amfora/subscriptions/subscriptions.go

377 lines
7.9 KiB
Go
Raw Normal View History

2020-11-28 01:01:29 +03:00
package subscriptions
import (
2020-08-17 00:42:45 +03:00
"crypto/sha256"
"encoding/json"
"errors"
"fmt"
"io"
"io/ioutil"
2020-08-17 00:42:45 +03:00
"mime"
"os"
2020-08-17 00:42:45 +03:00
"path"
"reflect"
"strings"
2020-08-17 00:42:45 +03:00
"sync"
"time"
2020-08-17 00:42:45 +03:00
"github.com/makeworld-the-better-one/amfora/client"
"github.com/makeworld-the-better-one/amfora/config"
2020-08-17 00:42:45 +03:00
"github.com/makeworld-the-better-one/go-gemini"
"github.com/mmcdole/gofeed"
2020-11-19 05:24:26 +03:00
"github.com/spf13/viper"
)
2020-08-17 00:42:45 +03:00
var (
ErrSaving = errors.New("couldn't save JSON to disk")
ErrNotSuccess = errors.New("status 20 not returned")
ErrNotFeed = errors.New("not a valid feed")
)
2020-11-28 01:01:29 +03:00
var writeMu = sync.Mutex{} // Prevent concurrent writes to subscriptions.json file
// LastUpdated is the time when the in-memory data was last updated.
2020-11-28 01:01:29 +03:00
// It can be used to know if the subscriptions page should be regenerated.
var LastUpdated time.Time
// Init should be called after config.Init.
func Init() error {
2020-11-28 01:01:29 +03:00
f, err := os.Open(config.SubscriptionPath)
if err == nil {
// File exists and could be opened
fi, err := f.Stat()
if err == nil && fi.Size() > 0 {
// File is not empty
jsonBytes, err := ioutil.ReadAll(f)
f.Close()
if err != nil {
return fmt.Errorf("read subscriptions.json error: %w", err)
}
err = json.Unmarshal(jsonBytes, &data)
if err != nil {
2020-12-07 06:15:24 +03:00
return fmt.Errorf("subscriptions.json is corrupted: %w", err)
}
}
f.Close()
} else if !os.IsNotExist(err) {
// There's an error opening the file, but it's not bc is doesn't exist
2020-12-07 06:15:24 +03:00
return fmt.Errorf("open subscriptions.json error: %w", err)
} else {
// File does not exist, initialize maps
data.Feeds = make(map[string]*gofeed.Feed)
data.Pages = make(map[string]*pageJSON)
}
2020-08-17 00:42:45 +03:00
LastUpdated = time.Now()
2020-11-19 05:24:26 +03:00
2020-11-28 01:01:29 +03:00
if viper.GetInt("subscriptions.update_interval") > 0 {
// Update subscriptions every so often
2020-11-19 05:24:26 +03:00
go func() {
for {
updateAll()
2020-11-28 01:01:29 +03:00
time.Sleep(time.Duration(viper.GetInt("subscriptions.update_interval")) * time.Second)
2020-11-19 05:24:26 +03:00
}
}()
} else {
2020-11-28 01:01:29 +03:00
// User disabled automatic updates
2020-11-19 05:24:26 +03:00
// So just update once at the beginning
go updateAll()
}
2020-08-17 22:33:53 +03:00
return nil
}
2020-11-28 01:01:29 +03:00
// IsSubscribed returns true if the URL is already subscribed to,
// whether a feed or page.
func IsSubscribed(url string) bool {
2020-08-17 00:42:45 +03:00
data.feedMu.RLock()
for u := range data.Feeds {
if url == u {
2020-08-17 00:42:45 +03:00
data.feedMu.RUnlock()
return true
}
}
2020-08-17 00:42:45 +03:00
data.feedMu.RUnlock()
data.pageMu.RLock()
for u := range data.Pages {
if url == u {
2020-08-17 00:42:45 +03:00
data.pageMu.RUnlock()
return true
}
}
2020-08-17 00:42:45 +03:00
data.pageMu.RUnlock()
return false
}
// GetFeed returns a Feed object and a bool indicating whether the passed
// content was actually recognized as a feed.
