Make recent-changes feeds able to look farther back for revisions.

Before, grouped feeds would only use the last 30 revisions.
If part of a group was outside these revisions, it would be handled incorrectly.
Now, grouped feeds contain the last 30 groups, and use as many revisions as needed for this.

history/feed.go

@@ -12,14 +12,16 @@ import (
 	"github.com/gorilla/feeds"
 )
 
+const changeGroupMaxSize = 100
+
 func recentChangesFeed(opts FeedOptions) *feeds.Feed {
 	feed := &feeds.Feed{
 		Title:       cfg.WikiName + " (recent changes)",
 		Link:        &feeds.Link{Href: cfg.URL},
-		Description: "List of 30 recent changes on the wiki",
+		Description: fmt.Sprintf("List of %d recent changes on the wiki", changeGroupMaxSize),
 		Updated:     time.Now(),
 	}
-	revs := RecentChanges(30)
+	revs := newRecentChangesStream()
 	groups := opts.grouping.Group(revs)
 	for _, grp := range groups {
 		item := grp.feedItem(opts)
@@ -43,6 +45,7 @@ func RecentChangesJSON(opts FeedOptions) (string, error) {
 	return recentChangesFeed(opts).ToJSON()
 }
 
+// revisionGroup is a slice of revisions, ordered most recent first.
 type revisionGroup []Revision
 
 func newRevisionGroup(rev Revision) revisionGroup {
@@ -70,25 +73,34 @@ func groupRevisionsByMonth(revs []Revision) (res []revisionGroup) {
 	return res
 }
 
-// groupRevisionsByPeriodFromNow groups close-together revisions.
+// groupRevisionsByPeriodFromNow groups close-together revisions and returns the first changeGroupMaxSize (30) groups.
 // If two revisions happened within period of each other, they are put in the same group.
-func groupRevisionsByPeriod(revs []Revision, period time.Duration) (res []revisionGroup) {
-	if len(revs) == 0 {
+func groupRevisionsByPeriod(revs recentChangesStream, period time.Duration) (res []revisionGroup) {
+	nextRev := revs.iterator()
+	rev, empty := nextRev()
+	if empty {
 		return res
 	}
 
-	currTime := revs[0].Time
-	currGroup := newRevisionGroup(revs[0])
-	for _, rev := range revs[1:] {
+	currTime := rev.Time
+	currGroup := newRevisionGroup(rev)
+	for {
+		rev, done := nextRev()
+		if done {
+			return append(res, currGroup)
+		}
+
 		if currTime.Sub(rev.Time) < period && currGroup[0].Username == rev.Username {
 			currGroup.addRevision(rev)
 		} else {
 			res = append(res, currGroup)
+			if len(res) == changeGroupMaxSize {
+				return res
+			}
 			currGroup = newRevisionGroup(rev)
 		}
 		currTime = rev.Time
 	}
-	return res
 }
 
 func (grp revisionGroup) feedItem(opts FeedOptions) feeds.Item {
@@ -154,7 +166,7 @@ func ParseFeedOptions(query url.Values) (FeedOptions, error) {
 }
 
 type FeedGrouping interface {
-	Group([]Revision) []revisionGroup
+	Group(recentChangesStream) []revisionGroup
 }
 
 func parseFeedGrouping(query url.Values) (FeedGrouping, error) {
@@ -171,8 +183,8 @@ func parseFeedGrouping(query url.Values) (FeedGrouping, error) {
 
 type NormalFeedGrouping struct{}
 
-func (NormalFeedGrouping) Group(revs []Revision) (res []revisionGroup) {
-	for _, rev := range revs {
+func (NormalFeedGrouping) Group(revs recentChangesStream) (res []revisionGroup) {
+	for _, rev := range revs.next(changeGroupMaxSize) {
 		res = append(res, newRevisionGroup(rev))
 	}
 	return res
@@ -182,7 +194,7 @@ type PeriodFeedGrouping struct {
 	Period time.Duration
 }
 
-func (g PeriodFeedGrouping) Group(revs []Revision) (res []revisionGroup) {
+func (g PeriodFeedGrouping) Group(revs recentChangesStream) (res []revisionGroup) {
 	return groupRevisionsByPeriod(revs, g.Period)
 }
 
