store

package
v0.15.2 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Aug 20, 2026 License: Apache-2.0 Imports: 14 Imported by: 0

Documentation

Overview

Package store persists governed telemetry records (what optic.yaml allowed through) and answers the dashboard's query/analytics needs.

Contract: everything stored here is ALREADY restricted/redacted by the engine. The store is deliberately downstream of governance so no driver can ever see raw sensitive payloads.

Index

Constants

View Source 
const (
	DefaultAnalysisMaxRows = ext.DefaultAnalysisMaxRows
	MaxAnalysisMaxRows     = ext.MaxAnalysisMaxRows
)

Variables

This section is empty.

Functions

func AnalysisLimit added in v0.8.0 

func AnalysisLimit(limit int) int

AnalysisLimit resolves a requested limit against the defaults.

Types

type AppLog added in v0.9.0 

type AppLog = ext.AppLog

AppLog is one application log line correlated to a span. AppLogStore is OPTIONAL — a driver is complete without it; detect support with a type assertion rather than assuming.

type AppLogFilter added in v0.9.0 

type AppLogFilter = ext.AppLogFilter

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type AppLogStore added in v0.9.0 

type AppLogStore = ext.AppLogStore

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type AppLogSummary added in v0.9.0 

type AppLogSummary = ext.AppLogSummary

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type AsyncOption 

type AsyncOption func(*AsyncWriter)

func WithAppLogMaxAge added in v0.9.0 

func WithAppLogMaxAge(d time.Duration) AsyncOption

WithAppLogMaxAge sets the retention horizon for application log lines. It is a no-op unless the store implements ext.AppLogStore.

func WithDropCallback 

func WithDropCallback(fn func()) AsyncOption

WithDropCallback wires drop accounting into the metrics collector.

func WithMaxAge 

func WithMaxAge(d time.Duration) AsyncOption

WithMaxAge deletes records older than d on the same schedule. Row-count and age limits compose: whichever removes a record first wins.

func WithRetention 

func WithRetention(maxRows int64) AsyncOption

WithRetention prunes the store down to maxRows periodically.

func WithSpanMaxAge added in v0.15.0 

func WithSpanMaxAge(d time.Duration) AsyncOption

WithSpanMaxAge sets the retention horizon for inner spans. It is a no-op unless the store implements ext.SpanStore.

type AsyncWriter 

type AsyncWriter struct {
	// contains filtered or unexported fields
}

AsyncWriter decouples the request hot path from persistence: Enqueue is a non-blocking channel send, and a single background worker drains into the LogStore. Under backpressure records are DROPPED (and counted) — an observability tool must never become the bottleneck it measures.

func NewAsyncWriter 

func NewAsyncWriter(s LogStore, queueSize int, logger *slog.Logger, opts ...AsyncOption) *AsyncWriter

func (*AsyncWriter) Close 

func (w *AsyncWriter) Close() error

Close drains outstanding records, then closes the underlying store.

func (*AsyncWriter) Enqueue 

func (w *AsyncWriter) Enqueue(rec *Record)

Enqueue hands off a record without ever blocking the caller.

type ClickHouseStore added in v0.8.0 

type ClickHouseStore struct {
	// contains filtered or unexported fields
}

ClickHouseStore is the column-oriented LogStore driver.

Why a third driver: every dashboard query here is a time-bucketed aggregation over a table that only grows — Stats, RouteStats, ServiceStats, UsageByLabel. SQLite and Postgres are row stores answering those with sorts and scans; ClickHouse answers them with column reads, and is a common choice for teams already keeping API traffic at volume.

Two things work differently here than in the row stores, and both are visible in this file rather than hidden:

  • There is no autoincrement. IDs are generated client-side (see nextID) rather than by the database.
  • DELETE is a mutation, applied asynchronously by default. Every delete path runs with mutations_sync=2 so Purge and the retention functions behave synchronously, as the shared conformance suite requires and as an erasure request demands.

func NewClickHouse added in v0.8.0 

func NewClickHouse(dsn string) (*ClickHouseStore, error)

