Add Neo4j integration tests and query rate-limiting logic
Some checks failed
Go / build-and-release (push) Has been cancelled

Introduce comprehensive integration tests for Neo4j bug fixes covering batching, event relationships, and processing logic. Add rate-limiting to Neo4j queries using semaphores and retry policies to prevent authentication rate limiting and connection exhaustion, ensuring system stability under load.
This commit is contained in:
2025-12-07 00:07:25 +00:00
parent 8ea91e39d8
commit 95271cbc81
21 changed files with 1925 additions and 1594 deletions

View File

@@ -16,12 +16,19 @@ func parseInt64(s string) (int64, error) {
return strconv.ParseInt(s, 10, 64)
}
// tagBatchSize is the maximum number of tags to process in a single transaction
// This prevents Neo4j stack overflow errors with events that have thousands of tags
const tagBatchSize = 500
// SaveEvent stores a Nostr event in the Neo4j database.
// It creates event nodes and relationships for authors, tags, and references.
// This method leverages Neo4j's graph capabilities to model Nostr's social graph naturally.
//
// For social graph events (kinds 0, 3, 1984, 10000), it additionally processes them
// to maintain NostrUser nodes and FOLLOWS/MUTES/REPORTS relationships with event traceability.
//
// To prevent Neo4j stack overflow errors with events containing thousands of tags,
// tags are processed in batches using UNWIND instead of generating inline Cypher.
func (n *N) SaveEvent(c context.Context, ev *event.E) (exists bool, err error) {
eventID := hex.Enc(ev.ID[:])
@@ -42,7 +49,7 @@ func (n *N) SaveEvent(c context.Context, ev *event.E) (exists bool, err error) {
if ev.Kind == 0 || ev.Kind == 3 || ev.Kind == 1984 || ev.Kind == 10000 {
processor := NewSocialEventProcessor(n)
if err := processor.ProcessSocialEvent(c, ev); err != nil {
n.Logger.Warningf("failed to reprocess social event %s: %v", eventID[:16], err)
n.Logger.Warningf("failed to reprocess social event %s: %v", safePrefix(eventID, 16), err)
// Don't fail the whole save, social processing is supplementary
}
}
@@ -55,14 +62,20 @@ func (n *N) SaveEvent(c context.Context, ev *event.E) (exists bool, err error) {
return false, fmt.Errorf("failed to get serial number: %w", err)
}
// Build and execute Cypher query to create event with all relationships
// This creates Event and Author nodes for NIP-01 query support
cypher, params := n.buildEventCreationCypher(ev, serial)
// Step 1: Create base event with author (small, fixed-size query)
cypher, params := n.buildBaseEventCypher(ev, serial)
if _, err = n.ExecuteWrite(c, cypher, params); err != nil {
return false, fmt.Errorf("failed to save event: %w", err)
}
// Step 2: Process tags in batches to avoid stack overflow
if ev.Tags != nil {
if err := n.addTagsInBatches(c, eventID, ev); err != nil {
// Log but don't fail - base event is saved, tags are supplementary for queries
n.Logger.Errorf("failed to add tags for event %s: %v", safePrefix(eventID, 16), err)
}
}
// Process social graph events (kinds 0, 3, 1984, 10000)
// This creates NostrUser nodes and social relationships (FOLLOWS, MUTES, REPORTS)
// with event traceability for diff-based updates
@@ -72,7 +85,7 @@ func (n *N) SaveEvent(c context.Context, ev *event.E) (exists bool, err error) {
// Log error but don't fail the whole save
// NIP-01 queries will still work even if social processing fails
n.Logger.Errorf("failed to process social event kind %d, event %s: %v",
ev.Kind, eventID[:16], err)
ev.Kind, safePrefix(eventID, 16), err)
// Consider: should we fail here or continue?
// For now, continue - social graph is supplementary to base relay
}
@@ -81,13 +94,20 @@ func (n *N) SaveEvent(c context.Context, ev *event.E) (exists bool, err error) {
return false, nil
}
// buildEventCreationCypher constructs a Cypher query to create an event node with all relationships
// This is a single atomic operation that creates:
// safePrefix returns up to n characters from a string, handling short strings gracefully
func safePrefix(s string, n int) string {
if len(s) <= n {
return s
}
return s[:n]
}
// buildBaseEventCypher constructs a Cypher query to create just the base event node and author.
// Tags are added separately in batches to prevent stack overflow with large tag sets.
