import NewNotesButton from '@/components/NewNotesButton' import { ExtendedKind, FIRST_RELAY_RESULT_GRACE_MS } from '@/constants' import { collectEmbeddedEventPrefetchTargets, getReplaceableCoordinateFromEvent, isMentioningMutedUsers, isReplaceableEvent, isReplyNoteEvent } from '@/lib/event' import { shouldFilterEvent } from '@/lib/event-filtering' import { isRelayUrlStrictSupersetIdentityKey, stableSpellFeedFilterKey } from '@/lib/spell-feed-request-identity' import { normalizeUrl } from '@/lib/url' import { getZapInfoFromEvent } from '@/lib/event-metadata' import { isTouchDevice } from '@/lib/utils' import { useContentPolicy } from '@/providers/ContentPolicyProvider' import { useDeletedEvent } from '@/providers/DeletedEventProvider' import { useMuteList } from '@/providers/MuteListProvider' import { useNostr } from '@/providers/NostrProvider' import { useUserTrust } from '@/providers/UserTrustProvider' import { useZap } from '@/providers/ZapProvider' import client from '@/services/client.service' import { TFeedSubRequest } from '@/types' import dayjs from 'dayjs' import { Event, kinds } from 'nostr-tools' import { decode } from 'nostr-tools/nip19' import { forwardRef, useCallback, useEffect, useImperativeHandle, useLayoutEffect, useMemo, useRef, useState } from 'react' import { useTranslation } from 'react-i18next' import PullToRefresh from 'react-simple-pull-to-refresh' import { formatPubkey, pubkeyToNpub } from '@/lib/pubkey' import { NoteFeedProfileContext, type NoteFeedProfileContextValue } from '@/providers/NoteFeedProfileContext' import type { TProfile } from '@/types' import NoteCard, { NoteCardLoadingSkeleton } from '../NoteCard' const LIMIT = 100 // Increased from 200 to load more events per request const ALGO_LIMIT = 200 // Increased from 500 for algorithm feeds const SHOW_COUNT = 20 // Increased from 10 to show more events at once, reducing scroll load frequency /** Hard cap after merging parallel one-shot fetches (e.g. interests = one REQ per topic). */ const ONE_SHOT_MERGED_CAP =100 const FEED_PROFILE_BATCH_DEBOUNCE_MS = 120 const FEED_PROFILE_CHUNK = 36 function mergeEventBatchesById(prev: Event[], incoming: Event[], cap: number): Event[] { const byId = new Map() for (const e of prev) { byId.set(e.id, e) } for (const e of incoming) { byId.set(e.id, e) } return Array.from(byId.values()) .sort((a, b) => b.created_at - a.created_at) .slice(0, cap) } const NoteList = forwardRef( ( { subRequests, showKinds, showKind1OPs = true, showKind1Replies = true, showKind1111 = true, filterMutedNotes = true, hideReplies = false, hideUntrustedNotes = false, areAlgoRelays = false, pinnedEventIds = [], useFilterAsIs = false, extraShouldHideEvent, /** When set (e.g. Spells page), timeline subscription keys off this string instead of `subRequests` reference churn. */ feedSubscriptionKey, /** * When true (e.g. Explore relay reviews), `subRequests` may grow after first paint (bootstrap relays → full list). * Re-subscribe when URLs change but **merge** new timeline batches into existing rows by event id instead of clearing. */ preserveTimelineOnSubRequestsChange = false, /** * Spells page: after this many ms, clear the loading skeleton so the list area renders; subscription keeps running. */ spellFetchTimeoutMs, /** Spells page: bumps when user picks a feed; used with {@link onSpellFeedFirstPaint}. */ spellFeedInstrumentToken, /** Spells page: fired once when the filtered list first has rows after a picker change. */ onSpellFeedFirstPaint, /** * When true, load events with parallel {@link client.fetchEvents} per subRequest instead of * {@link client.subscribeTimeline}. No live stream or `loadMore` timeline pagination; use for faux spells * (except Following). Refresh re-fetches. */ oneShotFetch = false }: { subRequests: TFeedSubRequest[] showKinds: number[] showKind1OPs?: boolean showKind1Replies?: boolean showKind1111?: boolean filterMutedNotes?: boolean hideReplies?: boolean hideUntrustedNotes?: boolean areAlgoRelays?: boolean pinnedEventIds?: string[] /** When true, use filter from subRequests as-is (kinds, limit) instead of showKinds. For spell feeds. */ useFilterAsIs?: boolean /** When provided and returns true, the event is omitted from the feed (in addition to built-in rules). */ extraShouldHideEvent?: (evt: Event) => boolean feedSubscriptionKey?: string preserveTimelineOnSubRequestsChange?: boolean /** When set (spells), max time to show the initial loading skeleton (ms). */ spellFetchTimeoutMs?: number spellFeedInstrumentToken?: number onSpellFeedFirstPaint?: (detail: { eventCount: number; firstEventId: string }) => void oneShotFetch?: boolean }, ref ) => { const { t } = useTranslation() const { startLogin, pubkey } = useNostr() const { isUserTrusted } = useUserTrust() const { mutePubkeySet } = useMuteList() const { hideContentMentioningMutedUsers } = useContentPolicy() const { isEventDeleted } = useDeletedEvent() const { zapReplyThreshold } = useZap() const [events, setEvents] = useState([]) const eventsRef = useRef([]) const [newEvents, setNewEvents] = useState([]) const [hasMore, setHasMore] = useState(true) const [loading, setLoading] = useState(true) const [timelineKey, setTimelineKey] = useState(undefined) const [refreshCount, setRefreshCount] = useState(0) const [showCount, setShowCount] = useState(SHOW_COUNT) const supportTouch = useMemo(() => isTouchDevice(), []) const bottomRef = useRef(null) const topRef = useRef(null) const spellFeedFirstPaintLoggedKeyRef = useRef('') const consecutiveEmptyRef = useRef(0) // Track consecutive empty results to prevent infinite retries const loadMoreTimeoutRef = useRef(null) // Throttle loadMore calls to prevent stuttering /** Batched profile + embed prefetch after timeline updates (avoids N×9s profile storms while relays stream). */ const timelinePrefetchDebounceRef = useRef | null>(null) const lastEventsForTimelinePrefetchRef = useRef([]) const [feedProfileBatch, setFeedProfileBatch] = useState<{ profiles: Map pending: Set version: number }>(() => ({ profiles: new Map(), pending: new Set(), version: 0 })) const feedProfileLoadedRef = useRef>(new Set()) const feedProfileBatchGenRef = useRef(0) const noteFeedProfileContextValue = useMemo( () => ({ profiles: feedProfileBatch.profiles, pendingPubkeys: feedProfileBatch.pending, version: feedProfileBatch.version }), [feedProfileBatch] ) // Memoize subRequests serialization to avoid expensive JSON.stringify on every render const subRequestsKey = useMemo(() => { return JSON.stringify( subRequests.map((req) => ({ urls: [...req.urls].map((u) => normalizeUrl(u) || u).filter(Boolean).sort(), filter: stableSpellFeedFilterKey(req.filter) })) ) }, [subRequests]) const timelineSubscriptionKey = feedSubscriptionKey ?? subRequestsKey const prevSubRequestsKeyForTimelineRef = useRef(null) /** Detect pull-to-refresh so preserve-mode feeds still clear; unrelated dep changes must not clear. */ const timelineEffectLastRefreshCountRef = useRef(refreshCount) useEffect(() => { feedProfileBatchGenRef.current += 1 feedProfileLoadedRef.current.clear() setFeedProfileBatch({ profiles: new Map(), pending: new Set(), version: 0 }) }, [timelineSubscriptionKey, refreshCount]) /** Pending pubkeys sync with rows so useFetchProfile skips per-note fetches before the debounced batch. */ useLayoutEffect(() => { const candidates = new Set() const addPk = (p: string | undefined) => { if (!p) return const t = p.trim() if (t.length === 64 && /^[0-9a-f]{64}$/i.test(t)) { candidates.add(t.toLowerCase()) } } for (const e of events) { addPk(e.pubkey) } for (const e of newEvents) { addPk(e.pubkey) } setFeedProfileBatch((prev) => { const pending = new Set(prev.pending) let changed = false for (const pk of candidates) { if (!prev.profiles.has(pk) && !pending.has(pk)) { pending.add(pk) changed = true } } if (!changed) return prev return { ...prev, pending, version: prev.version + 1 } }) }, [events, newEvents]) const subRequestsRef = useRef(subRequests) subRequestsRef.current = subRequests // Stable key for kind filter so subscription effect doesn't re-run on parent re-renders with same kinds // Use sorted array and JSON.stringify to create a stable key that only changes when content changes const showKindsKey = useMemo(() => { if (!showKinds || showKinds.length === 0) return '' return JSON.stringify([...showKinds].sort((a, b) => a - b)) }, [showKinds]) const shouldHideEvent = useCallback( (evt: Event) => { const pinnedEventHexIdSet = new Set() pinnedEventIds.forEach((id) => { try { const { type, data } = decode(id) if (type === 'nevent') { pinnedEventHexIdSet.add(data.id) } } catch { // ignore } }) if (pinnedEventHexIdSet.has(evt.id)) return true if (isEventDeleted(evt)) return true if (hideReplies && isReplyNoteEvent(evt)) return true if (hideUntrustedNotes && !isUserTrusted(evt.pubkey)) return true if (filterMutedNotes && mutePubkeySet.has(evt.pubkey)) return true if ( filterMutedNotes && hideContentMentioningMutedUsers && isMentioningMutedUsers(evt, mutePubkeySet) ) { return true } // Filter out expired events if (shouldFilterEvent(evt)) return true // Filter out zap receipts below the zap threshold (superzaps) if (evt.kind === ExtendedKind.ZAP_RECEIPT) { const zapInfo = getZapInfoFromEvent(evt) // Hide zap receipts if amount is missing, 0, or below the threshold if (!zapInfo || zapInfo.amount === undefined || zapInfo.amount === 0 || zapInfo.amount < zapReplyThreshold) { return true } } if (extraShouldHideEvent?.