(context.Background(), "repair_torrent"); err != nil {
			t.log.Errorf("Failed to mark torrent %s as under repair: %v", t.GetKey(torrent), err)
			return
		}
		t.repair(torrent)
	})
}

func (t *TorrentManager) repair(torrent *Torrent) {
	torrentIDs := torrent.Components.Keys()
	t.log.Infof("Started repair process for torrent %s (ids=%v)", t.GetKey(torrent), torrentIDs)

	// handle torrents with incomplete links for selected files
	// torrent can be rar'ed by RD, so we need to check for that
	if torrent.UnassignedLinks.Cardinality() > 0 && !t.assignUnassignedLinks(torrent) {
		return
	}

	// get all broken files
	brokenFiles, allBroken := getBrokenFiles(torrent)

	// first step: redownload the whole torrent

	t.log.Debugf("Torrent %s has %d broken files (out of %d), repairing by redownloading whole torrent", t.GetKey(torrent), len(brokenFiles), torrent.SelectedFiles.Count())

	info, err := t.redownloadTorrent(torrent, []string{}) // reinsert the whole torrent, passing empty selection
	if info != nil && info.Progress == 100 && !t.isStillBroken(info, brokenFiles) {
		// successful repair
		if err := torrent.State.Event(context.Background(), "mark_as_repaired"); err != nil {
			t.log.Errorf("Cannot repair torrent %s: %v", torrent.Hash, err)
			return
		}
		// delete old torrents
		torrent.Components.IterCb(func(id string, _ *realdebrid.TorrentInfo) {
			if id != info.ID {
				t.api.DeleteTorrent(id)
			}
		})
		t.log.Infof("Successfully repaired torrent %s by redownloading all files", t.GetKey(torrent))
		return
	} else if info != nil && info.Progress != 100 {
		t.log.Infof("Torrent %s is still in progress after redownloading but it should be repaired once done", t.GetKey(torrent))
		return
	}

	if err != nil {
		t.log.Warnf("Cannot repair torrent %s by redownloading all files (error=%s)", t.GetKey(torrent), err.Error())
	} else {
		t.log.Warnf("Cannot repair torrent %s by redownloading all files", t.GetKey(torrent))
	}

	if torrent.UnrepairableReason != "" {
		t.log.Debugf("Torrent %s has been marked as unfixable during redownload (%s), ending repair process early", t.GetKey(torrent), torrent.UnrepairableReason)
		return
	}

	// second step: download just the broken files

	if len(brokenFiles) == 1 && allBroken {
		// if all files are broken, we can't do anything!
		t.log.Warnf("Torrent %s has only 1 cached file and it's broken; marking as unfixable (to fix, select other files)", t.GetKey(torrent))
		t.markAsUnfixable(torrent, "the lone cached file is broken")
		return
	}

	t.log.Infof("Torrent %s will be repaired by downloading %d batches of the %d broken files", int(math.Ceil(float64(len(brokenFiles))/100)), len(brokenFiles), t.GetKey(torrent))

	newlyDownloadedIds := make([]string, 0)
	batchNum := 1
	brokenFileIDs := getFileIDs(brokenFiles)
	var group []string
	for _, fileIDStr := range brokenFileIDs {
		group = append(group, fileIDStr)
		if len(group) >= 100 {
			t.log.Debugf("Downloading batch %d of broken files of torrent %s", batchNum, t.GetKey(torrent))
			batchNum++
			redownloadedInfo, err := t.redownloadTorrent(torrent, group)
			if err != nil {
				t.log.Warnf("Cannot repair torrent %s by downloading broken files (error=%s) giving up", t.GetKey(torrent), err.Error())
				for _, newId := range newlyDownloadedIds {
					t.trash(newId)
				}
				return
			}
			newlyDownloadedIds = append(newlyDownloadedIds, redownloadedInfo.ID)
			group = make([]string, 0)
		}
	}

	t.log.Debugf("Downloading last batch of broken files of torrent %s", t.GetKey(torrent))