NewClickHouse opens (and migrates) a ClickHouse-backed store. dsn is a clickhouse-go URL: clickhouse://user:pass@host:9000/database

func (*ClickHouseStore) AppLogStats added in v0.11.0 

func (s *ClickHouseStore) AppLogStats(ctx context.Context, since time.Time) (*ext.AppLogSummary, error)

AppLogStats aggregates in the database rather than by counting rows in the caller — a summary of the first page is wrong at exactly the volumes where a summary starts to matter.

func (*ClickHouseStore) Close added in v0.8.0 

func (s *ClickHouseStore) Close() error

func (*ClickHouseStore) Count added in v0.8.0 

func (s *ClickHouseStore) Count(ctx context.Context) (int64, error)

func (*ClickHouseStore) CountAppLogs added in v0.11.0 

func (s *ClickHouseStore) CountAppLogs(ctx context.Context) (int64, error)

func (*ClickHouseStore) CountSpans added in v0.15.0 

func (s *ClickHouseStore) CountSpans(ctx context.Context) (int64, error)

func (*ClickHouseStore) Get added in v0.8.0 

func (s *ClickHouseStore) Get(ctx context.Context, id int64) (*Record, error)

func (*ClickHouseStore) Prune added in v0.8.0 

func (s *ClickHouseStore) Prune(ctx context.Context, maxRows int64) (int64, error)

func (*ClickHouseStore) PruneAppLogsBefore added in v0.11.0 

func (s *ClickHouseStore) PruneAppLogsBefore(ctx context.Context, cutoff time.Time) (int64, error)

PruneAppLogsBefore enforces the app-log retention horizon.

mutations_sync=2 for the same reason the record path uses it: an unsynchronised ALTER ... DELETE returns before the data is gone, so retention would report success while the rows were still readable.

func (*ClickHouseStore) PruneBefore added in v0.8.0 

func (s *ClickHouseStore) PruneBefore(ctx context.Context, cutoff time.Time) (int64, error)

func (*ClickHouseStore) PruneSpansBefore added in v0.15.0 

func (s *ClickHouseStore) PruneSpansBefore(ctx context.Context, cutoff time.Time) (int64, error)

func (*ClickHouseStore) Purge added in v0.8.0 

func (s *ClickHouseStore) Purge(ctx context.Context, label, value string, before time.Time) (int64, error)

Purge implements erasure requests. The label match is on the exact `"label":"value"` pair extracted from the JSON document, not a substring or a LIKE pattern — deleting a neighbouring tenant's data is the one mistake an erasure tool must never make, and a LIKE with unescaped metacharacters is exactly how the SQLite driver once made it.

func (*ClickHouseStore) Query added in v0.8.0 

func (s *ClickHouseStore) Query(ctx context.Context, f Filter) ([]Record, int64, error)

func (*ClickHouseStore) QueryAppLogs added in v0.11.0 

func (s *ClickHouseStore) QueryAppLogs(ctx context.Context, f ext.AppLogFilter) ([]ext.AppLog, int64, error)

QueryAppLogs returns matching lines oldest-first.

func (*ClickHouseStore) QuerySpans added in v0.15.0 

func (s *ClickHouseStore) QuerySpans(ctx context.Context, f ext.SpanFilter) ([]ext.Span, int64, error)

func (*ClickHouseStore) Recent added in v0.8.0 

func (s *ClickHouseStore) Recent(ctx context.Context, since time.Time, limit int) ([]Record, error)

func (*ClickHouseStore) RecentFunc added in v0.8.0 

func (s *ClickHouseStore) RecentFunc(ctx context.Context, since time.Time, limit int, fn func(*Record) error) error

func (*ClickHouseStore) RouteStats added in v0.8.0 

func (s *ClickHouseStore) RouteStats(ctx context.Context, since time.Time) ([]RouteDetail, error)

func (*ClickHouseStore) RuleMatchCounts added in v0.8.0 

func (s *ClickHouseStore) RuleMatchCounts(ctx context.Context, since time.Time, ruleNames []string) ([]RuleMatch, error)

RuleMatchCounts counts rule firings in a single pass. matched_rules is a JSON array of strings; arrayJoin over JSONExtract expands it so the counting happens in the database rather than in N round trips.

func (*ClickHouseStore) Save added in v0.8.0 

func (s *ClickHouseStore) Save(ctx context.Context, r *Record) error

func (*ClickHouseStore) SaveAppLogs added in v0.11.0 

func (s *ClickHouseStore) SaveAppLogs(ctx context.Context, lines []ext.AppLog) error

SaveAppLogs writes a batch.

