Skip to content

The Postgres development platform. Supabase gives you a dedicated Postgres database to build your web, mobile, and AI applications.

7.0 KB
Raw Blame
import { type SafeSqlFragment } from '@supabase/pg-meta'

import type { QueryPerformanceRow } from '../../QueryPerformance/QueryPerformance.types'
import type { Logs } from '../../Settings/Logs/Logs.types'
import {
  SCHEMA_INTROSPECTION_REGEX,
  SUPAMONITOR_EXCLUDED_APP_NAMES,
  SUPAMONITOR_EXCLUDED_ROLES,
  TRANSACTION_CONTROL_REGEX,
} from '../QueryInsights.constants'
import type { ChartDataPoint, ParsedLogEntry } from '../QueryInsights.types'

export function filterSystemLogs(
  logs: ParsedLogEntry[],
  { includeIntrospection = false }: { includeIntrospection?: boolean } = {}
): ParsedLogEntry[] {
  return logs.filter((log) => {
    if (log.user_name && (SUPAMONITOR_EXCLUDED_ROLES as readonly string[]).includes(log.user_name))
      return false
    if (
      log.application_name &&
      (SUPAMONITOR_EXCLUDED_APP_NAMES as readonly string[]).includes(log.application_name)
    )
      return false
    if (log.query && TRANSACTION_CONTROL_REGEX.test(log.query)) return false
    if (!includeIntrospection && log.query && SCHEMA_INTROSPECTION_REGEX.test(log.query))
      return false
    return true
  })
}

function asString(unknown: unknown): string | undefined {
  if (typeof unknown === 'string') return unknown
  if (unknown === null || unknown === undefined) return undefined
  return String(unknown)
}

function asNumber(unknown: unknown): number | undefined {
  if (typeof unknown === 'number') return unknown
  if (unknown === null || unknown === undefined) return undefined
  const parsed = Number(unknown)
  return Number.isNaN(parsed) ? undefined : parsed
}

export function parseSupamonitorLogs(logData: Logs['result']): ParsedLogEntry[] {
  if (!logData || logData.length === 0) return []

  return logData.map((log) => ({
    timestamp: asString(log.timestamp),
    application_name: asString(log.application_name),
    calls: asNumber(log.calls),
    database_name: asString(log.database_name),
    query: log.query,
    query_id: asNumber(log.query_id),
    total_exec_time: asNumber(log.total_exec_time),
    total_plan_time: asNumber(log.total_plan_time),
    user_name: asString(log.user_name),
    mean_exec_time: asNumber(log.mean_exec_time),
    mean_plan_time: asNumber(log.mean_plan_time),
    min_exec_time: asNumber(log.min_exec_time),
    max_exec_time: asNumber(log.max_exec_time),
    min_plan_time: asNumber(log.min_plan_time),
    max_plan_time: asNumber(log.max_plan_time),
    p50_exec_time: asNumber(log.p50_exec_time),
    p95_exec_time: asNumber(log.p95_exec_time),
    p50_plan_time: asNumber(log.p50_plan_time),
    p95_plan_time: asNumber(log.p95_plan_time),
  }))
}

export function transformLogsToChartData(parsedLogs: ParsedLogEntry[]): ChartDataPoint[] {
  if (!parsedLogs || parsedLogs.length === 0) return []

  return parsedLogs
    .map((log: ParsedLogEntry) => {
      if (!log.timestamp) return null

      const periodStart = new Date(log.timestamp).getTime()
      if (isNaN(periodStart)) return null

      const meanExecTime = parseFloat(String(log.mean_exec_time ?? 0))
      const meanPlanTime = parseFloat(String(log.mean_plan_time ?? 0))
      const calls = parseInt(String(log.calls ?? 0), 10)