func GetFeed(mediatype, filename string, r io.Reader) (*gofeed.Feed, bool) {
if r == nil {
return nil, false
}
// Check mediatype and filename
2020-11-17 19:26:49 +03:00
if mediatype != "application/atom+xml" && mediatype != "application/rss+xml" && mediatype != "application/json+feed" &&
filename != "atom.xml" && filename != "feed.xml" && filename != "feed.json" &&
2020-12-17 20:00:34 +03:00
!strings.HasSuffix(filename, ".atom") && !strings.HasSuffix(filename, ".rss") && !strings.HasSuffix(filename, ".xml") {
// No part of the above is true
return nil, false
}
feed, err := gofeed.NewParser().Parse(r)
2020-08-17 00:42:45 +03:00
if feed == nil {
return nil, false
}
return feed, err == nil
}
2020-08-28 19:18:30 +03:00
func writeJSON() error {
2020-08-17 00:42:45 +03:00
writeMu.Lock()
defer writeMu.Unlock()
data.Lock()
jsonBytes, err := json.MarshalIndent(&data, "", " ")
data.Unlock()
if err != nil {
return err
}
err = ioutil.WriteFile(config.SubscriptionPath, jsonBytes, 0666)
if err != nil {
return err
}
return nil
}
// AddFeed stores a feed.
2020-08-17 00:42:45 +03:00
// It can be used to update a feed for a URL, although the package
// will handle that on its own.
func AddFeed(url string, feed *gofeed.Feed) error {
2020-08-17 00:42:45 +03:00
if feed == nil {
panic("feed is nil")
}
// Remove any unused fields to save memory and disk space
feed.Image = nil
feed.Generator = ""
feed.Categories = nil
feed.DublinCoreExt = nil
feed.ITunesExt = nil
feed.Custom = nil
feed.Link = ""
feed.Links = nil
2020-08-17 00:42:45 +03:00
for _, item := range feed.Items {
item.Description = ""
2020-08-17 00:42:45 +03:00
item.Content = ""
item.Image = nil
item.Categories = nil
item.Enclosures = nil
item.DublinCoreExt = nil
item.ITunesExt = nil
item.Extensions = nil
item.Custom = nil
item.Link = "" // Links is used instead
2020-08-17 00:42:45 +03:00
}
data.feedMu.Lock()
oldFeed, ok := data.Feeds[url]
if !ok || !reflect.DeepEqual(feed, oldFeed) {
// Feeds are different, or there was never an old one
LastUpdated = time.Now()
data.Feeds[url] = feed
data.feedMu.Unlock()
err := writeJSON()
if err != nil {
return ErrSaving
}
} else {
2020-08-17 00:42:45 +03:00
data.feedMu.Unlock()
}
return nil
}
// AddPage stores a page to track for changes.
// It can be used to update the page as well, although the package
// will handle that on its own.
func AddPage(url string, r io.Reader) error {
if r == nil {
return nil
}
h := sha256.New()
if _, err := io.Copy(h, r); err != nil {
return err
}
newHash := fmt.Sprintf("%x", h.Sum(nil))
2020-08-17 00:42:45 +03:00
data.pageMu.Lock()
_, ok := data.Pages[url]
if !ok || data.Pages[url].Hash != newHash {
// Page content is different, or it didn't exist
LastUpdated = time.Now()
data.Pages[url] = &pageJSON{
Hash: newHash,
Changed: time.Now().UTC(),
}
2020-08-17 00:42:45 +03:00
data.pageMu.Unlock()
err := writeJSON()
if err != nil {
return ErrSaving
}
} else {
data.pageMu.Unlock()
}
return nil
}
2020-08-17 00:42:45 +03:00
func updateFeed(url string) error {
res, err := client.Fetch(url)
if err != nil {
if res != nil {
res.Body.Close()
}
return err
}
defer res.Body.Close()
if res.Status != gemini.StatusSuccess {
return ErrNotSuccess
}
mediatype, _, err := mime.ParseMediaType(res.Meta)
if err != nil {
return err
}
filename := path.Base(url)
feed, ok := GetFeed(mediatype, filename, res.Body)
if !ok {
return ErrNotFeed
}
return AddFeed(url, feed)
}
func updatePage(url string) error {
res, err := client.Fetch(url)
if err != nil {
if res != nil {
res.Body.Close()
}
return err
}
defer res.Body.Close()
if res.Status != gemini.StatusSuccess {
return ErrNotSuccess
}
2020-08-17 20:31:45 +03:00
return AddPage(url, res.Body)
2020-08-17 00:42:45 +03:00
}
2020-11-28 01:01:29 +03:00
// updateAll updates all subscriptions using workers.