ClickHouse has no autoincrement, so ids come from the same timestamp+sequence generator the record path uses — monotonic enough to order lines written in the same millisecond, which is exactly when a request logs several times.

func (*ClickHouseStore) SaveSpans added in v0.15.0 

func (s *ClickHouseStore) SaveSpans(ctx context.Context, list []ext.Span) error

func (*ClickHouseStore) ServiceStats added in v0.8.0 

func (s *ClickHouseStore) ServiceStats(ctx context.Context, since time.Time) ([]ServiceStat, error)

func (*ClickHouseStore) SpanBreakdown added in v0.15.0 

func (s *ClickHouseStore) SpanBreakdown(ctx context.Context, since time.Time, route string, limit int) ([]ext.SpanBreakdown, error)

func (*ClickHouseStore) SpanStats added in v0.15.0 

func (s *ClickHouseStore) SpanStats(ctx context.Context, since time.Time) (*ext.SpanSummary, error)

func (*ClickHouseStore) Stats added in v0.8.0 

func (s *ClickHouseStore) Stats(ctx context.Context, since time.Time, bucket time.Duration) (*Stats, error)

func (*ClickHouseStore) Traces added in v0.14.0 

func (s *ClickHouseStore) Traces(ctx context.Context, f ext.TraceFilter) ([]ext.TraceSummary, int64, error)

Traces rolls the recorded hops up into one row per trace.

ClickHouse has no DISTINCT ON and correlated subqueries are a trap here, so the root hop is picked with argMin over a sort key that expresses both cases at once: `(parent_span != ”, ts)` puts a parentless hop first and falls back to the earliest hop when the entry service was never instrumented.

func (*ClickHouseStore) UsageByLabel added in v0.8.0 

func (s *ClickHouseStore) UsageByLabel(ctx context.Context, since time.Time, label string) ([]Usage, error)

UsageByLabel aggregates in Go for the same reason the other drivers do: labels and meters are stored as JSON documents whose keys are not known in advance, so grouping in SQL would need a per-key query.

type Filter 

type Filter = ext.Filter

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type LogStore 

type LogStore = ext.Store

LogStore is the persistence contract. Implementations must be safe for concurrent use. Save is called from the async writer, never from the request hot path.

type PostgresStore 

type PostgresStore struct {
	// contains filtered or unexported fields
}

PostgresStore is the multi-node LogStore. SQLite is perfect for a sidecar with one writer; the moment several agents (or several replicas of one agent) need to share history, they need a server they can all reach.

Notes on the schema choice:

  • JSONB rather than TEXT for headers/labels/meters, so usage aggregation and label filtering happen in the database instead of in Go.
  • Timestamps stay as bigint milliseconds, matching the SQLite driver, so records are portable between the two without conversion surprises.

func NewPostgres 

func NewPostgres(dsn string) (*PostgresStore, error)

NewPostgres opens (and migrates) a Postgres-backed store. dsn is a standard libpq/pgx URL: postgres://user:pass@host:5432/db?sslmode=disable

func (*PostgresStore) AppLogStats added in v0.11.0 

func (s *PostgresStore) AppLogStats(ctx context.Context, since time.Time) (*ext.AppLogSummary, error)

AppLogStats aggregates in the database rather than by counting rows in the caller: a dashboard that summarises the first page and calls it a total is quietly lying at exactly the volumes where the summary starts to matter.

func (*PostgresStore) Close 

func (s *PostgresStore) Close() error

func (*PostgresStore) Count 

func (s *PostgresStore) Count(ctx context.Context) (int64, error)

func (*PostgresStore) CountAppLogs added in v0.11.0 

func (s *PostgresStore) CountAppLogs(ctx context.Context) (int64, error)

func (*PostgresStore) CountSpans added in v0.15.0 

func (s *PostgresStore) CountSpans(ctx context.Context) (int64, error)

func (*PostgresStore) Get 

func (s *PostgresStore) Get(ctx context.Context, id int64) (*Record, error)

func (*PostgresStore) Prune 

func (s *PostgresStore) Prune(ctx context.Context, maxRows int64) (int64, error)

func (*PostgresStore) PruneAppLogsBefore added in v0.11.0 

func (s *PostgresStore) PruneAppLogsBefore(ctx context.Context, cutoff time.Time) (int64, error)