	if len(group) > 0 {
		redownloadedInfo, err := t.redownloadTorrent(torrent, group)
		if err != nil {
			t.log.Warnf("Cannot repair torrent %s by downloading broken files (error=%s) giving up", t.GetKey(torrent), err.Error())
			for _, newId := range newlyDownloadedIds {
				t.trash(newId)
			}
			return
		}
		newlyDownloadedIds = append(newlyDownloadedIds, redownloadedInfo.ID)
	}

	for _, newId := range newlyDownloadedIds {
		t.trashOnceCompleted(newId)
	}
}

func (t *TorrentManager) assignUnassignedLinks(torrent *Torrent) bool {
	unassignedTotal := torrent.UnassignedLinks.Cardinality()
	t.log.Infof("Trying to assign %d links to the %d selected of incomplete torrent %s", unassignedTotal, torrent.SelectedFiles.Count(), t.GetKey(torrent))

	// handle torrents with incomplete links for selected files
	assignedCount := 0
	expiredCount := 0
	rarCount := 0
	unassignedCount := 0
	newUnassignedLinks := cmap.New[*realdebrid.Download]()

	torrent.UnassignedLinks.Each(func(link string) bool {
		// unrestrict each unassigned link that was filled out during torrent init
		unrestrict := t.UnrestrictLinkUntilOk(link)
		if unrestrict == nil {
			expiredCount++
			return false // next unassigned link
		}

		// try to assign to a selected file
		assigned := false
		torrent.SelectedFiles.IterCb(func(_ string, file *File) {
			// base it on size because why not?
			if !assigned && file.State.Is("broken_file") && file.Bytes == unrestrict.Filesize && strings.HasSuffix(strings.ToLower(file.Path), strings.ToLower(unrestrict.Filename)) {
				file.Link = link
				if strings.HasPrefix(file.Link, "https://real-debrid.com/d/") {
					file.Link = file.Link[0:39]
				}
				if err := file.State.Event(context.Background(), "repair_file"); err != nil {
					t.log.Errorf("Failed to mark file %s as repaired: %v", file.Path, err)
					return
				}
				assigned = true
				assignedCount++
			}
		})

		if !assigned {
			// if not assigned and is a rar, likely it was rar'ed by RD
			if strings.HasSuffix(strings.ToLower(unrestrict.Filename), ".rar") {
				rarCount++
			} else {
				t.log.Warnf("Cannot assign %s to any file in torrent %s", unrestrict.Filename, t.GetKey(torrent))
			}
			newUnassignedLinks.Set(link, unrestrict)
		}

		processedCount := assignedCount + unassignedCount + expiredCount
		if processedCount%10 == 0 || processedCount == unassignedTotal {
			t.log.Infof("Processed %d out of %d links (%d expired) to broken torrent %s", processedCount, unassignedTotal, expiredCount, t.GetKey(torrent))
		}

		return false // next unassigned link
	})
	// magnet:?xt=urn:btih:ba8720dde2472e650a87efbb78efb4fbcea3f7ee

	// empty/reset the unassigned links as we have assigned them already
	if unassignedTotal > 0 {
		torrent.UnassignedLinks = mapset.NewSet[string]()
		t.writeTorrentToFile(torrent)
	}

	// this is a rar'ed torrent, nothing we can do
	if assignedCount == 0 && rarCount == 1 {
		if t.Config.ShouldDeleteRarFiles() {
			t.log.Warnf("Torrent %s is rar'ed and we cannot repair it, deleting it as configured", t.GetKey(torrent))
			t.Delete(t.GetKey(torrent), true)
		} else {
			t.log.Warnf("Torrent %s is rar'ed and we cannot repair it", t.GetKey(torrent))
			newUnassignedLinks.IterCb(func(_ string, unassigned *realdebrid.Download) {
				// if unassigned == nil {
				// 	return
				// }
				newFile := &File{
					File: realdebrid.File{
						ID:       0,
						Path:     unassigned.Filename,
						Bytes:    unassigned.Filesize,
						Selected: 0,
					},
					Ended: torrent.Added,
					Link:  unassigned.Link,
					State: NewFileState("ok_file"),
				}
				torrent.SelectedFiles.Set(unassigned.Filename, newFile)
			})
			torrent.UnassignedLinks = mapset.NewSet[string]()
			t.markAsUnfixable(torrent, "rar'ed by RD")
			t.markAsUnplayable(torrent, "rar'ed by RD")
			if err := torrent.State.Event(context.Background(), "mark_as_repaired"); err != nil {
				t.log.Errorf("Cannot repair torrent %s: %v", torrent.Hash, err)
			}
		}
		return false // end repair
	}

