RSS show author & ability to filter items (#311)

* Show author of RSS feed item if available

closes #298

Signed-off-by: Nikos Filippakis <me@nfil.dev>

* Filter RSS feed items by keywords

closes #236, closes #296

Signed-off-by: Nikos Filippakis <me@nfil.dev>

config.sample.yaml

@@ -100,6 +100,16 @@ services:
       feeds:
         "http://lorem-rss.herokuapp.com/feed?unit=second&interval=60":
           rooms: ["!qmElAGdFYCHoCJuaNt:localhost"]
+          must_include:
+            author:
+              - author1
+            description:
+              - lorem
+              - ipsum
+          must_not_include:
+            title:
+              - Lorem
+              - Ipsum
 
   - ID: "github_cmd_service"
     Type: "github"

src/github.com/matrix-org/go-neb/services/rssbot/rssbot.go

@@ -7,7 +7,9 @@ import (
 	"html"
 	"net/http"
 	"strconv"
+	"strings"
 	"time"
+	"unicode"
 
 	log "github.com/Sirupsen/logrus"
 	"github.com/die-net/lrucache"
@@ -34,6 +36,30 @@ var (
 
 const minPollingIntervalSeconds = 60 * 5 // 5 min (News feeds can be genuinely spammy)
 
+// includeRules contains the rules for including or excluding a feed item. For the fields Author, Title
+// and Description in a feed item, there can be some words specified in the config that determine whether
+// the item will be displayed or not, depending on whether these words are included in that field.
+//
+//   - If specified in the `must_include` field, the feed item must include at least one word for each field
+//     that has been specified. This means that if some words have been specified for both Author and Title,
+//     both the Author and Title must contain at least one of their respective words or the item will be skipped.
+//   - If specified in the `must_not_include` field, the feed item fields must not contain any of the words
+//     that were specified for each field. This means that if some words have been specified for both Author
+//     and Title, if either of them includes at least one of their respective words, the item will be skipped,
+//     even in the case that the item matched the `must_include` rules.
+//
+//   In both cases, specifying an empty list for a field or not specifying anything causes the field to be ignored.
+//   The field being checked each time will be split into words (any non-alphanumeric character starts a new word)
+//   and they will be checked against the provided list.
+type includeRules struct {
+	// Author is a case-sensitive list of words that the author name must contain or not contain.
+	Author []string `json:"author"`
+	// Title is a case-sensitive list of words that the author name must contain or not contain.
+	Title []string `json:"title"`
+	// Description is a case-sensitive list of words that the author name must contain or not contain.
+	Description []string `json:"description"`
+}
+
 // Service contains the Config fields for this service.
 //
 // Example request:
@@ -62,6 +88,10 @@ type Service struct {
 		// The time of the last successful poll. This is populated by Go-NEB. Use /getService to retrieve
 		// this value.
 		FeedUpdatedTimestampSecs int64 `json:"last_updated_ts_secs"`
+		// Specified fields must each include at least one of these words.
+		MustInclude includeRules `json:"must_include"`
+		// None of the specified fields must include any of these words.
+		MustNotInclude includeRules `json:"must_not_include"`
 		// Internal field. When we should poll again.
 		NextPollTimestampSecs int64
 		// Internal field. The most recently seen GUIDs. Sized to the number of items in the feed.
@@ -302,7 +332,44 @@ func (s *Service) queryFeed(feedURL string) (*gofeed.Feed, []gofeed.Item, error)
 	return feed, items, nil
 }
 
