[performance] add statusfilter cache to cache calculated status filtering results (#4303)

this adds another 'filter' type cache, similar to the visibility and mute caches, to cache the results of status filtering checks. for the moment this keeps all the check calls themselves within the frontend typeconversion code, but i may move this out of the typeconverter in a future PR (also removing the ErrHideStatus means of propagating a hidden status).

also tweaks some of the cache invalidation hooks to not make unnecessary calls.

Reviewed-on: https://codeberg.org/superseriousbusiness/gotosocial/pulls/4303
Co-authored-by: kim <grufwub@gmail.com>
Co-committed-by: kim <grufwub@gmail.com>

internal/cache/cache.go

@@ -41,21 +41,22 @@ 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]
-
-	// Timelines ...
+	// Timelines provides access to the
+	// collection of timeline object caches,
+	// used in timeline lookups and streaming.
 	Timelines TimelineCaches
 
 	// Mutes provides access to the item mutes
 	// cache. (used by the item mutes filter).
-	Mutes MutesCache
+	Mutes StructCache[*CachedMute]
+
+	// StatusFilter provides access to the status filter
+	// cache. (used by the status-filter results filter).
+	StatusFilter StructCache[*CachedStatusFilterResults]
 
 	// Visibility provides access to the item visibility
 	// cache. (used by the visibility filter).
-	Visibility VisibilityCache
+	Visibility StructCache[*CachedVisibility]
 
 	// Webfinger provides access to the webfinger URL cache.
 	Webfinger *ttl.Cache[string, string] // TTL=24hr, sweep=5min
@@ -119,7 +120,6 @@ func (c *Caches) Init() {
 	c.initStatusEdit()
 	c.initStatusFave()
 	c.initStatusFaveIDs()
-	c.initStatusesFilterableFields()
 	c.initTag()
 	c.initThreadMute()
 	c.initToken()
@@ -131,6 +131,7 @@ func (c *Caches) Init() {
 	c.initWebPushSubscription()
 	c.initWebPushSubscriptionIDs()
 	c.initMutes()
+	c.initStatusFilter()
 	c.initVisibility()
 }
 
@@ -143,10 +144,6 @@ func (c *Caches) Start() error {
 		return gtserror.New("could not start webfinger cache")
 	}
 
-	if !c.StatusesFilterableFields.Start(5 * time.Minute) {
-		return gtserror.New("could not start statusesFilterableFields cache")
-	}
-
 	return nil
 }
 
@@ -158,9 +155,6 @@ func (c *Caches) Stop() {
 	if c.Webfinger != nil {
 		_ = c.Webfinger.Stop()
 	}
-	if c.StatusesFilterableFields != nil {
-		_ = c.StatusesFilterableFields.Stop()
-	}
 }
 
 // Sweep will sweep all the available caches to ensure none
@@ -183,6 +177,7 @@ func (c *Caches) Sweep(threshold float64) {
 	c.DB.Emoji.Trim(threshold)
 	c.DB.EmojiCategory.Trim(threshold)
 	c.DB.Filter.Trim(threshold)
+	c.DB.FilterIDs.Trim(threshold)
 	c.DB.FilterKeyword.Trim(threshold)
 	c.DB.FilterStatus.Trim(threshold)
 	c.DB.Follow.Trim(threshold)
@@ -218,20 +213,13 @@ func (c *Caches) Sweep(threshold float64) {
 	c.DB.User.Trim(threshold)
 	c.DB.UserMute.Trim(threshold)
 	c.DB.UserMuteIDs.Trim(threshold)
+	c.Mutes.Trim(threshold)
+	c.StatusFilter.Trim(threshold)
 	c.Timelines.Home.Trim()
 	c.Timelines.List.Trim()
 	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(

internal/cache/invalidate.go

@@ -26,19 +26,23 @@ import (
 // as an invalidation indicates a database INSERT / UPDATE / DELETE.
 // NOTE THEY ARE ONLY CALLED WHEN THE ITEM IS IN THE CACHE, SO FOR
 // HOOKS TO BE CALLED ON DELETE YOU MUST FIRST POPULATE IT IN THE CACHE.
+//
+// Also note that while Timelines are a part of the Caches{} object,
+// they are generally not modified as part of side-effects here, as
+// they often need specific IDs or more information that can only be
+// fetched from the database. As such, they are generally handled as
+// side-effects in the ./internal/processor/workers/ package.
 