// This creates:
// - Event node with all properties
// - NostrUser node and AUTHORED_BY relationship (unified author + WoT node)
// - Tag nodes and TAGGED_WITH relationships
// - Reference relationships (REFERENCES for 'e' tags, MENTIONS for 'p' tags)
func (n *N) buildEventCreationCypher(ev *event.E, serial uint64) (string, map[string]any) {
func (n *N) buildBaseEventCypher(ev *event.E, serial uint64) (string, map[string]any) {
params := make(map[string]any)
// Event properties
@@ -123,7 +143,7 @@ func (n *N) buildEventCreationCypher(ev *event.E, serial uint64) (string, map[st
}
params["tags"] = string(tagsJSON)
// Start building the Cypher query
// Build Cypher query - just event + author, no tags (tags added in batches)
// Use MERGE to ensure idempotency for NostrUser nodes
// NostrUser serves both NIP-01 author tracking and WoT social graph
cypher := `
@@ -146,143 +166,180 @@ CREATE (e:Event {
// Link event to author
CREATE (e)-[:AUTHORED_BY]->(a)
`
// Process tags to create relationships
// Different tag types create different relationship patterns
tagNodeIndex := 0
eTagIndex := 0
pTagIndex := 0
// Track if we need to add WITH clause before OPTIONAL MATCH
// This is required because Cypher doesn't allow MATCH after CREATE without WITH
needsWithClause := true
// Collect all e-tags, p-tags, and other tags first so we can generate proper Cypher
// Neo4j requires WITH clauses between certain clause types (FOREACH -> MATCH/MERGE)
type tagInfo struct {
tagType string
value string
}
var eTags, pTags, otherTags []tagInfo
// Only process tags if they exist
if ev.Tags != nil {
for _, tagItem := range *ev.Tags {
if len(tagItem.T) < 2 {
continue
}
tagType := string(tagItem.T[0])
switch tagType {
case "e": // Event reference
tagValue := ExtractETagValue(tagItem)
if tagValue != "" {
eTags = append(eTags, tagInfo{"e", tagValue})
}
case "p": // Pubkey mention
tagValue := ExtractPTagValue(tagItem)
if tagValue != "" {
pTags = append(pTags, tagInfo{"p", tagValue})
}
default: // Other tags
tagValue := string(tagItem.T[1])
otherTags = append(otherTags, tagInfo{tagType, tagValue})
}
}
}
// Generate Cypher for e-tags (OPTIONAL MATCH + FOREACH pattern)
// These need WITH clause before first one, and WITH after all FOREACHes
for i, tag := range eTags {
paramName := fmt.Sprintf("eTag_%d", eTagIndex)
params[paramName] = tag.value
// Add WITH clause before first OPTIONAL MATCH only
if needsWithClause {
cypher += `
// Carry forward event and author nodes for tag processing
WITH e, a
`
needsWithClause = false
}
cypher += fmt.Sprintf(`
// Reference to event (e-tag)
OPTIONAL MATCH (ref%d:Event {id: $%s})
FOREACH (ignoreMe IN CASE WHEN ref%d IS NOT NULL THEN [1] ELSE [] END |
CREATE (e)-[:REFERENCES]->(ref%d)
)
`, eTagIndex, paramName, eTagIndex, eTagIndex)
eTagIndex++
// After the last e-tag FOREACH, add WITH clause if there are p-tags or other tags
if i == len(eTags)-1 && (len(pTags) > 0 || len(otherTags) > 0) {
cypher += `
// Required WITH after FOREACH before MERGE/MATCH
WITH e, a
`
}
}
// Generate Cypher for p-tags (MERGE pattern)
for _, tag := range pTags {
paramName := fmt.Sprintf("pTag_%d", pTagIndex)
params[paramName] = tag.value
// If no e-tags were processed, we still need the initial WITH
if needsWithClause {
cypher += `
// Carry forward event and author nodes for tag processing
WITH e, a
`
needsWithClause = false
}
cypher += fmt.Sprintf(`
// Mention of NostrUser (p-tag)
MERGE (mentioned%d:NostrUser {pubkey: $%s})
ON CREATE SET mentioned%d.created_at = timestamp()
CREATE (e)-[:MENTIONS]->(mentioned%d)
`, pTagIndex, paramName, pTagIndex, pTagIndex)
pTagIndex++
}
// Generate Cypher for other tags (MERGE pattern)
for _, tag := range otherTags {
typeParam := fmt.Sprintf("tagType_%d", tagNodeIndex)
valueParam := fmt.Sprintf("tagValue_%d", tagNodeIndex)
params[typeParam] = tag.tagType
params[valueParam] = tag.value
// If no e-tags or p-tags were processed, we still need the initial WITH
if needsWithClause {
cypher += `
// Carry forward event and author nodes for tag processing
WITH e, a
`
needsWithClause = false
}
cypher += fmt.Sprintf(`
// Generic tag relationship
MERGE (tag%d:Tag {type: $%s, value: $%s})
CREATE (e)-[:TAGGED_WITH]->(tag%d)
`, tagNodeIndex, typeParam, valueParam, tagNodeIndex)
tagNodeIndex++
}
// Return the created event
cypher += `
RETURN e.id AS id`
return cypher, params
}
// tagTypeValue represents a generic tag with type and value for batch processing
type tagTypeValue struct {
Type string
Value string
}
// addTagsInBatches processes event tags in batches using UNWIND to prevent Neo4j stack overflow.