+// containsAny takes a string and an array of words and returns whether any of the words
+// in the list are contained in the string. The words in the string are considered to be
+// separated by any non-alphanumeric character.
+func containsAny(item string, filterWords []string) bool {
+	itemWords := strings.FieldsFunc(item, func(c rune) bool {
+		return !unicode.IsLetter(c) && !unicode.IsNumber(c)
+	})
+	for _, itemWord := range itemWords {
+		for _, filterWord := range filterWords {
+			if filterWord == itemWord {
+				return true
+			}
+		}
+	}
+	return false
+}
+
+func itemFiltered(i *gofeed.Item, mustInclude, mustNotInclude *includeRules) bool {
+	// At least one word for each field that has been specified must be included for an item to pass the filter.
+	if (i.Author != nil && len(mustInclude.Author) > 0 && !containsAny(i.Author.Name, mustInclude.Author)) ||
+		(len(mustInclude.Title) > 0 && !containsAny(i.Title, mustInclude.Title)) ||
+		(len(mustInclude.Description) > 0 && !containsAny(i.Description, mustInclude.Description)) {
+		return true
+	}
+
+	// If at least one word of any field that has been specified is included in the item, it doesn't pass the filter.
+	if (i.Author != nil && containsAny(i.Author.Name, mustNotInclude.Author)) ||
+		containsAny(i.Title, mustNotInclude.Title) ||
+		containsAny(i.Description, mustNotInclude.Description) {
+		return true
+	}
+	return false
+}
+
 func (s *Service) newItems(feedURL string, allItems []*gofeed.Item) (items []gofeed.Item) {
+	mustInclude := s.Feeds[feedURL].MustInclude
+	mustNotInclude := s.Feeds[feedURL].MustNotInclude
+
 	for _, i := range allItems {
 		if i == nil {
 			continue
@@ -327,9 +394,15 @@ func (s *Service) newItems(feedURL string, allItems []*gofeed.Item) (items []gof
 		//   This will inevitably break for some people, but that group of people are probably smaller, so *shrug*.
 		i.Title = html.UnescapeString(i.Title)
 		i.Description = html.UnescapeString(i.Description)
+		if i.Author != nil {
+			i.Author.Name = html.UnescapeString(i.Author.Name)
+			i.Author.Email = html.UnescapeString(i.Author.Email)
+		}
 
+		if !itemFiltered(i, &mustInclude, &mustNotInclude) {
 			items = append(items, *i)
 		}
+	}
 	return
 }
 
@@ -356,13 +429,25 @@ func itemToHTML(feed *gofeed.Feed, item gofeed.Item) gomatrix.HTMLMessage {
 		itemTitle = feed.Title
 	}
 
+	fmtBody := fmt.Sprintf("<strong>%s</strong>:<br><a href=\"%s\"><strong>%s</strong></a>",
+		html.EscapeString(feed.Title), html.EscapeString(item.Link), html.EscapeString(itemTitle))
+	if item.Author != nil {
+		if len(item.Author.Name) > 0 && len(item.Author.Email) > 0 {
+			fmtBody += fmt.Sprintf(" by <a href=\"mailto:%s\">%s</a>", html.EscapeString(item.Author.Email),
+				html.EscapeString(item.Author.Name))
+		} else if len(item.Author.Name) > 0 {
+			fmtBody += fmt.Sprintf(" by %s", html.EscapeString(item.Author.Name))
+		} else if len(item.Author.Email) > 0 {
+			fmtBody += fmt.Sprintf(" by <a href=\"mailto:%s\">%s</a>", html.EscapeString(item.Author.Email),
+				html.EscapeString(item.Author.Email))
+		}
+	}
 	return gomatrix.HTMLMessage{
 		Body: fmt.Sprintf("%s: %s ( %s )",
-			html.EscapeString(feed.Title), html.EscapeString(item.Title), html.EscapeString(item.Link)),
+			html.EscapeString(feed.Title), html.EscapeString(itemTitle), html.EscapeString(item.Link)),
 		MsgType:       "m.notice",
 		Format:        "org.matrix.custom.html",
-		FormattedBody: fmt.Sprintf("<strong>%s</strong>:<br><a href=\"%s\"><strong>%s</strong></a>",
-			html.EscapeString(feed.Title), html.EscapeString(item.Link), html.EscapeString(itemTitle)),
+		FormattedBody: fmtBody,
 		// <strong>FeedTitle</strong>:
 		// <br>
 		// <a href="url-of-the-entry"><strong>Title of the Entry</strong></a>

src/github.com/matrix-org/go-neb/services/rssbot/rssbot_test.go

@@ -32,13 +32,13 @@ const rssFeedXML = `
 	<item>
 		<title>New Item: Majora&#8217;s Mask</title>
 		<link>http://go.neb/rss/majoras-mask</link>
+		<author>The Skullkid!</author>
 	</item>
 </channel>
 </rss>`
 