 func (c *Caches) OnInvalidateAccount(account *gtsmodel.Account) {
-	// Invalidate cached stats objects for this account.
-	c.DB.AccountStats.Invalidate("AccountID", account.ID)
-
 	// Invalidate as possible visibility target result.
 	c.Visibility.Invalidate("ItemID", account.ID)
 
 	// If account is local, invalidate as
-	// possible mute / visibility result requester.
+	// possible visibility result requester,
+	// also, invalidate any cached stats.
 	if account.IsLocal() {
+		c.DB.AccountStats.Invalidate("AccountID", account.ID)
 		c.Visibility.Invalidate("RequesterID", account.ID)
-		c.Mutes.Invalidate("RequesterID", account.ID)
 	}
 
 	// Invalidate this account's
@@ -94,9 +98,8 @@ func (c *Caches) OnInvalidateBlock(block *gtsmodel.Block) {
 		localAccountIDs = append(localAccountIDs, block.TargetAccountID)
 	}
 
-	// Now perform local mute / visibility result invalidations.
+	// Now perform local visibility result invalidations.
 	c.Visibility.InvalidateIDs("RequesterID", localAccountIDs)
-	c.Mutes.InvalidateIDs("RequesterID", localAccountIDs)
 
 	// Invalidate source account's block lists.
 	c.DB.BlockIDs.Invalidate(block.AccountID)
@@ -120,9 +123,8 @@ func (c *Caches) OnInvalidateFilter(filter *gtsmodel.Filter) {
 	c.DB.FilterKeyword.InvalidateIDs("ID", filter.KeywordIDs)
 	c.DB.FilterStatus.InvalidateIDs("ID", filter.StatusIDs)
 
-	// Invalidate account's timelines (in case local).
-	c.Timelines.Home.Unprepare(filter.AccountID)
-	c.Timelines.List.Unprepare(filter.AccountID)
+	// Invalidate account's status filter cache.
+	c.StatusFilter.Invalidate("RequesterID", filter.AccountID)
 }
 
 func (c *Caches) OnInvalidateFilterKeyword(filterKeyword *gtsmodel.FilterKeyword) {
@@ -161,9 +163,8 @@ func (c *Caches) OnInvalidateFollow(follow *gtsmodel.Follow) {
 		localAccountIDs = append(localAccountIDs, follow.TargetAccountID)
 	}
 
-	// Now perform local mute / visibility result invalidations.
+	// Now perform local visibility result invalidations.
 	c.Visibility.InvalidateIDs("RequesterID", localAccountIDs)
-	c.Mutes.InvalidateIDs("RequesterID", localAccountIDs)
 
 	// Invalidate ID slice cache.
 	c.DB.FollowIDs.Invalidate(
@@ -295,6 +296,9 @@ func (c *Caches) OnInvalidateStatus(status *gtsmodel.Status) {
 	// Invalidate cached stats objects for this account.
 	c.DB.AccountStats.Invalidate("AccountID", status.AccountID)
 
+	// Invalidate filter results targeting status.
+	c.StatusFilter.Invalidate("StatusID", status.ID)
+
 	// Invalidate status ID cached visibility.
 	c.Visibility.Invalidate("ItemID", status.ID)
 

internal/cache/mutes.go

@@ -25,10 +25,6 @@ import (
 	"codeberg.org/gruf/go-structr"
 )
 
-type MutesCache struct {
-	StructCache[*CachedMute]
-}
-
 func (c *Caches) initMutes() {
 	// Calculate maximum cache size.
 	cap := calculateResultCacheMax(

internal/cache/size.go

@@ -25,6 +25,7 @@ import (
 	"unsafe"
 
 	"code.superseriousbusiness.org/gotosocial/internal/ap"
+	apimodel "code.superseriousbusiness.org/gotosocial/internal/api/model"
 	"code.superseriousbusiness.org/gotosocial/internal/config"
 	"code.superseriousbusiness.org/gotosocial/internal/gtsmodel"
 	"code.superseriousbusiness.org/gotosocial/internal/id"
@@ -653,6 +654,20 @@ func sizeofStatusFave() uintptr {
 	}))
 }
 