2020-11-17 19:26:49 +03:00
// It only returns once all the workers are done.
2020-08-17 00:42:45 +03:00
func updateAll() {
2020-08-17 20:31:45 +03:00
worker := func(jobs <-chan [2]string, wg *sync.WaitGroup) {
// Each job is: [2]string{<type>, "url"}
2020-08-17 20:31:45 +03:00
// where <type> is "feed" or "page"
defer wg.Done()
for j := range jobs {
if j[0] == "feed" {
2020-08-28 19:18:30 +03:00
updateFeed(j[1]) //nolint:errcheck
} else if j[0] == "page" {
2020-08-28 19:18:30 +03:00
updatePage(j[1]) //nolint:errcheck
2020-08-17 20:31:45 +03:00
}
}
}
var wg sync.WaitGroup
data.RLock()
numJobs := len(data.Feeds) + len(data.Pages)
jobs := make(chan [2]string, numJobs)
if numJobs == 0 {
data.RUnlock()
return
}
2020-11-28 01:01:29 +03:00
numWorkers := viper.GetInt("subscriptions.workers")
2020-11-19 05:24:26 +03:00
if numWorkers < 1 {
numWorkers = 1
}
// Start workers, waiting for jobs
for w := 0; w < numWorkers; w++ {
2020-08-17 20:31:45 +03:00
wg.Add(1)
2020-12-07 06:15:24 +03:00
go func() {
worker(jobs, &wg)
2020-12-07 06:15:24 +03:00
}()
2020-08-17 20:31:45 +03:00
}
// Get map keys in a slice
feedKeys := make([]string, len(data.Feeds))
i := 0
for k := range data.Feeds {
feedKeys[i] = k
i++
}
pageKeys := make([]string, len(data.Pages))
i = 0
for k := range data.Pages {
pageKeys[i] = k
i++
}
data.RUnlock()
for j := 0; j < numJobs; j++ {
if j < len(feedKeys) {
jobs <- [2]string{"feed", feedKeys[j]}
} else {
// In the Pages
jobs <- [2]string{"page", pageKeys[j-len(feedKeys)]}
}
}
close(jobs)
2020-08-17 00:42:45 +03:00
2020-08-17 20:31:45 +03:00
wg.Wait()
2020-08-17 00:42:45 +03:00
}
2020-12-07 04:57:57 +03:00
// AllURLs returns all the subscribed-to URLS.
func AllURLS() []string {
data.RLock()
defer data.RUnlock()
urls := make([]string, len(data.Feeds)+len(data.Pages))
i := 0
for k := range data.Feeds {
urls[i] = k
i++
}
for k := range data.Pages {
urls[i] = k
i++
}
return urls
}
// Remove removes a subscription from memory and from the disk.
// The URL must be provided. It will do nothing if the URL is
// not an actual subscription.
//
2020-12-07 04:57:57 +03:00
// It returns any errors that occurred when saving to disk.
func Remove(u string) error {
data.Lock()
// Just delete from both instead of using a loop to find it
delete(data.Feeds, u)
delete(data.Pages, u)
data.Unlock()
return writeJSON()
}