-func TestHTMLEntities(t *testing.T) {
+func createRSSClient(t *testing.T, feedURL string) *Service {
 	database.SetServiceDB(&database.NopStorage{})
-	feedURL := "https://thehappymaskshop.hyrule"
 	// Replace the cachingClient with a mock so we can intercept RSS requests
 	rssTrans := testutils.NewRoundTripper(func(req *http.Request) (*http.Response, error) {
 		if req.URL.String() != feedURL {
@@ -55,9 +55,11 @@ func TestHTMLEntities(t *testing.T) {
 	srv, err := types.CreateService("id", "rssbot", "@happy_mask_salesman:hyrule", []byte(
 		`{"feeds": {"`+feedURL+`":{}}}`, // no config yet
 	))
+
 	if err != nil {
-		t.Fatal("Failed to create RSS bot: ", err)
+		t.Fatal(err)
 	}
+
 	rssbot := srv.(*Service)
 
 	// Configure the service to force OnPoll to query the RSS feed and attempt to send results
@@ -67,6 +69,14 @@ func TestHTMLEntities(t *testing.T) {
 	f.NextPollTimestampSecs = time.Now().Unix()
 	rssbot.Feeds[feedURL] = f
 
+	return rssbot
+}
+
+func TestHTMLEntities(t *testing.T) {
+	feedURL := "https://thehappymaskshop.hyrule"
+
+	rssbot := createRSSClient(t, feedURL)
+
 	// Create the Matrix client which will send the notification
 	wg := sync.WaitGroup{}
 	wg.Add(1)
@@ -103,3 +113,59 @@ func TestHTMLEntities(t *testing.T) {
 	// Check that the Matrix client sent a message
 	wg.Wait()
 }
+
+func TestFeedItemFiltering(t *testing.T) {
+	feedURL := "https://thehappymaskshop.hyrule"
+
+	// Create rssbot client
+	rssbot := createRSSClient(t, feedURL)
+
+	feed := rssbot.Feeds[feedURL]
+	feed.MustInclude.Title = []string{"Zelda"}
+	rssbot.Feeds[feedURL] = feed
+
+	_, items, _ := rssbot.queryFeed(feedURL)
+	// Expect that we get no items if we filter for 'Zelda' in title
+	if len(items) != 0 {
+		t.Errorf("Expected 0 items, got %v", items)
+	}
+
+	// Recreate rssbot client
+	rssbot = createRSSClient(t, feedURL)
+
+	feed = rssbot.Feeds[feedURL]
+	feed.MustInclude.Title = []string{"Majora"}
+	rssbot.Feeds[feedURL] = feed
+
+	_, items, _ = rssbot.queryFeed(feedURL)
+	// Expect one item if we filter for 'Majora' in title
+	if len(items) != 1 {
+		t.Errorf("Expected 1 item, got %d", len(items))
+	}
+
+	// Recreate rssbot client
+	rssbot = createRSSClient(t, feedURL)
+
+	feed = rssbot.Feeds[feedURL]
+	feed.MustNotInclude.Author = []string{"kid"}
+	rssbot.Feeds[feedURL] = feed
+
+	_, items, _ = rssbot.queryFeed(feedURL)
+	// 'kid' does not match an entire word in the author name, so it's not filtered
+	if len(items) != 1 {
+		t.Errorf("Expected 1 item, got %d", len(items))
+	}
+
+	// Recreate rssbot client
+	rssbot = createRSSClient(t, feedURL)
+
+	feed = rssbot.Feeds[feedURL]
+	feed.MustNotInclude.Author = []string{"Skullkid"}
+	rssbot.Feeds[feedURL] = feed
+
+	_, items, _ = rssbot.queryFeed(feedURL)
+	// Expect no items if we filter for 'Skullkid' not in author name
+	if len(items) != 0 {
+		t.Errorf("Expected 0 items, got %v", items)
+	}
+}