(evt)) return true return false }, [ hideReplies, hideUntrustedNotes, mutePubkeySet, pinnedEventIds, isEventDeleted, zapReplyThreshold, extraShouldHideEvent ] ) const shouldHideEventRef = useRef(shouldHideEvent) useEffect(() => { shouldHideEventRef.current = shouldHideEvent }, [shouldHideEvent]) const filteredEvents = useMemo(() => { const idSet = new Set() return events.slice(0, showCount).filter((evt) => { if (!showKinds.includes(evt.kind)) return false // Kind 1: show only OPs if showKind1OPs, only replies if showKind1Replies if (evt.kind === kinds.ShortTextNote) { const isReply = isReplyNoteEvent(evt) if (isReply && !showKind1Replies) return false if (!isReply && !showKind1OPs) return false } // Kind 1111 (comments): show only if showKind1111 if (evt.kind === ExtendedKind.COMMENT && !showKind1111) return false if (shouldHideEvent(evt)) return false const id = isReplaceableEvent(evt.kind) ? getReplaceableCoordinateFromEvent(evt) : evt.id if (idSet.has(id)) { return false } idSet.add(id) return true }) }, [events, showCount, shouldHideEvent, showKinds, showKind1OPs, showKind1Replies, showKind1111]) const filteredNewEvents = useMemo(() => { const idSet = new Set() return newEvents.filter((event: Event) => { if (!showKinds.includes(event.kind)) return false if (event.kind === kinds.ShortTextNote) { const isReply = isReplyNoteEvent(event) if (isReply && !showKind1Replies) return false if (!isReply && !showKind1OPs) return false } if (event.kind === ExtendedKind.COMMENT && !showKind1111) return false if (shouldHideEvent(event)) return false const id = isReplaceableEvent(event.kind) ? getReplaceableCoordinateFromEvent(event) : event.id if (idSet.has(id)) { return false } idSet.add(id) return true }) }, [newEvents, shouldHideEvent, showKinds, showKind1OPs, showKind1Replies, showKind1111]) useLayoutEffect(() => { if (!onSpellFeedFirstPaint || spellFeedInstrumentToken === undefined) return if (filteredEvents.length === 0) return const first = filteredEvents[0] if (!first) return const fpKey = `${spellFeedInstrumentToken}|${timelineSubscriptionKey ?? ''}` if (spellFeedFirstPaintLoggedKeyRef.current === fpKey) return spellFeedFirstPaintLoggedKeyRef.current = fpKey onSpellFeedFirstPaint({ eventCount: filteredEvents.length, firstEventId: first.id }) }, [ onSpellFeedFirstPaint, spellFeedInstrumentToken, timelineSubscriptionKey, filteredEvents.length, filteredEvents[0]?.id ]) useEffect(() => { const handle = window.setTimeout(() => { const gen = feedProfileBatchGenRef.current const candidates = new Set() const addPk = (p: string | undefined) => { if (p && p.length === 64 && /^[0-9a-f]{64}$/.test(p)) { candidates.add(p) } } for (const e of events) { addPk(e.pubkey) } for (const e of newEvents) { addPk(e.pubkey) } const need = [...candidates].filter((pk) => !feedProfileLoadedRef.current.has(pk)) if (need.length === 0) return need.forEach((pk) => feedProfileLoadedRef.current.add(pk)) setFeedProfileBatch((prev) => { const pending = new Set(prev.pending) let pendingChanged = false for (const pk of need) { if (!pending.has(pk)) { pending.add(pk) pendingChanged = true } } if (!pendingChanged) return prev return { ...prev, pending, version: prev.version + 1 } }) void (async () => { for (let i = 0; i < need.length; i += FEED_PROFILE_CHUNK) { if (gen !== feedProfileBatchGenRef.current) return const chunk = need.slice(i, i + FEED_PROFILE_CHUNK) try { const profiles = await client.fetchProfilesForPubkeys(chunk) if (gen !== feedProfileBatchGenRef.current) return setFeedProfileBatch((prev) => { const next = new Map(prev.profiles) const pend = new Set(prev.pending) for (const p of profiles) { next.set(p.pubkey, p) pend.delete(p.pubkey) } for (const pk of chunk) { pend.delete(pk) if (!next.has(pk)) { next.set(pk, { pubkey: pk, npub: pubkeyToNpub(pk) ?? '', username: formatPubkey(pk) }) } } return { profiles: next, pending: pend, version: prev.version + 1 } }) } catch { chunk.forEach((pk) => feedProfileLoadedRef.current.delete(pk)) if (gen !