PruneAppLogsBefore enforces the app-log retention horizon, which is separate from the record horizon because the volumes differ by orders of magnitude.

func (*PostgresStore) PruneBefore 

func (s *PostgresStore) PruneBefore(ctx context.Context, cutoff time.Time) (int64, error)

func (*PostgresStore) PruneSpansBefore added in v0.15.0 

func (s *PostgresStore) PruneSpansBefore(ctx context.Context, cutoff time.Time) (int64, error)

func (*PostgresStore) Purge 

func (s *PostgresStore) Purge(ctx context.Context, label, value string, before time.Time) (int64, error)

func (*PostgresStore) Query 

func (s *PostgresStore) Query(ctx context.Context, f Filter) ([]Record, int64, error)

func (*PostgresStore) QueryAppLogs added in v0.11.0 

func (s *PostgresStore) QueryAppLogs(ctx context.Context, f ext.AppLogFilter) ([]ext.AppLog, int64, error)

QueryAppLogs returns matching lines oldest-first: reading what a request did means reading it in the order it happened.

func (*PostgresStore) QuerySpans added in v0.15.0 

func (s *PostgresStore) QuerySpans(ctx context.Context, f ext.SpanFilter) ([]ext.Span, int64, error)

func (*PostgresStore) Recent 

func (s *PostgresStore) Recent(ctx context.Context, since time.Time, limit int) ([]Record, error)

func (*PostgresStore) RecentFunc added in v0.8.0 

func (s *PostgresStore) RecentFunc(ctx context.Context, since time.Time, limit int, fn func(*Record) error) error

RecentFunc streams records one at a time so an analysis pass costs one record of memory rather than the whole window. Recent() materialises the same rows; prefer this wherever the caller only folds over them.

func (*PostgresStore) RouteStats 

func (s *PostgresStore) RouteStats(ctx context.Context, since time.Time) ([]RouteDetail, error)

func (*PostgresStore) RuleMatchCounts 

func (s *PostgresStore) RuleMatchCounts(ctx context.Context, since time.Time, ruleNames []string) ([]RuleMatch, error)

func (*PostgresStore) Save 

func (s *PostgresStore) Save(ctx context.Context, r *Record) error

func (*PostgresStore) SaveAppLogs added in v0.11.0 

func (s *PostgresStore) SaveAppLogs(ctx context.Context, lines []ext.AppLog) error

SaveAppLogs writes a batch in one round trip.

Batching is the point: application logs arrive at many lines per request, and a statement per line makes the store the slowest part of the system.

func (*PostgresStore) SaveSpans added in v0.15.0 

func (s *PostgresStore) SaveSpans(ctx context.Context, list []ext.Span) error

func (*PostgresStore) ServiceStats 

func (s *PostgresStore) ServiceStats(ctx context.Context, since time.Time) ([]ServiceStat, error)

func (*PostgresStore) SpanBreakdown added in v0.15.0 

func (s *PostgresStore) SpanBreakdown(ctx context.Context, since time.Time, route string, limit int) ([]ext.SpanBreakdown, error)

SpanBreakdown uses percentile_cont, so the p95 comes from the same pass rather than one ordered subquery per operation name.

func (*PostgresStore) SpanStats added in v0.15.0 

func (s *PostgresStore) SpanStats(ctx context.Context, since time.Time) (*ext.SpanSummary, error)

func (*PostgresStore) Stats 

func (s *PostgresStore) Stats(ctx context.Context, since time.Time, bucket time.Duration) (*Stats, error)

func (*PostgresStore) Traces added in v0.14.0 

func (s *PostgresStore) Traces(ctx context.Context, f ext.TraceFilter) ([]ext.TraceSummary, int64, error)

Traces rolls the recorded hops up into one row per trace. See the SQLite implementation for the shape; the difference here is DISTINCT ON, which picks the root hop without a correlated subquery per trace.