	return true
}

func (t *TorrentManager) redownloadTorrent(torrent *Torrent, selection []string) (*realdebrid.TorrentInfo, error) {
	// broken files means broken links
	// if brokenFiles is not provided, we will redownload all files
	finalSelection := strings.Join(selection, ",")
	if len(selection) == 0 {
		torrent.SelectedFiles.IterCb(func(_ string, file *File) {
			selection = append(selection, fmt.Sprintf("%d", file.ID))
		})
		if len(selection) == 0 {
			return nil, nil
		} else {
			finalSelection = strings.Join(selection, ",")
		}
	}

	// redownload torrent
	resp, err := t.api.AddMagnetHash(torrent.Hash)
	if err != nil {
		if strings.Contains(err.Error(), "infringing") {
			t.markAsUnfixable(torrent, "infringing torrent")
		} else if strings.Contains(err.Error(), "unsupported") {
			t.markAsUnfixable(torrent, "unsupported torrent")
		} else if strings.Contains(err.Error(), "unavailable") {
			t.markAsUnfixable(torrent, "unavailable torrent")
		} else if strings.Contains(err.Error(), "invalid") {
			t.markAsUnfixable(torrent, "invalid torrent")
		} else if strings.Contains(err.Error(), "big") {
			t.markAsUnfixable(torrent, "torrent too big")
		} else if strings.Contains(err.Error(), "allowed") {
			t.markAsUnfixable(torrent, "torrent not allowed")
		}
		return nil, fmt.Errorf("cannot redownload torrent: %v", err)
	}

	newTorrentID := resp.ID

	// sleep for 1 second to let RD process the magnet
	time.Sleep(1 * time.Second)

	var info *realdebrid.TorrentInfo
	for {
		// select files
		err = t.api.SelectTorrentFiles(newTorrentID, finalSelection)
		if err != nil {
			t.trash(newTorrentID)
			return nil, fmt.Errorf("cannot start redownloading torrent %s (id=%s): %v", t.GetKey(torrent), newTorrentID, err)
		}
		// sleep for 2 second to let RD process the magnet
		time.Sleep(2 * time.Second)

		// see if the torrent is ready
		info, err = t.api.GetTorrentInfo(newTorrentID)
		if err != nil {
			t.trash(newTorrentID)
			return nil, fmt.Errorf("cannot get info on redownloaded torrent %s (id=%s) : %v", t.GetKey(torrent), newTorrentID, err)
		}

		if info.Status == "magnet_conversion" {
			time.Sleep(60 * time.Second)
			continue
		} else if info.Status == "waiting_files_selection" {
			time.Sleep(10 * time.Second)
			return nil, fmt.Errorf("torrent %s (id=%s) is stuck on waiting_files_selection", t.GetKey(torrent), newTorrentID)
		}
		break
	}

	// documented status: magnet_error, magnet_conversion, waiting_files_selection, queued, downloading, downloaded, error, virus, compressing, uploading, dead
	isOkStatus := false
	okStatuses := []string{"downloading", "downloaded", "uploading", "queued", "compressing"}
	// not compressing because we need playable files
	for _, status := range okStatuses {
		if info.Status == status {
			isOkStatus = true
			break
		}
	}

	if !isOkStatus {
		t.trash(info.ID)
		return nil, fmt.Errorf("the redownloaded torrent %s is in a non-OK state: %s", t.GetKey(torrent), info.Status)
	}