== feedProfileBatchGenRef.current) return setFeedProfileBatch((prev) => { const pend = new Set(prev.pending) chunk.forEach((pk) => pend.delete(pk)) return { ...prev, pending: pend, version: prev.version + 1 } }) } } })() }, FEED_PROFILE_BATCH_DEBOUNCE_MS) return () => window.clearTimeout(handle) }, [events, newEvents]) const scrollToTop = useCallback((behavior: ScrollBehavior = 'instant') => { setTimeout(() => { topRef.current?.scrollIntoView({ behavior, block: 'start' }) }, 20) }, []) const refresh = useCallback(() => { scrollToTop() setTimeout(() => { setRefreshCount((count) => count + 1) }, 500) }, [scrollToTop]) useImperativeHandle(ref, () => ({ scrollToTop, refresh }), [scrollToTop, refresh]) useEffect(() => { const currentSubRequests = subRequestsRef.current if (!currentSubRequests.length) { setLoading(false) setEvents([]) // Return a no-op closer function to satisfy the cleanup function return () => {} } const prevSubKey = prevSubRequestsKeyForTimelineRef.current const userPulledRefresh = refreshCount !== timelineEffectLastRefreshCountRef.current if (userPulledRefresh) { timelineEffectLastRefreshCountRef.current = refreshCount } const keepExistingTimelineEvents = preserveTimelineOnSubRequestsChange && !userPulledRefresh && (prevSubKey === subRequestsKey || isRelayUrlStrictSupersetIdentityKey(prevSubKey, subRequestsKey)) prevSubRequestsKeyForTimelineRef.current = subRequestsKey /** False after cleanup so stale timeline callbacks cannot overwrite state after switching feeds (e.g. Spells discussions → notifications). */ let effectActive = true async function init() { // Re-subscribe with rows visible (e.g. relay URL expansion): don't flash global loading / skeleton. const keepRowsVisible = preserveTimelineOnSubRequestsChange && keepExistingTimelineEvents && eventsRef.current.length > 0 if (!keepRowsVisible) { setLoading(true) } if (!keepExistingTimelineEvents) { setEvents([]) setNewEvents([]) } setHasMore(true) consecutiveEmptyRef.current = 0 // Reset counter on refresh const mappedSubRequests = subRequestsRef.current.map(({ urls, filter }) => { // CRITICAL: Always ensure filter has kinds - relays require this to return events const defaultKinds = showKinds.length > 0 ? showKinds : [kinds.ShortTextNote] const finalFilter = useFilterAsIs ? { ...filter, // If filter doesn't have kinds, add them (required for relay queries) kinds: filter.kinds && filter.kinds.length > 0 ? filter.kinds : defaultKinds, limit: filter.limit ?? (areAlgoRelays ? ALGO_LIMIT : LIMIT) } : { ...filter, // If showKinds is empty, default to kind 1 (ShortTextNote) only kinds: defaultKinds, limit: areAlgoRelays ? ALGO_LIMIT : LIMIT } // CRITICAL: Validate filter has kinds before subscribing if (!finalFilter.kinds || finalFilter.kinds.length === 0) { finalFilter.kinds = [kinds.ShortTextNote] } return { urls, filter: finalFilter } }) // CRITICAL: Validate all filters have kinds before subscribing const invalidFilters = mappedSubRequests.filter(({ filter }) => !filter.kinds || filter.kinds.length === 0) if (invalidFilters.length > 0) { // Don't subscribe with invalid filters - this would return no events setLoading(false) setEvents([]) // Return a no-op closer function to satisfy the cleanup function return () => {} } if (oneShotFetch) { if (!keepExistingTimelineEvents) { setEvents([]) setNewEvents([]) } setHasMore(false) try { const batches = await Promise.all( mappedSubRequests.map(({ urls, filter }) => client.fetchEvents(urls, filter, { // Was `false`, which disabled feed grace and forced a wait for every relay EOSE (very slow). firstRelayResultGraceMs: FIRST_RELAY_RESULT_GRACE_MS, globalTimeout: 14_000, eoseTimeout: 2_000, cache: true }) ) ) if (!effectActive) return undefined const byId = new Map() for (const ev of batches.flat()) { const prev = byId.get(ev.id) if (!prev || ev.created_at > prev.created_at) { byId.set(ev.id, ev) } } const merged = [...byId.values()] .sort((a, b) => b.created_at - a.created_at) .slice(0, ONE_SHOT_MERGED_CAP) setEvents(merged) lastEventsForTimelinePrefetchRef.current = merged } catch { if (effectActive) setEvents([]) } finally { if (effectActive) { setLoading(false) setHasMore(false) setTimelineKey(undefined) } } return undefined } const totalRelayUrls = mappedSubRequests.reduce((n, r) => n + r.urls.length, 0) // Explore-style feeds merge many read relays; subscribeTimeline awaits every ensureRelay — 5s often loses the race. const subscribeSetupRaceMs = totalRelayUrls > 24 ? 