minstrel/internal/prowlarr/prowlarr.go

// Package prowlarr searches through Prowlarr's native API (M502 #5518).
// Ported from Fabled Librarian's internal/prowlarr.
//
// Native rather than Torznab because Prowlarr has no aggregate Torznab feed:
// `/api/v1/indexer/all/newznab` 404s exactly like an indexer id that does not
// exist (measured by Fabled Librarian against a real Prowlarr, its #4304).
// `/api/v1/search` is the one place Prowlarr asks every indexer at once.
package prowlarr

import (
	"context"
	"encoding/json"
	"fmt"
	"net/http"
	"net/url"
	"strings"
	"time"

	"git.fabledsword.com/bvandeusen/minstrel/internal/indexer"
)

// SearchPath is Prowlarr's aggregate search.
const SearchPath = "/api/v1/search"

// DefaultTimeout is longer than a Torznab feed's: Prowlarr waits on every
// indexer it has before answering.
const DefaultTimeout = 60 * time.Second

// Config is one Prowlarr.
type Config struct {
	// Code is the acquisition_services row, stamped onto every Result.
	Code    string
	BaseURL string
	// APIKey is required: Prowlarr's API refuses every call without one.
	APIKey  string
	Timeout time.Duration
}

// Client searches one Prowlarr and fetches .torrent files through it.
type Client struct {
	Config Config
	// HTTP is injectable for tests. Nil takes http.DefaultClient.
	HTTP *http.Client
}

type searchResult struct {
	Title       string `json:"title"`
	InfoHash    string `json:"infoHash"`
	GUID        string `json:"guid"`
	MagnetURL   string `json:"magnetUrl"`
	DownloadURL string `json:"downloadUrl"`
	Size        int64  `json:"size"`
	Seeders     int    `json:"seeders"`
	Leechers    int    `json:"leechers"`
	Indexer     string `json:"indexer"`
	PublishDate string `json:"publishDate"`
	Categories  []struct {
		ID int `json:"id"`
	} `json:"categories"`
}

// Search asks every indexer in this Prowlarr. `type=search` is free text;
// `type=music` takes artist and album separately, and a release name does not
// reliably say which part is which.
func (c *Client) Search(ctx context.Context, q indexer.Query) ([]indexer.Result, error) {
	params := url.Values{"query": {q.Text}, "type": {"search"}}
	for _, cat := range q.Categories {
		params.Add("categories", fmt.Sprint(cat))
	}
	endpoint := strings.TrimRight(c.Config.BaseURL, "/") + SearchPath + "?" + params.Encode()
	body, err := c.get(ctx, endpoint)
	if err != nil {
		return nil, err
	}
	out, err := parseResults(body)
	if err != nil {
		return nil, err
	}
	for i := range out {
		out[i].Source = c.Config.Code
	}
	return out, nil
}

func parseResults(body []byte) ([]indexer.Result, error) {
	var rows []searchResult
	if err := json.Unmarshal(body, &rows); err != nil {
		return nil, fmt.Errorf("%w: parse search results: %v", indexer.ErrUnreachable, err)
	}
	out := make([]indexer.Result, 0, len(rows))
	for _, row := range rows {
		title := strings.TrimSpace(row.Title)
		if title == "" {
			continue
		}
		r := indexer.Result{
			Title:       title,
			SizeBytes:   row.Size,
			Seeders:     row.Seeders,
			Peers:       row.Leechers,
			Indexer:     strings.TrimSpace(row.Indexer),
			Magnet:      firstMagnet(row.MagnetURL, row.GUID),
			DownloadURL: strings.TrimSpace(row.DownloadURL),
		}
		if len(row.Categories) > 0 {
			r.Category = row.Categories[0].ID
		}
		r.InfoHash = indexer.NormalizeHash(row.InfoHash)
		if r.InfoHash == "" {
			r.InfoHash = indexer.MagnetHash(r.Magnet)
		}
		if r.InfoHash == "" && indexer.IsInfoHash(row.GUID) {
			r.InfoHash = indexer.NormalizeHash(row.GUID)
		}
		if t, err := time.Parse(time.RFC3339, strings.TrimSpace(row.PublishDate)); err == nil {
			r.Published = t
		}
		out = append(out, r)
	}
	return out, nil
}

func firstMagnet(candidates ...string) string {
	for _, c := range candidates {
		c = strings.TrimSpace(c)
		if strings.HasPrefix(c, "magnet:") {
			return c
		}
	}
	return ""
}

// Download fetches a result's .torrent through Prowlarr, which holds the
// tracker credentials. A public tracker's link often redirects to a magnet,
// which comes back as an *indexer.MagnetLink.
func (c *Client) Download(ctx context.Context, r indexer.Result) ([]byte, error) {
	if r.DownloadURL == "" {
		return nil, fmt.Errorf("%w: %q has no download URL", indexer.ErrUnreachable, r.Title)
	}
	body, err := c.get(ctx, r.DownloadURL)
	if err != nil {
		return nil, err
	}
	return indexer.TorrentBody(r.Title, body)
}

func (c *Client) get(ctx context.Context, endpoint string) ([]byte, error) {
	if c.Config.BaseURL == "" || c.Config.APIKey == "" {
		return nil, indexer.ErrUnconfigured
	}
	timeout := c.Config.Timeout
	if timeout <= 0 {
		timeout = DefaultTimeout
	}
	ctx, cancel := context.WithTimeout(ctx, timeout)
	defer cancel()
	req, err := http.NewRequestWithContext(ctx, http.MethodGet, endpoint, nil)
	if err != nil {
		return nil, fmt.Errorf("%w: build request: %v", indexer.ErrUnreachable, err)
	}
	req.Header.Set("User-Agent", indexer.UserAgent)
	// In a header, not ?apikey=, so the key stays out of URLs, logs and
	// error strings.
	req.Header.Set("X-Api-Key", c.Config.APIKey)
	req.Header.Set("Accept", "application/json")
	resp, body, err := indexer.Send(c.HTTP, req)
	if err != nil {
		return nil, err
	}
	if resp.StatusCode == http.StatusUnauthorized || resp.StatusCode == http.StatusForbidden {
		return nil, fmt.Errorf("%w: HTTP %d", indexer.ErrAuth, resp.StatusCode)
	}
	if err := indexer.StatusError(resp, "Prowlarr"); err != nil {
		return nil, err
	}
	return body, nil
}