import { AnyFunction } from '../types'; import { isDebugTest } from '../config'; const isDebug = isDebugTest; // const orderReducedBy = 1; // reduction of bigO's order compared to the baseline bigO export const magnitude = { // CONSTANT: Math.pow(10, 9), // LOG_N: Math.pow(10, 8 - orderReducedBy), // LINEAR: Math.pow(10, 7 - orderReducedBy), // N_LOG_N: Math.pow(10, 4 - orderReducedBy), // SQUARED: Math.pow(10, 3 - orderReducedBy), // CUBED: Math.pow(10, 2 - orderReducedBy), // FACTORIAL: 20 - orderReducedBy, THOUSAND: 1000, TEN_THOUSAND: 10000, HUNDRED_THOUSAND: 100000, MILLION: 1000000, TEN_MILLION: 10000000, BILLION: 100000000 }; export const bigO = { // CONSTANT: magnitude.CONSTANT / 100000, // LOG_N: Math.log2(magnitude.LOG_N) / 1000, // LINEAR: magnitude.LINEAR / 1000, // N_LOG_N: (magnitude.N_LOG_N * Math.log2(magnitude.LOG_N)) / 1000, // SQUARED: Math.pow(magnitude.SQUARED, 2) / 1000, // CUBED: Math.pow(magnitude.SQUARED, 3) / 1000, FACTORIAL: 10000 }; function findPotentialN(input: any): number { let longestArray: any[] = []; let mostProperties: { [key: string]: any; } = {}; function recurse(obj: any) { if (Array.isArray(obj)) { if (obj.length > longestArray.length) { longestArray = obj; } } else if (typeof obj === 'object' && obj !== null) { const keys = Object.keys(obj); if (keys.length > Object.keys(mostProperties).length) { mostProperties = obj; } keys.forEach(key => { recurse(obj[key]); }); } } if (Array.isArray(input)) { input.forEach(item => { recurse(item); }); } else { recurse(input); } // return [longestArray, mostProperties] : [any[], { [key: string]: any }]; return Math.max(longestArray.length, Object.keys(mostProperties).length); } function linearRegression(x: number[], y: number[]) { const n = x.length; const sumX = x.reduce((acc, value) => acc + value, 0); const sumY = y.reduce((acc, value) => acc + value, 0); const sumXSquared = x.reduce((acc, value) => acc + value ** 2, 0); const sumXY = x.reduce((acc, value, i) => acc + value * y[i], 0); const slope = (n * sumXY - sumX * sumY) / (n * sumXSquared - sumX ** 2); const intercept = (sumY - slope * sumX) / n; const yHat = x.map(value => slope * value + intercept); const totalVariation = y.map((value, i) => (value - yHat[i]) ** 2).reduce((acc, value) => acc + value, 0); const explainedVariation = y.map(value => (value - sumY / n) ** 2).reduce((acc, value) => acc + value, 0); const rSquared = 1 - totalVariation / explainedVariation; return { slope, intercept, rSquared }; } function estimateBigO(runtimes: number[], dataSizes: number[]): string { // Make sure the input runtimes and data sizes have the same length if (runtimes.length !== dataSizes.length) { return 'Lengths of input arrays do not match'; } // Create an array to store the computational complexity of each data point const complexities: string[] = []; // Traverse different possible complexities const complexitiesToCheck: string[] = [ 'O(1)', // constant time complexity 'O(log n)', // Logarithmic time complexity 'O(n)', // linear time complexity 'O(n log n)', // linear logarithmic time complexity 'O(n^2)' // squared time complexity ]; for (const complexity of complexitiesToCheck) { // Calculate data points for fitting const fittedData: number[] = dataSizes.map(size => { if (complexity === 'O(1)') { return 1; // constant time complexity } else if (complexity === 'O(log n)') { return Math.log(size); } else if (complexity === 'O(n)') { return size; } else if (complexity === 'O(n log n)') { return size * Math.log(size); } else if (complexity === 'O(n^2)') { return size ** 2; } else { return size ** 10; } }); // Fit the data points using linear regression analysis const regressionResult = linearRegression(fittedData, runtimes); // Check the R-squared value of the fit. It is usually considered a valid fit if it is greater than 0.9. if (regressionResult.rSquared >= 0.9) { complexities.push(complexity); } } // If there is no valid fitting result, return "cannot estimate", otherwise return the estimated time complexity if (complexities.length === 0) { return 'Unable to estimate'; } else { return complexities.join(' or '); } } const methodLogs: Map = new Map(); export function logBigOMetricsWrap(fn: F, args: Parameters, fnName: string) { const startTime = performance.now(); const result = fn(args); const endTime = performance.now(); const runTime = endTime - startTime; const methodName = `${fnName}`; if (!methodLogs.has(methodName)) { methodLogs.set(methodName, []); } const methodLog = methodLogs.get(methodName); const maxDataSize = args.length === 1 && typeof args[0] === 'number' ? args[0] : findPotentialN(args); if (methodLog) { methodLog.push([runTime, maxDataSize]); if (methodLog.length >= 20) { if (isDebug) console.log('triggered', methodName, methodLog); const bigO = estimateBigO( methodLog.map(([runTime]) => runTime), methodLog.map(([runTime]) => runTime) ); if (isDebug) console.log(`Estimated Big O: ${bigO}`); methodLogs.delete(methodName); } } return result; } export function logBigOMetrics(target: any, propertyKey: string, descriptor: PropertyDescriptor) { const originalMethod = descriptor.value; descriptor.value = function (...args: any[]) { const startTime = performance.now(); const result = originalMethod.apply(this, args); const endTime = performance.now(); const runTime = endTime - startTime; const methodName = `${target.constructor.name}.${propertyKey}`; if (!methodLogs.has(methodName)) { methodLogs.set(methodName, []); } const methodLog = methodLogs.get(methodName); const maxDataSize = args.length === 1 && typeof args[0] === 'number' ? args[0] : findPotentialN(args); if (methodLog) { methodLog.push([runTime, maxDataSize]); if (methodLog.length >= 20) { if (isDebug) console.log('triggered', methodName, methodLog); const bigO = estimateBigO( methodLog.map(([runTime]) => runTime), methodLog.map(([runTime]) => runTime) ); if (isDebug) console.log(`Estimated Big O: ${bigO}`); methodLogs.delete(methodName); } } return result; }; return descriptor; } export const logPerf = function ( label: string = 'function running cost', fn: (...args: any[]) => any, args: any[], thisArg?: any ) { const start = performance.now(); let result: any; if (thisArg) { if (args && args.length > 0) result = fn.apply(thisArg, args); else result = fn.apply(thisArg); } else { if (args && args.length > 0) result = fn(...args); else result = fn(); } console.log(`${(performance.now() - start).toFixed(2)} ms, ${label}, ${result}`); };