30_000 : 5000 let closer: (() => void) | undefined let timelineKey: string | undefined let timelineSubscribePromise: | Promise<{ closer: () => void; timelineKey: string }> | undefined try { // Opening many relay subs can exceed 2s on spell feeds; a short race // rejects, the catch closes the late subscription, and the list stays empty after refresh. const timeoutPromise = new Promise((_, reject) => { setTimeout(() => { reject(new Error(`subscribeTimeline timeout after ${subscribeSetupRaceMs}ms`)) }, subscribeSetupRaceMs) }) const eventCap = areAlgoRelays ? ALGO_LIMIT : LIMIT timelineSubscribePromise = client.subscribeTimeline( mappedSubRequests, { onEvents: (batch: Event[], eosed: boolean) => { if (!effectActive) return if (batch.length > 0) { if (preserveTimelineOnSubRequestsChange) { setEvents((prev) => { const next = mergeEventBatchesById(prev, batch, eventCap) lastEventsForTimelinePrefetchRef.current = next return next }) } else { setEvents(batch) lastEventsForTimelinePrefetchRef.current = batch } // Do not wait for full EOSE across many relays — otherwise loading/skeleton stays up for 10–30s+ setLoading(false) // Defer profile + embed prefetch: streaming timelines fire onEvents often; starting // fetchProfilesForPubkeys on every update spams relays (multi-second each) and cancels hooks. if (timelinePrefetchDebounceRef.current) { clearTimeout(timelinePrefetchDebounceRef.current) } timelinePrefetchDebounceRef.current = setTimeout(() => { timelinePrefetchDebounceRef.current = null if (!effectActive) return const evs = lastEventsForTimelinePrefetchRef.current if (evs.length === 0) return const { hexIds, nip19Pointers } = mergePrefetchTargetsFromEvents(evs.slice(0, 50)) const hexIdsToFetch = hexIds.filter((id) => !prefetchedEventIdsRef.current.has(id)) const nip19ToFetch = nip19Pointers.filter((p) => !prefetchedEventIdsRef.current.has(p)) if (hexIdsToFetch.length > 0 || nip19ToFetch.length > 0) { hexIdsToFetch.forEach((id) => prefetchedEventIdsRef.current.add(id)) nip19ToFetch.forEach((p) => prefetchedEventIdsRef.current.add(p)) const run = async () => { try { await client.prefetchHexEventIds(hexIdsToFetch) await Promise.all(nip19ToFetch.map((p) => client.fetchEvent(p))) } catch { hexIdsToFetch.forEach((id) => prefetchedEventIdsRef.current.delete(id)) nip19ToFetch.forEach((p) => prefetchedEventIdsRef.current.delete(p)) } } void run() } }, 450) } else if (eosed) { if (!preserveTimelineOnSubRequestsChange) { setEvents([]) } setLoading(false) } if (areAlgoRelays) { // Algorithm feeds typically return all results at once setHasMore(false) } else if (eosed) { setLoading(false) // CRITICAL FIX: For non-algo feeds, always assume there might be more events // The initial load might only return a few events due to filtering or relay limits // We should still try to load more on scroll - the loadMore logic will handle stopping // Only set to false if we explicitly know there are no more events (handled in loadMore) // If we got a full limit of events, there's likely more available if (batch.length >= (areAlgoRelays ? ALGO_LIMIT : LIMIT)) { setHasMore(true) } else { // Even with fewer events, there might be more (filtering, slow relays, etc.) // Let loadMore determine if we've reached the end setHasMore(true) } } }, onNew: (event: Event) => { if (!effectActive) return if (!useFilterAsIs && !showKinds.includes(event.kind)) return if (event.kind === kinds.ShortTextNote) { const isReply = isReplyNoteEvent(event) if (isReply && !showKind1Replies) return if (!isReply && !showKind1OPs) return } if (event.kind === ExtendedKind.COMMENT && !showKind1111) return if (shouldHideEventRef.current(event)) return if (pubkey && event.pubkey === pubkey) { // If the new event is from the current user, insert it directly into the feed setEvents((oldEvents) => oldEvents.some((e) => e.id === event.id) ? oldEvents : [event, ...oldEvents] ) } else { // Otherwise, buffer it and show the New Notes button setNewEvents((oldEvents) => [event, ...oldEvents].sort((a, b) => b.created_at - a.created_at) ) } }, }, { startLogin, needSort: !areAlgoRelays, firstRelayResultGraceMs: FIRST_RELAY_RESULT_GRACE_MS } ) const result = await Promise.race([timelineSubscribePromise, timeoutPromise]) if (!effectActive) { result.closer() return () => {} } closer = result.closer timelineKey = result.timelineKey setTimelineKey(timelineKey) return closer } catch (_error) { setLoading(false) // Race timeout or subscribe failure: if the timeline promise later resolves, close or subs leak (relay slots + stale setEvents). if (timelineSubscribePromise) { void timelineSubscribePromise .then((r) => { r.closer() }) .catch(() => {}) } return () => {} } } const promise = init() return () => { effectActive = false if (timelinePrefetchDebounceRef.current) { clearTimeout(timelinePrefetchDebounceRef.current) timelinePrefetchDebounceRef.current = null } promise.then((closer) => closer?.