@@ -195,8 +207,7 @@ const (
 
 func parseFeedGroupOrder(query url.Values) (FeedGroupOrder, error) {
 	switch query.Get("order") {
-	case "old-to-new":
-	case "":
+	case "", "old-to-new":
 		return OldToNew, nil
 	case "new-to-old":
 		return NewToOld, nil

history/revision.go

@@ -21,53 +21,96 @@ type Revision struct {
 	hyphaeAffectedBuf []string
 }
 
+// gitLog calls `git log` and parses the results.
+func gitLog(args ...string) ([]Revision, error) {
+	args = append([]string{
+		"log", "--abbrev-commit", "--no-merges",
+		"--pretty=format:%h\t%ae\t%at\t%s",
+	}, args...)
+	out, err := silentGitsh(args...)
+	if err != nil {
+		return nil, err
+	}
+
+	outStr := out.String()
+	if outStr == "" {
+		// if there are no commits to return
+		return nil, nil
+	}
+
+	var revs []Revision
+	for _, line := range strings.Split(outStr, "\n") {
+		revs = append(revs, parseRevisionLine(line))
+	}
+	return revs, nil
+}
+
+type recentChangesStream struct {
+	currHash string
+}
+
+func newRecentChangesStream() recentChangesStream {
+	// next returns the next n revisions from the stream, ordered most recent first.
+	// If there are less than n revisions remaining, it will return only those.
+	return recentChangesStream{currHash: ""}
+}
+
+func (stream *recentChangesStream) next(n int) []Revision {
+	args := []string{"--max-count=" + strconv.Itoa(n)}
+	if stream.currHash == "" {
+		args = append(args, "HEAD")
+	} else {
+		// currHash is the last revision from the last call, so skip it
+		args = append(args, "--skip=1", stream.currHash)
+	}
+	// I don't think this can fail, so ignore the error
+	res, _ := gitLog(args...)
+	if len(res) != 0 {
+		stream.currHash = res[len(res)-1].Hash
+	}
+	return res
+}
+
+// recentChangesIterator returns a function that returns successive revisions from the stream.
+// It buffers revisions to avoid calling git every time.
+func (stream recentChangesStream) iterator() func() (Revision, bool) {
+	var buf []Revision
+	return func() (Revision, bool) {
+		if len(buf) == 0 {
+			// no real reason to choose 30, just needs some large number
+			buf = stream.next(30)
+			if len(buf) == 0 {
+				// revs has no revisions left
+				return Revision{}, true
+			}
+		}
+		rev := buf[0]
+		buf = buf[1:]
+		return rev, false
+	}
+}
+
 // RecentChanges gathers an arbitrary number of latest changes in form of revisions slice, ordered most recent first.
 func RecentChanges(n int) []Revision {
-	var (
-		out, err = silentGitsh(
-			"log", "--oneline", "--no-merges",
-			"--pretty=format:%h\t%ae\t%at\t%s",
-			"--max-count="+strconv.Itoa(n),
-		)
-		revs []Revision
-	)
-	if err == nil {
-		for _, line := range strings.Split(out.String(), "\n") {
-			revs = append(revs, parseRevisionLine(line))
-		}
-	}
+	stream := newRecentChangesStream()
+	revs := stream.next(n)
 	log.Printf("Found %d recent changes", len(revs))
 	return revs
 }
 
+// Revisions returns slice of revisions for the given hypha name, ordered most recent first.
+func Revisions(hyphaName string) ([]Revision, error) {
+	revs, err := gitLog("--", hyphaName+".*")
+	log.Printf("Found %d revisions for ‘%s’\n", len(revs), hyphaName)
+	return revs, err
+}
+
 // FileChanged tells you if the file has been changed since the last commit.
 func FileChanged(path string) bool {
 	_, err := gitsh("diff", "--exit-code", path)
 	return err != nil
 }
 
-// Revisions returns slice of revisions for the given hypha name, ordered most recent first.
-func Revisions(hyphaName string) ([]Revision, error) {
-	var (
-		out, err = silentGitsh(
-			"log", "--oneline", "--no-merges",
-			// Hash, author email, author time, commit msg separated by tab
-			"--pretty=format:%h\t%ae\t%at\t%s",
-			"--", hyphaName+".*",
-		)
-		revs []Revision
-	)
-	if err == nil {
-		for _, line := range strings.Split(out.String(), "\n") {
-			if line != "" {
-				revs = append(revs, parseRevisionLine(line))
-			}
-		}
-	}
-	log.Printf("Found %d revisions for ‘%s’\n", len(revs), hyphaName)
-	return revs, err
-}
-
 // Return time like dd — 13:42
 func (rev *Revision) timeToDisplay() string {
 	D := rev.Time.Day()