// This handles e-tags (event references), p-tags (pubkey mentions), and other tags separately.
func (n *N) addTagsInBatches(c context.Context, eventID string, ev *event.E) error {
if ev.Tags == nil {
return nil
}
// Collect tags by type
var eTags, pTags []string
var otherTags []tagTypeValue
for _, tagItem := range *ev.Tags {
if len(tagItem.T) < 2 {
continue
}
tagType := string(tagItem.T[0])
switch tagType {
case "e": // Event reference
tagValue := ExtractETagValue(tagItem)
if tagValue != "" {
eTags = append(eTags, tagValue)
}
case "p": // Pubkey mention
tagValue := ExtractPTagValue(tagItem)
if tagValue != "" {
pTags = append(pTags, tagValue)
}
default: // Other tags
tagValue := string(tagItem.T[1])
otherTags = append(otherTags, tagTypeValue{Type: tagType, Value: tagValue})
}
}
// Add p-tags in batches (creates MENTIONS relationships)
if len(pTags) > 0 {
if err := n.addPTagsInBatches(c, eventID, pTags); err != nil {
return fmt.Errorf("failed to add p-tags: %w", err)
}
}
// Add e-tags in batches (creates REFERENCES relationships)
if len(eTags) > 0 {
if err := n.addETagsInBatches(c, eventID, eTags); err != nil {
return fmt.Errorf("failed to add e-tags: %w", err)
}
}
// Add other tags in batches (creates TAGGED_WITH relationships)
if len(otherTags) > 0 {
if err := n.addOtherTagsInBatches(c, eventID, otherTags); err != nil {
return fmt.Errorf("failed to add other tags: %w", err)
}
}
return nil
}
// addPTagsInBatches adds p-tag (pubkey mention) relationships using UNWIND for efficiency.
// Creates NostrUser nodes for mentioned pubkeys and MENTIONS relationships.
func (n *N) addPTagsInBatches(c context.Context, eventID string, pTags []string) error {
// Process in batches to avoid memory issues
for i := 0; i < len(pTags); i += tagBatchSize {
end := i + tagBatchSize
if end > len(pTags) {
end = len(pTags)
}
batch := pTags[i:end]
// Use UNWIND to process multiple p-tags in a single query
cypher := `
MATCH (e:Event {id: $eventId})
UNWIND $pubkeys AS pubkey
MERGE (u:NostrUser {pubkey: pubkey})
ON CREATE SET u.created_at = timestamp()
CREATE (e)-[:MENTIONS]->(u)`
params := map[string]any{
"eventId": eventID,
"pubkeys": batch,
}
if _, err := n.ExecuteWrite(c, cypher, params); err != nil {
return fmt.Errorf("batch %d-%d: %w", i, end, err)
}
}
return nil
}
// addETagsInBatches adds e-tag (event reference) relationships using UNWIND for efficiency.
// Only creates REFERENCES relationships if the referenced event exists.
func (n *N) addETagsInBatches(c context.Context, eventID string, eTags []string) error {
// Process in batches to avoid memory issues
for i := 0; i < len(eTags); i += tagBatchSize {
end := i + tagBatchSize
if end > len(eTags) {
end = len(eTags)
}
batch := eTags[i:end]
// Use UNWIND to process multiple e-tags in a single query
// OPTIONAL MATCH ensures we only create relationships if referenced event exists
cypher := `
MATCH (e:Event {id: $eventId})
UNWIND $eventIds AS refId
OPTIONAL MATCH (ref:Event {id: refId})
WITH e, ref
WHERE ref IS NOT NULL
CREATE (e)-[:REFERENCES]->(ref)`
params := map[string]any{
"eventId": eventID,
"eventIds": batch,
}
if _, err := n.ExecuteWrite(c, cypher, params); err != nil {
return fmt.Errorf("batch %d-%d: %w", i, end, err)
}
}
return nil
}
// addOtherTagsInBatches adds generic tag relationships using UNWIND for efficiency.
// Creates Tag nodes with type and value, and TAGGED_WITH relationships.
func (n *N) addOtherTagsInBatches(c context.Context, eventID string, tags []tagTypeValue) error {
// Process in batches to avoid memory issues
for i := 0; i < len(tags); i += tagBatchSize {
end := i + tagBatchSize
if end > len(tags) {
end = len(tags)
}
batch := tags[i:end]
// Convert to map slice for Neo4j parameter passing
tagMaps := make([]map[string]string, len(batch))
for j, t := range batch {
tagMaps[j] = map[string]string{"type": t.Type, "value": t.Value}
}
// Use UNWIND to process multiple tags in a single query
cypher := `
MATCH (e:Event {id: $eventId})
UNWIND $tags AS tag
MERGE (t:Tag {type: tag.type, value: tag.value})
CREATE (e)-[:TAGGED_WITH]->(t)`
params := map[string]any{
"eventId": eventID,
"tags": tagMaps,
}
if _, err := n.ExecuteWrite(c, cypher, params); err != nil {
return fmt.Errorf("batch %d-%d: %w", i, end, err)
}
}
return nil
}
// GetSerialsFromFilter returns event serials matching a filter
func (n *N) GetSerialsFromFilter(f *filter.F) (serials types.Uint40s, err error) {
// Use QueryForSerials with background context