()) } }, [ timelineSubscriptionKey, subRequestsKey, preserveTimelineOnSubRequestsChange, refreshCount, showKindsKey, showKind1OPs, showKind1Replies, showKind1111, useFilterAsIs, areAlgoRelays, spellFetchTimeoutMs, oneShotFetch ]) useEffect(() => { eventsRef.current = events }, [events]) useEffect(() => { if (!subRequestsRef.current.length) return let cancelled = false const timer = window.setTimeout(() => { if (cancelled) return setLoading((prev) => (prev ? false : prev)) // hasMore defaults true; if timeline never sends eosed (slow/hung relays), we would keep a // bottom skeleton forever while loading is false — unblock empty state / reload. if (eventsRef.current.length === 0) { setHasMore(false) } }, 15_000) return () => { cancelled = true clearTimeout(timer) } }, [timelineSubscriptionKey, refreshCount]) /** Spells: drop loading skeleton quickly so rows (or empty + reload) appear while REQ continues. */ useEffect(() => { if (spellFetchTimeoutMs == null || spellFetchTimeoutMs <= 0) return if (!subRequestsRef.current.length) return let cancelled = false const id = window.setTimeout(() => { if (cancelled) return setLoading(false) }, spellFetchTimeoutMs) return () => { cancelled = true clearTimeout(id) } }, [timelineSubscriptionKey, refreshCount, spellFetchTimeoutMs]) // Use refs to avoid dependency issues and ensure latest values in async callbacks const showCountRef = useRef(showCount) const loadingRef = useRef(loading) const hasMoreRef = useRef(hasMore) const timelineKeyRef = useRef(timelineKey) useEffect(() => { showCountRef.current = showCount }, [showCount]) useEffect(() => { loadingRef.current = loading }, [loading]) useEffect(() => { hasMoreRef.current = hasMore }, [hasMore]) useEffect(() => { timelineKeyRef.current = timelineKey }, [timelineKey]) useEffect(() => { const options: IntersectionObserverInit = { root: null, // Trigger when user is 400px from the bottom so we start loading before they reach the end rootMargin: '0px 0px 400px 0px', threshold: 0 } const loadMore = async (): Promise => { const currentEvents = eventsRef.current const currentShowCount = showCountRef.current const currentLoading = loadingRef.current const currentHasMore = hasMoreRef.current const currentTimelineKey = timelineKeyRef.current // CRITICAL: Throttle loadMore calls to prevent stuttering during rapid scrolling if (loadMoreTimeoutRef.current) { return // Already scheduled, skip } // Show more events immediately if we have them cached if (currentShowCount < currentEvents.length) { // Show more aggressively: increase by SHOW_COUNT, but also check if we should show even more const remaining = currentEvents.length - currentShowCount const increment = Math.min(SHOW_COUNT * 2, remaining) // Show up to 2x SHOW_COUNT if available setShowCount((prev) => prev + increment) // Only preload more if we have plenty cached (more than 3/4 of LIMIT) // BUT: Always try to load more if we have very few events (might be due to filtering) if (currentEvents.length - currentShowCount > LIMIT * 0.75 && currentEvents.length >= 50) { return } // If we have very few events, always try to load more (might be aggressive filtering) if (currentEvents.length < 50) { // Continue to loadMore below even if we have cached events // This ensures we keep loading when filtering is aggressive } } if (!currentTimelineKey || currentLoading || !currentHasMore) return // Schedule loadMore with a small delay to throttle rapid calls loadMoreTimeoutRef.current = setTimeout(async () => { loadMoreTimeoutRef.current = null const latestEvents = eventsRef.current const latestTimelineKey = timelineKeyRef.current const latestLoading = loadingRef.current const latestHasMore = hasMoreRef.current if (!latestTimelineKey || latestLoading || !latestHasMore) return setLoading(true) let newEvents: Event[] = [] try { const until = latestEvents.length ? latestEvents[latestEvents.length - 1].created_at - 1 : dayjs().unix() newEvents = await client.loadMoreTimeline( latestTimelineKey, until, LIMIT ) // CRITICAL FIX: Be extremely conservative about stopping the feed // Only stop if we're absolutely certain there are no more events if (newEvents.length === 0) { // Check if timeline has more cached refs that we haven't loaded yet const hasMoreCached = client.hasMoreTimelineEvents?.(latestTimelineKey, until) ?? false if (hasMoreCached) { // There are more cached events, keep hasMore true and try again setLoading(false) // Retry after a short delay to allow IndexedDB to catch up setTimeout(() => { if (hasMoreRef.current && !loadingRef.current) { loadMore() } }, 300) return } // No cached events and network returned empty // Be VERY patient - don't stop too early, especially when we have few events // This prevents stopping due to temporary relay issues or slow relays consecutiveEmptyRef.current += 1 // CRITICAL FIX: Only stop if we have MANY consecutive empty results AND we have a reasonable number of events // This ensures we don't stop prematurely when relays are slow or filtering is aggressive // If we have very few events (< 50), keep trying longer in case filtering is aggressive const eventCount = latestEvents.length const shouldStop = consecutiveEmptyRef.current >= (eventCount < 50 ? 