func (*PostgresStore) UsageByLabel 

func (s *PostgresStore) UsageByLabel(ctx context.Context, since time.Time, label string) ([]Usage, error)

type Record 

type Record = ext.Record

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type RouteDetail 

type RouteDetail = ext.RouteDetail

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type RouteStat 

type RouteStat = ext.RouteStat

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type RuleMatch 

type RuleMatch = ext.RuleMatch

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type SQLiteStore 

type SQLiteStore struct {
	// contains filtered or unexported fields
}

SQLiteStore is the default embedded LogStore. WAL mode keeps concurrent dashboard reads from blocking the single async writer.

func NewSQLite 

func NewSQLite(dsn string) (*SQLiteStore, error)

func (*SQLiteStore) AppLogStats added in v0.9.0 

func (s *SQLiteStore) AppLogStats(ctx context.Context, since time.Time) (*ext.AppLogSummary, error)

AppLogStats aggregates stored lines. Three cheap grouped counts rather than one walk: the point of asking the database is not having to carry the rows.

func (*SQLiteStore) Close 

func (s *SQLiteStore) Close() error

func (*SQLiteStore) Count 

func (s *SQLiteStore) Count(ctx context.Context) (int64, error)

func (*SQLiteStore) CountAppLogs added in v0.9.0 

func (s *SQLiteStore) CountAppLogs(ctx context.Context) (int64, error)

func (*SQLiteStore) CountSpans added in v0.15.0 

func (s *SQLiteStore) CountSpans(ctx context.Context) (int64, error)

func (*SQLiteStore) Get 

func (s *SQLiteStore) Get(ctx context.Context, id int64) (*Record, error)

func (*SQLiteStore) Prune 

func (s *SQLiteStore) Prune(ctx context.Context, maxRows int64) (int64, error)

func (*SQLiteStore) PruneAppLogsBefore added in v0.9.0 

func (s *SQLiteStore) PruneAppLogsBefore(ctx context.Context, cutoff time.Time) (int64, error)

PruneAppLogsBefore enforces the app-log retention horizon, which is separate from the record horizon because the volumes differ by orders of magnitude.

func (*SQLiteStore) PruneBefore 

func (s *SQLiteStore) PruneBefore(ctx context.Context, cutoff time.Time) (int64, error)

func (*SQLiteStore) PruneSpansBefore added in v0.15.0 

func (s *SQLiteStore) PruneSpansBefore(ctx context.Context, cutoff time.Time) (int64, error)

func (*SQLiteStore) Purge 

func (s *SQLiteStore) Purge(ctx context.Context, label, value string, before time.Time) (int64, error)

Purge implements erasure requests. Labels are stored as a JSON object, so the match is on the exact `"label":"value"` pair — substring matching on the value alone would delete another tenant's data, which is the one mistake an erasure tool must never make.

func (*SQLiteStore) Query 

func (s *SQLiteStore) Query(ctx context.Context, f Filter) ([]Record, int64, error)

func (*SQLiteStore) QueryAppLogs added in v0.9.0 

func (s *SQLiteStore) QueryAppLogs(ctx context.Context, f ext.AppLogFilter) ([]ext.AppLog, int64, error)

QueryAppLogs returns matching lines oldest-first: reading what a request did means reading it in the order it happened.

func (*SQLiteStore) QuerySpans added in v0.15.0 

func (s *SQLiteStore) QuerySpans(ctx context.Context, f ext.SpanFilter) ([]ext.Span, int64, error)

QuerySpans returns matching spans OLDEST-FIRST: reading a request's work means reading it in the order it happened.

func (*SQLiteStore) Recent 

func (s *SQLiteStore) Recent(ctx context.Context, since time.Time, limit int) ([]Record, error)

func (*SQLiteStore) RecentFunc added in v0.8.0 

func (s *SQLiteStore) RecentFunc(ctx context.Context, since time.Time, limit int, fn func(*Record) error) error

RecentFunc streams records one at a time so an analysis pass costs one record of memory rather than the whole window. Recent() materialises the same rows; prefer this wherever the caller only folds over them.

func (*SQLiteStore) RouteStats 

func (s *SQLiteStore) RouteStats(ctx context.Context, since time.Time) ([]RouteDetail, error)

func (*SQLiteStore) RuleMatchCounts 

func (s *SQLiteStore) RuleMatchCounts(ctx context.Context, since time.Time, ruleNames []string) ([]RuleMatch, error)

RuleMatchCounts counts how often each named rule fired.