+func sizeofStatusFilterResults() uintptr {
+	return uintptr(size.Of(&CachedStatusFilterResults{
+		StatusID:    exampleID,
+		RequesterID: exampleID,
+		Results: [5][]StatusFilterResult{
+			{{Result: &apimodel.FilterResult{KeywordMatches: []string{"key", "word"}}}, {Result: &apimodel.FilterResult{StatusMatches: []string{exampleID, exampleID}}}, {}},
+			{{Result: &apimodel.FilterResult{KeywordMatches: []string{"key", "word"}}}, {Result: &apimodel.FilterResult{StatusMatches: []string{exampleID, exampleID}}}, {}},
+			{{Result: &apimodel.FilterResult{KeywordMatches: []string{"key", "word"}}}, {Result: &apimodel.FilterResult{StatusMatches: []string{exampleID, exampleID}}}, {}},
+			{{Result: &apimodel.FilterResult{KeywordMatches: []string{"key", "word"}}}, {Result: &apimodel.FilterResult{StatusMatches: []string{exampleID, exampleID}}}, {}},
+			{{Result: &apimodel.FilterResult{KeywordMatches: []string{"key", "word"}}}, {Result: &apimodel.FilterResult{StatusMatches: []string{exampleID, exampleID}}}, {}},
+		},
+	}))
+}
+
 func sizeofTag() uintptr {
 	return uintptr(size.Of(&gtsmodel.Tag{
 		ID:        exampleID,

internal/cache/statusfilter.go

@@ -0,0 +1,107 @@
+// GoToSocial
+// Copyright (C) GoToSocial Authors admin@gotosocial.org
+// SPDX-License-Identifier: AGPL-3.0-or-later
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cache
+
+import (
+	"time"
+
+	apimodel "code.superseriousbusiness.org/gotosocial/internal/api/model"
+	"code.superseriousbusiness.org/gotosocial/internal/config"
+	"code.superseriousbusiness.org/gotosocial/internal/log"
+	"codeberg.org/gruf/go-structr"
+)
+
+func (c *Caches) initStatusFilter() {
+	// Calculate maximum cache size.
+	cap := calculateResultCacheMax(
+		sizeofStatusFilterResults(), // model in-mem size.
+		config.GetCacheStatusFilterMemRatio(),
+	)
+
+	log.Infof(nil, "cache size = %d", cap)
+
+	copyF := func(r1 *CachedStatusFilterResults) *CachedStatusFilterResults {
+		r2 := new(CachedStatusFilterResults)
+		*r2 = *r1
+		return r2
+	}
+
+	c.StatusFilter.Init(structr.CacheConfig[*CachedStatusFilterResults]{
+		Indices: []structr.IndexConfig{
+			{Fields: "RequesterID,StatusID"},
+			{Fields: "StatusID", Multiple: true},
+			{Fields: "RequesterID", Multiple: true},
+		},
+		MaxSize: cap,
+		IgnoreErr: func(err error) bool {
+			// don't cache any errors,
+			// it gets a little too tricky
+			// otherwise with ensuring
+			// errors are cleared out
+			return true
+		},
+		Copy: copyF,
+	})
+}
+
+const (
+	KeyContextHome = iota
+	KeyContextPublic
+	KeyContextNotifs
+	KeyContextThread
+	KeyContextAccount
+	keysLen // must always be last in list
+)
+
+// CachedStatusFilterResults contains the
+// results of a cached status filter lookup.
+type CachedStatusFilterResults struct {
+
+	// StatusID is the ID of the
+	// status this is a result for.
+	StatusID string
+
+	// RequesterID is the ID of the requesting
+	// account for this status filter lookup.
+	RequesterID string
+
+	// Results is a map (int-key-array) of status filter
+	// result slices in all possible filtering contexts.
+	Results [keysLen][]StatusFilterResult
+}
+
+// StatusFilterResult stores a single (positive,
+// i.e. match) filter result for a status by a filter.
+type StatusFilterResult struct {
+
+	// Expiry stores the time at which
+	// (if any) the filter result expires.
+	Expiry time.Time
+
+	// Result stores any generated filter result for
+	// this match intended to be shown at the frontend.
+	// This can be used to determine the filter action:
+	// - value => gtsmodel.FilterActionWarn
+	// - nil => gtsmodel.FilterActionHide
+	Result *apimodel.FilterResult
+}
+
+// Expired returns whether the filter result has expired.
+func (r *StatusFilterResult) Expired(now time.Time) bool {
+	return !r.Expiry.IsZero() && !r.Expiry.After(now)
+}

internal/cache/visibility.go

@@ -23,10 +23,6 @@ import (
 	"codeberg.org/gruf/go-structr"
 )
 
-type VisibilityCache struct {
-	StructCache[*CachedVisibility]
-}
-
 func (c *Caches) initVisibility() {
 	// Calculate maximum cache size.
 	cap := calculateResultCacheMax(