30 : 15) if (shouldStop) { // After many consecutive empty results, assume we've reached the end setHasMore(false) } // Otherwise, keep hasMore true to allow retry on next scroll // This ensures the feed continues trying even if relays are slow setLoading(false) return } // Reset consecutive empty counter on success consecutiveEmptyRef.current = 0 setEvents((oldEvents) => [...oldEvents, ...newEvents]) // After appending, the bottom sentinel may have moved below the fold. Re-check after // paint: if it's still in/near view, trigger loadMore again so user doesn't have to scroll. setTimeout(() => { const bottomEl = bottomRef.current if (bottomEl && hasMoreRef.current && !loadingRef.current) { const rect = bottomEl.getBoundingClientRect() if (rect.top < window.innerHeight + 200) { loadMore() } } }, 150) // NEVER automatically set hasMore to false based on result count // Only stop when we get consecutive empty results // This ensures the feed continues loading even with partial results // CRITICAL: Prefetch profiles for newly loaded events (optimized to reduce stuttering) // Only prefetch if we're not currently loading to avoid blocking scroll if (newEvents.length > 0 && !loadingRef.current) { // Use requestIdleCallback if available, otherwise setTimeout with longer delay const schedulePrefetch = (callback: () => void) => { if (typeof requestIdleCallback !== 'undefined') { requestIdleCallback(callback, { timeout: 500 }) } else { setTimeout(callback, 300) } } schedulePrefetch(() => { const { hexIds, nip19Pointers } = mergePrefetchTargetsFromEvents(newEvents.slice(0, 30)) const hexIdsToFetch = hexIds.filter((id) => !prefetchedEventIdsRef.current.has(id)) const nip19ToFetch = nip19Pointers.filter((p) => !prefetchedEventIdsRef.current.has(p)) if (hexIdsToFetch.length === 0 && nip19ToFetch.length === 0) return hexIdsToFetch.forEach((id) => prefetchedEventIdsRef.current.add(id)) nip19ToFetch.forEach((p) => prefetchedEventIdsRef.current.add(p)) const run = async () => { try { await client.prefetchHexEventIds(hexIdsToFetch) await Promise.all(nip19ToFetch.map((p) => client.fetchEvent(p))) } catch { hexIdsToFetch.forEach((id) => prefetchedEventIdsRef.current.delete(id)) nip19ToFetch.forEach((p) => prefetchedEventIdsRef.current.delete(p)) } } void run() }) } } catch (_error) { // On error, don't set hasMore to false - might be temporary network issue consecutiveEmptyRef.current += 1 // Only stop after MANY consecutive errors - be very patient with network issues // This prevents stopping when relays are temporarily down or slow if (consecutiveEmptyRef.current >= 25) { // Increased from 15 to 25 to be even more patient with network issues setHasMore(false) } } finally { setLoading(false) } }, 50) // Reduced delay from 100ms to 50ms for more responsive scrolling } const observerInstance = new IntersectionObserver((entries) => { if (entries[0].isIntersecting && hasMoreRef.current && !loadingRef.current) { // Throttle: only trigger if not already loading and not already scheduled loadMore() } }, options) const currentBottomRef = bottomRef.current if (currentBottomRef) { observerInstance.observe(currentBottomRef) } return () => { if (observerInstance && currentBottomRef) { observerInstance.unobserve(currentBottomRef) } // Clean up timeout on unmount if (loadMoreTimeoutRef.current) { clearTimeout(loadMoreTimeoutRef.current) loadMoreTimeoutRef.current = null } } // Dependencies are handled via refs to avoid stale closures in async callbacks // eslint-disable-next-line react-hooks/exhaustive-deps }, []) // CRITICAL: Prefetch embedded events (referenced in e tags, a tags, and content) // This ensures embedded events are ready before user scrolls to them const prefetchedEventIdsRef = useRef>(new Set()) const prefetchEmbeddedEventsTimeoutRef = useRef(null) const mergePrefetchTargetsFromEvents = useCallback((evts: Event[]) => { const hex = new Set() const nip19 = new Set() for (const e of evts) { const t = collectEmbeddedEventPrefetchTargets(e) t.hexIds.forEach((id) => hex.add(id)) t.nip19Pointers.forEach((p) => nip19.add(p)) } return { hexIds: Array.from(hex), nip19Pointers: Array.from(nip19) } }, []) // CRITICAL: Prefetch embedded events for