This used to run one `matched_rules LIKE '%"name"%'` per rule — N full table scans per dashboard load, since a LIKE pattern with a leading wildcard can never use an index. It is now a single pass: json_each expands the JSON array into rows, the ts index bounds the window, and GROUP BY does the counting in SQLite instead of N round trips.

func (*SQLiteStore) Save 

func (s *SQLiteStore) Save(ctx context.Context, r *Record) error

func (*SQLiteStore) SaveAppLogs added in v0.9.0 

func (s *SQLiteStore) SaveAppLogs(ctx context.Context, lines []ext.AppLog) error

SaveAppLogs writes a batch in one transaction. Batching is the point — application logs arrive at many lines per request, and a transaction per line makes the store the slowest part of the system.

func (*SQLiteStore) SaveSpans added in v0.15.0 

func (s *SQLiteStore) SaveSpans(ctx context.Context, list []ext.Span) error

SaveSpans persists a batch in one transaction. Called from the async writer, never from a request path.

func (*SQLiteStore) ServiceStats 

func (s *SQLiteStore) ServiceStats(ctx context.Context, since time.Time) ([]ServiceStat, error)

func (*SQLiteStore) SpanBreakdown added in v0.15.0 

func (s *SQLiteStore) SpanBreakdown(ctx context.Context, since time.Time, route string, limit int) ([]ext.SpanBreakdown, error)

SpanBreakdown answers "where did the time go" for a window.

Requests is counted alongside Count on purpose: Count/Requests is the per-request multiplier, and a multiplier of 40 on a query named once in the source is the shape an N+1 makes. A total alone hides it — 4,000 calls looks like busy traffic until you know it was 100 requests.

func (*SQLiteStore) SpanStats added in v0.15.0 

func (s *SQLiteStore) SpanStats(ctx context.Context, since time.Time) (*ext.SpanSummary, error)

func (*SQLiteStore) Stats 

func (s *SQLiteStore) Stats(ctx context.Context, since time.Time, bucket time.Duration) (*Stats, error)

func (*SQLiteStore) Traces added in v0.14.0 

func (s *SQLiteStore) Traces(ctx context.Context, f ext.TraceFilter) ([]ext.TraceSummary, int64, error)

func (*SQLiteStore) UsageByLabel 

func (s *SQLiteStore) UsageByLabel(ctx context.Context, since time.Time, label string) ([]Usage, error)

UsageByLabel aggregates in Go rather than SQL: labels/meters are stored as JSON text, and the window is bounded by retention (~100k rows), so a single narrow scan is simpler and fast enough.

type ServiceStat 

type ServiceStat = ext.ServiceStat

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type Span added in v0.15.0 

type Span = ext.Span

Span is one operation inside a request. SpanStore is OPTIONAL for the same reason the two above are.

type SpanBreakdown added in v0.15.0 

type SpanBreakdown = ext.SpanBreakdown

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type SpanFilter added in v0.15.0 

type SpanFilter = ext.SpanFilter

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type SpanStore added in v0.15.0 

type SpanStore = ext.SpanStore

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type SpanSummary added in v0.15.0 

type SpanSummary = ext.SpanSummary

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type Stats 

type Stats = ext.Stats

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type TimeBucket 

type TimeBucket = ext.TimeBucket

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type TraceFilter added in v0.14.0 

type TraceFilter = ext.TraceFilter

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type TraceStore added in v0.14.0 

type TraceStore = ext.TraceStore

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

type TraceSummary added in v0.14.0 

type TraceSummary = ext.TraceSummary

TraceSummary rolls every hop of one request into a row. TraceStore is OPTIONAL for the same reason AppLogStore is: Store is implemented outside this module, so it cannot grow methods.

type Usage 

type Usage = ext.Usage

The record types and the store contract live in ext, the public extension surface, so an out-of-tree driver can name them. These aliases are not conversions — Record and ext.Record are the same type — so nothing in the rest of the codebase changes, and there is exactly one Record in the program.

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.