	// check if incorrect number of links
	if info.Progress == 100 && len(info.Links) != len(selection) {
		t.trash(newTorrentID)
		return nil, fmt.Errorf("torrent %s only got %d links but we need %d", t.GetKey(torrent), len(info.Links), len(selection))
	}

	t.log.Infof("Redownloading torrent %s successful (id=%s, progress=%d)", t.GetKey(torrent), info.ID, info.Progress)

	return info, nil
}

func (t *TorrentManager) canCapacityHandle() bool {
	// max waiting time is 45 minutes
	const maxRetries = 50
	const baseDelay = 1 * time.Second
	const maxDelay = 60 * time.Second
	retryCount := 0
	for {
		count, err := t.api.GetActiveTorrentCount()
		if err != nil {
			t.log.Warnf("Cannot get active downloads count: %v", err)
			if retryCount >= maxRetries {
				t.log.Error("Max retries reached. Exiting.")
				return false
			}
			delay := time.Duration(math.Pow(2, float64(retryCount))) * baseDelay
			if delay > maxDelay {
				delay = maxDelay
			}
			time.Sleep(delay)
			retryCount++
			continue
		}

		if count.DownloadingCount < count.MaxNumberOfTorrents-1 {
			return true
		}

		delay := time.Duration(math.Pow(2, float64(retryCount))) * baseDelay
		if delay > maxDelay {
			delay = maxDelay
		}
		t.log.Infof("We have reached the max number of active torrents, waiting for %s seconds before retrying", delay)

		if retryCount >= maxRetries {
			t.log.Error("Max retries reached. Exiting.")
			return false
		}

		time.Sleep(delay)
		retryCount++
	}
}

func (t *TorrentManager) markAsUnplayable(torrent *Torrent, reason string) {
	t.log.Warnf("Torrent %s is unplayable (reason: %s), moving to unplayable directory", t.GetKey(torrent), reason)
	// reassign to unplayable torrents directory
	t.DirectoryMap.IterCb(func(directory string, torrents cmap.ConcurrentMap[string, *Torrent]) {
		if strings.HasPrefix(directory, "int__") {
			return
		}
		torrents.Remove(t.GetKey(torrent))
	})
	torrents, _ := t.DirectoryMap.Get(config.UNPLAYABLE_TORRENTS)
	torrents.Set(t.GetKey(torrent), torrent)
}

func (t *TorrentManager) markAsUnfixable(torrent *Torrent, reason string) {
	t.log.Warnf("Marking torrent %s as unfixable - %s", t.GetKey(torrent), reason)
	torrent.UnrepairableReason = reason
	t.writeTorrentToFile(torrent)
}

// getBrokenFiles returns the files that are not http links and not deleted
func getBrokenFiles(torrent *Torrent) ([]*File, bool) {
	var brokenFiles []*File
	allBroken := true
	torrent.SelectedFiles.IterCb(func(_ string, file *File) {
		if file.State.Is("broken_file") {
			brokenFiles = append(brokenFiles, file)
		} else {
			allBroken = false
		}
	})
	return brokenFiles, allBroken
}

// isStillBroken checks if the torrent is still broken
// if it's not broken anymore, it will assign the links to the files
func (t *TorrentManager) isStillBroken(info *realdebrid.TorrentInfo, brokenFiles []*File) bool {
	var selectedFiles []*File
	for _, file := range info.Files {
		if file.Selected == 0 {
			continue
		}
		selectedFiles = append(selectedFiles, &File{
			File:  file,
			Ended: info.Ended,
			Link:  "",
			State: NewFileState("broken_file"),
		})
	}
	if len(selectedFiles) == len(info.Links) {
		// all links are still intact! good!
		for i, file := range selectedFiles {
			file.Link = info.Links[i]
			if strings.HasPrefix(file.Link, "https://real-debrid.com/d/") {
				file.Link = file.Link[0:39]
			}
			if err := file.State.Event(context.Background(), "repair_file"); err != nil {
				t.log.Errorf("Failed to mark file %s as repaired: %v", file.Path, err)
				return true
			}
		}
	} else {
		// if we can't assign links then it's still broken
		return true
	}

	if len(brokenFiles) == 0 {
		// just check for the last file
		brokenFiles = append(brokenFiles, selectedFiles[len(selectedFiles)-1])
	}

	// check if the broken files can now be unrestricted
	for _, oldFile := range brokenFiles {
		for idx, newFile := range selectedFiles {
			if oldFile.Bytes == newFile.Bytes {
				unrestrict := t.UnrestrictFileUntilOk(selectedFiles[idx])
				if unrestrict == nil || oldFile.Bytes != unrestrict.Filesize {
					return true
				}
			}
		}
	}
	return false
}