      return {
        period_start: periodStart,
        timestamp: log.timestamp,
        query_latency: meanExecTime + meanPlanTime,
        mean_time: meanExecTime,
        min_time: (log.min_exec_time ?? 0) + (log.min_plan_time ?? 0),
        max_time: (log.max_exec_time ?? 0) + (log.max_plan_time ?? 0),
        stddev_time: 0,
        p50_time: (log.p50_exec_time ?? 0) + (log.p50_plan_time ?? 0),
        p95_time: (log.p95_exec_time ?? 0) + (log.p95_plan_time ?? 0),
        rows_read: 0,
        calls,
        cache_hits: 0,
        cache_misses: 0,
      }
    })
    .filter((item): item is NonNullable<typeof item> => item !== null)
    .sort((a, b) => a.period_start - b.period_start)
}

function normalizeQuery(query: string): string {
  return query.replace(/\s+/g, ' ').trim()
}

export function aggregateLogsByQuery(parsedLogs: ParsedLogEntry[]): QueryPerformanceRow[] {
  if (!parsedLogs || parsedLogs.length === 0) return []

  const queryGroups = new Map<string, ParsedLogEntry[]>()

  parsedLogs.forEach((log) => {
    const query = normalizeQuery(log.query || '')
    if (!query) return

    if (!queryGroups.has(query)) {
      queryGroups.set(query, [])
    }
    queryGroups.get(query)!.push(log)
  })

  const aggregatedData: QueryPerformanceRow[] = []
  let totalExecutionTime = 0

  const queryStats = Array.from(queryGroups.entries()).map(([query, logs]) => {
    const count = logs.length
    let totalCalls = 0
    let totalExecTime = 0
    let totalPlanTime = 0
    let p95Sum = 0
    let p95Count = 0
    let minTime = Infinity
    let maxTime = -Infinity
    const rolname = logs[0]?.user_name || ''
    const applicationName = logs[0]?.application_name || ''
    let firstSeen = logs[0]?.timestamp ?? ''

    logs.forEach((log) => {
      if (log.timestamp && (!firstSeen || log.timestamp < firstSeen)) firstSeen = log.timestamp
      const logCalls = parseInt(String(log.calls ?? 0), 10)
      totalCalls += logCalls
      totalExecTime += parseFloat(String(log.total_exec_time ?? 0))
      totalPlanTime += parseFloat(String(log.total_plan_time ?? 0))
      const logP95 =
        parseFloat(String(log.p95_exec_time ?? 0)) + parseFloat(String(log.p95_plan_time ?? 0))
      if (logP95 > 0) {
        p95Sum += logP95
        p95Count++
      }
      minTime = Math.min(
        minTime,
        parseFloat(String(log.min_exec_time ?? 0)) + parseFloat(String(log.min_plan_time ?? 0))
      )
      maxTime = Math.max(
        maxTime,
        parseFloat(String(log.max_exec_time ?? 0)) + parseFloat(String(log.max_plan_time ?? 0))
      )
    })

    const totalTime = totalExecTime + totalPlanTime
    const avgMeanTime = totalCalls > 0 ? totalTime / totalCalls : 0
    const avgP95Time = p95Count > 0 ? p95Sum / p95Count : 0
    const finalMinTime = minTime === Infinity ? 0 : minTime
    const finalMaxTime = maxTime === -Infinity ? 0 : maxTime

    totalExecutionTime += totalTime

    return {
      query,
      rolname,
      applicationName,
      firstSeen,
      count,
      avgMeanTime,
      avgP95Time,
      minTime: finalMinTime,
      maxTime: finalMaxTime,
      totalCalls,
      totalTime,
    }
  })

  queryStats.forEach((stats) => {
    const propTotalTime = totalExecutionTime > 0 ? (stats.totalTime / totalExecutionTime) * 100 : 0

    aggregatedData.push({
      query: stats.query as SafeSqlFragment,
      rolname: stats.rolname,
      application_name: stats.applicationName,
      calls: stats.totalCalls,
      mean_time: stats.avgMeanTime,
      p95_time: stats.avgP95Time,
      min_time: stats.minTime,
      max_time: stats.maxTime,
      total_time: stats.totalTime,
      rows_read: 0,
      cache_hit_rate: 0,
      prop_total_time: propTotalTime,
      index_advisor_result: null,
      _total_cache_hits: 0,
      _total_cache_misses: 0,
      _count: stats.count,
      first_seen: stats.firstSeen,
    })
  })

  return aggregatedData.sort((a, b) => b.total_time - a.total_time)
}