visible events useEffect(() => { // Throttle embedded event prefetching to reduce frequency during rapid scrolling // Clear any existing timeout if (prefetchEmbeddedEventsTimeoutRef.current) { clearTimeout(prefetchEmbeddedEventsTimeoutRef.current) } // Debounce embedded event prefetching by 400ms to reduce frequency during rapid scrolling prefetchEmbeddedEventsTimeoutRef.current = setTimeout(() => { const visibleTargets = mergePrefetchTargetsFromEvents(filteredEvents.slice(0, 40)) const upcomingTargets = mergePrefetchTargetsFromEvents(events.slice(0, 80)) const hexIds = Array.from( new Set([...visibleTargets.hexIds, ...upcomingTargets.hexIds]) ) const nip19Pointers = Array.from( new Set([...visibleTargets.nip19Pointers, ...upcomingTargets.nip19Pointers]) ) const hexIdsToFetch = hexIds.filter((id) => !prefetchedEventIdsRef.current.has(id)) const nip19ToFetch = nip19Pointers.filter((p) => !prefetchedEventIdsRef.current.has(p)) if (hexIdsToFetch.length === 0 && nip19ToFetch.length === 0) return hexIdsToFetch.forEach((id) => prefetchedEventIdsRef.current.add(id)) nip19ToFetch.forEach((p) => prefetchedEventIdsRef.current.add(p)) const scheduleFetch = (callback: () => void) => { if (typeof requestIdleCallback !== 'undefined') { requestIdleCallback(callback, { timeout: 500 }) } else { setTimeout(callback, 0) } } scheduleFetch(() => { const run = async () => { try { await client.prefetchHexEventIds(hexIdsToFetch) await Promise.all(nip19ToFetch.map((p) => client.fetchEvent(p))) } catch { hexIdsToFetch.forEach((id) => prefetchedEventIdsRef.current.delete(id)) nip19ToFetch.forEach((p) => prefetchedEventIdsRef.current.delete(p)) } } void run() }) }, 400) // Debounce by 400ms to reduce frequency during rapid scrolling return () => { if (prefetchEmbeddedEventsTimeoutRef.current) { clearTimeout(prefetchEmbeddedEventsTimeoutRef.current) prefetchEmbeddedEventsTimeoutRef.current = null } } }, [filteredEvents, events, mergePrefetchTargetsFromEvents]) // Also prefetch when loading more events (scrolling down) // Throttled to reduce frequency during rapid scrolling const prefetchNewEventsTimeoutRef = useRef(null) useEffect(() => { if (loading || !hasMore) return // Clear any existing timeout if (prefetchNewEventsTimeoutRef.current) { clearTimeout(prefetchNewEventsTimeoutRef.current) } // Debounce embedded-event prefetch for newly revealed rows (profiles use NoteFeed batcher above) prefetchNewEventsTimeoutRef.current = setTimeout(() => { const { hexIds, nip19Pointers } = mergePrefetchTargetsFromEvents( events.slice(showCount, showCount + 50) ) const hexIdsToFetch = hexIds.filter((id) => !prefetchedEventIdsRef.current.has(id)) const nip19ToFetch = nip19Pointers.filter((p) => !prefetchedEventIdsRef.current.has(p)) if (hexIdsToFetch.length === 0 && nip19ToFetch.length === 0) return hexIdsToFetch.forEach((id) => prefetchedEventIdsRef.current.add(id)) nip19ToFetch.forEach((p) => prefetchedEventIdsRef.current.add(p)) const scheduleFetch = (callback: () => void) => { if (typeof requestIdleCallback !== 'undefined') { requestIdleCallback(callback, { timeout: 500 }) } else { setTimeout(callback, 0) } } scheduleFetch(() => { const run = async () => { try { await client.prefetchHexEventIds(hexIdsToFetch) await Promise.all(nip19ToFetch.map((p) => client.fetchEvent(p))) } catch { hexIdsToFetch.forEach((id) => prefetchedEventIdsRef.current.delete(id)) nip19ToFetch.forEach((p) => prefetchedEventIdsRef.current.delete(p)) } } void run() }) }, 400) // Debounce by 400ms to reduce frequency during rapid scrolling return () => { if (prefetchNewEventsTimeoutRef.current) { clearTimeout(prefetchNewEventsTimeoutRef.current) prefetchNewEventsTimeoutRef.current = null } } }, [events.length, showCount, loading, hasMore, mergePrefetchTargetsFromEvents]) const showNewEvents = () => { setEvents((oldEvents) => [...newEvents, ...oldEvents]) setNewEvents([]) setTimeout(() => { scrollToTop('smooth') }, 0) } const list = (
{filteredEvents.map((event) => ( ))} {events.length === 0 && loading ? (
) : events.length > 0 && (hasMore || loading) ? (
) : events.length > 0 ? (
{t('no more notes')}
) : (spellFetchTimeoutMs != null && spellFetchTimeoutMs > 0) || oneShotFetch ? (
{t('No posts loaded for this feed. Try refreshing.')}
) : (
)}
) return (
{supportTouch ? ( { refresh() await new Promise((resolve) => setTimeout(resolve, 1000)) }} pullingContent="" > {list} ) : ( list )}
{filteredNewEvents.length > 0 && ( )}
) } ) NoteList.displayName = 'NoteList' export default NoteList export type TNoteListRef = { scrollToTop: (behavior?: ScrollBehavior) => void refresh: () => void }