[performance] reduce InboxForward->Create calls by partially implemen…

…ting Exists() (#3647)

* alphabetical reordering

* keep a cache of activity IDs we have handled creates for

* reduce number of inbox forwarding create calls by partially implementing Exists()

* increase cache size, since all we're storing is string keys

internal/cache/cache.go

@@ -40,18 +40,18 @@ type Caches struct {
 	// the block []headerfilter.Filter cache.
 	BlockHeaderFilters headerfilter.Cache
 
+	// TTL cache of statuses -> filterable text fields.
+	// To ensure up-to-date fields, cache is keyed as:
+	// `[status.ID][status.UpdatedAt.Unix()]`
+	StatusesFilterableFields *ttl.Cache[string, []string]
+
 	// Visibility provides access to the item visibility
 	// cache. (used by the visibility filter).
 	Visibility VisibilityCache
 
 	// Webfinger provides access to the webfinger URL cache.
 	Webfinger *ttl.Cache[string, string] // TTL=24hr, sweep=5min
 
-	// TTL cache of statuses -> filterable text fields.
-	// To ensure up-to-date fields, cache is keyed as:
-	// `[status.ID][status.UpdatedAt.Unix()]`
-	StatusesFilterableFields *ttl.Cache[string, []string]
-
 	// prevent pass-by-value.
 	_ nocopy
 }
@@ -203,6 +203,15 @@ func (c *Caches) Sweep(threshold float64) {
 	c.Visibility.Trim(threshold)
 }
 
+func (c *Caches) initStatusesFilterableFields() {
+	c.StatusesFilterableFields = new(ttl.Cache[string, []string])
+	c.StatusesFilterableFields.Init(
+		0,
+		512,
+		1*time.Hour,
+	)
+}
+
 func (c *Caches) initWebfinger() {
 	// Calculate maximum cache size.
 	cap := calculateCacheMax(
@@ -219,12 +228,3 @@ func (c *Caches) initWebfinger() {
 		24*time.Hour,
 	)
 }
-
-func (c *Caches) initStatusesFilterableFields() {
-	c.StatusesFilterableFields = new(ttl.Cache[string, []string])
-	c.StatusesFilterableFields.Init(
-		0,
-		512,
-		1*time.Hour,
-	)
-}

internal/federation/federatingdb/create.go

@@ -63,6 +63,10 @@ func (f *federatingDB) Create(ctx context.Context, asType vocab.Type) error {
 		return nil
 	}
 
+	// Cache entry for this create activity ID for later
+	// checks in the Exist() function if we see it again.
+	f.activityIDs.Set(ap.GetJSONLDId(asType).String(), struct{}{})
+
 	switch name := asType.GetTypeName(); name {
 	case ap.ActivityBlock:
 		// BLOCK SOMETHING

internal/federation/federatingdb/db.go

@@ -21,6 +21,7 @@ import (
 	"context"
 	"net/url"
 
+	"codeberg.org/gruf/go-cache/v3/simple"
 	"github.com/superseriousbusiness/activity/pub"
 	"github.com/superseriousbusiness/activity/streams/vocab"
 	"github.com/superseriousbusiness/gotosocial/internal/filter/interaction"
@@ -61,6 +62,10 @@ type federatingDB struct {
 	visFilter  *visibility.Filter
 	intFilter  *interaction.Filter
 	spamFilter *spam.Filter
+
+	// tracks Activity IDs we have handled creates for,
+	// for use in the Exists() function during forwarding.
+	activityIDs simple.Cache[string, struct{}]
 }
 
 // New returns a DB that satisfies the pub.Database
@@ -79,5 +84,6 @@ func New(
 		intFilter:  intFilter,
 		spamFilter: spamFilter,
 	}
+	fdb.activityIDs.Init(0, 2048)
 	return &fdb
 }

internal/federation/federatingdb/exists.go

@@ -22,12 +22,8 @@ import (
 	"net/url"
 )
 
-// Exists returns true if the database has an entry for the specified
-// id. It may not be owned by this application instance.
-//
-// The library makes this call only after acquiring a lock first.
-//
-// Implementation note: this just straight up isn't implemented, and doesn't *really* need to be either.
+// Exists is an implementation of pub.Database{}.Exists(), optimized specifically for
+// the only usecase in which go-fed/activity/pub actually calls it. Do not use otherwise!
 func (f *federatingDB) Exists(ctx context.Context, id *url.URL) (exists bool, err error) {
-	return false, nil
+	return f.activityIDs.Has(id.String()), nil
 }

internal/federation/federator.go

@@ -33,7 +33,7 @@ import (
 var _ interface {
 	pub.CommonBehavior
 	pub.FederatingProtocol
-} = &Federator{}
+} = (*Federator)(nil)
 
 type Federator struct {
 	db                  db.DB

internal/federation/transport.go

@@ -49,7 +49,7 @@ import (
 // Note that the library will not maintain a long-lived pointer to the
 // returned Transport so that any private credentials are able to be
 // garbage collected.
-func (f *Federator) NewTransport(ctx context.Context, actorBoxIRI *url.URL, gofedAgent string) (pub.Transport, error) {
+func (f *Federator) NewTransport(ctx context.Context, actorBoxIRI *url.URL, _ string) (pub.Transport, error) {
 	var username string
 	var err error