- Update default admin email to admin@datastack.dev - Improve storage flush logic with partial success handling - Enhance CSV download reliability using fetch and Blob - Adjust simulator protocol resolution and timing - Fix date range filtering in data API query
590 lines
18 KiB
TypeScript
590 lines
18 KiB
TypeScript
import fs from 'fs';
|
|
import path from 'path';
|
|
import { Parser } from 'json2csv';
|
|
|
|
export interface DataEvent {
|
|
clientId: string;
|
|
eventId: string;
|
|
timestamp: string; // ISO 8601
|
|
payload: string; // JSON stringified
|
|
}
|
|
|
|
export interface StorageFileMeta {
|
|
clientId: string;
|
|
date: string;
|
|
filename: string;
|
|
filePath: string;
|
|
sizeBytes: number;
|
|
linesCount: number;
|
|
modifiedAt: Date;
|
|
}
|
|
|
|
interface RowCountCacheEntry {
|
|
sizeBytes: number;
|
|
mtimeMs: number;
|
|
linesCount: number;
|
|
}
|
|
|
|
const LOCK_RETRY_MIN_MS = 5;
|
|
const LOCK_RETRY_MAX_MS = 18;
|
|
const LOCK_TIMEOUT_MS = 10_000;
|
|
const STALE_LOCK_AGE_MS = 30_000;
|
|
|
|
function sleepMs(ms: number): Promise<void> {
|
|
return new Promise((resolve) => setTimeout(resolve, ms));
|
|
}
|
|
|
|
export class CSVStorageManager {
|
|
// Base storage directory: /storage/data
|
|
public static baseStorageDir = path.resolve(process.cwd(), 'storage', 'data');
|
|
public static maxFileSizeBytes = 100 * 1024 * 1024; // 100MB rotation limit
|
|
private static rowCountCache = new Map<string, RowCountCacheEntry>();
|
|
|
|
public static clearRowCountCache(): void {
|
|
this.rowCountCache.clear();
|
|
}
|
|
|
|
/**
|
|
* Ensures client directory exists
|
|
*/
|
|
public static ensureClientDir(clientId: string): string {
|
|
const safeClientId = clientId.replace(/[^a-zA-Z0-9_-]/g, '_');
|
|
const clientDir = path.join(this.baseStorageDir, safeClientId);
|
|
if (!fs.existsSync(clientDir)) {
|
|
fs.mkdirSync(clientDir, { recursive: true });
|
|
}
|
|
return clientDir;
|
|
}
|
|
|
|
/**
|
|
* Acquires an exclusive cooperative lock file for a client+date partition across Port 3000 & Port 4000
|
|
*/
|
|
public static async withPartitionLock<T>(
|
|
clientId: string,
|
|
dateStr: string,
|
|
fn: () => Promise<T> | T,
|
|
): Promise<T> {
|
|
const clientDir = this.ensureClientDir(clientId);
|
|
const date = this.extractSafePartitionDate(dateStr);
|
|
const lockPath = path.join(clientDir, `.${date}.csv.lock`);
|
|
const deadline = Date.now() + LOCK_TIMEOUT_MS;
|
|
|
|
let acquired = false;
|
|
while (Date.now() < deadline) {
|
|
try {
|
|
const fd = fs.openSync(lockPath, 'wx');
|
|
try {
|
|
fs.writeFileSync(fd, `${process.pid}:${Date.now()}`, 'utf8');
|
|
} finally {
|
|
fs.closeSync(fd);
|
|
}
|
|
acquired = true;
|
|
break;
|
|
} catch (err: unknown) {
|
|
const code = (err as NodeJS.ErrnoException)?.code;
|
|
if (code !== 'EEXIST') {
|
|
throw err;
|
|
}
|
|
try {
|
|
const lockStat = fs.statSync(lockPath);
|
|
if (Date.now() - lockStat.mtimeMs > STALE_LOCK_AGE_MS) {
|
|
fs.unlinkSync(lockPath);
|
|
continue;
|
|
}
|
|
} catch {
|
|
// Lock was released concurrently
|
|
}
|
|
const jitter =
|
|
LOCK_RETRY_MIN_MS +
|
|
Math.floor(Math.random() * (LOCK_RETRY_MAX_MS - LOCK_RETRY_MIN_MS));
|
|
await sleepMs(jitter);
|
|
}
|
|
}
|
|
|
|
if (!acquired) {
|
|
throw new Error(`Timed out waiting for CSV partition lock: ${lockPath}`);
|
|
}
|
|
|
|
try {
|
|
return await fn();
|
|
} finally {
|
|
try {
|
|
if (fs.existsSync(lockPath)) {
|
|
fs.unlinkSync(lockPath);
|
|
}
|
|
} catch {
|
|
// Ignore cleanup failure if already unlinked
|
|
}
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Normalizes and validates any timestamp or date string into a strict YYYY-MM-DD partition key.
|
|
* Returns fallback (default: today's date) if unparseable, or null when fallback is null.
|
|
*/
|
|
public static extractSafePartitionDate(rawTimestamp?: string, fallback?: string): string;
|
|
public static extractSafePartitionDate(rawTimestamp: string | undefined, fallback: null): string | null;
|
|
public static extractSafePartitionDate(
|
|
rawTimestamp?: string,
|
|
fallback: string | null = new Date().toISOString().slice(0, 10),
|
|
): string | null {
|
|
if (!rawTimestamp || typeof rawTimestamp !== 'string') {
|
|
return fallback;
|
|
}
|
|
const trimmed = rawTimestamp.trim();
|
|
const prefix = trimmed.slice(0, 10);
|
|
if (/^\d{4}-\d{2}-\d{2}$/.test(prefix)) {
|
|
const parsedPrefix = new Date(`${prefix}T00:00:00.000Z`);
|
|
if (!Number.isNaN(parsedPrefix.getTime())) {
|
|
return prefix;
|
|
}
|
|
}
|
|
if (/^\d+(\.\d+)?$/.test(trimmed)) {
|
|
const num = Number(trimmed);
|
|
const ms = num < 1e11 ? num * 1000 : num;
|
|
const d = new Date(ms);
|
|
if (!Number.isNaN(d.getTime())) {
|
|
return d.toISOString().slice(0, 10);
|
|
}
|
|
}
|
|
const parsed = new Date(trimmed);
|
|
if (!Number.isNaN(parsed.getTime())) {
|
|
return parsed.toISOString().slice(0, 10);
|
|
}
|
|
return fallback;
|
|
}
|
|
|
|
/**
|
|
* Stateful RFC 4180 quote-aware chunked row counter so multi-100MB CSVs never load wholly into RAM
|
|
*/
|
|
private static createChunkRowCounter() {
|
|
let inQuotes = false;
|
|
let pendingQuote = false;
|
|
let currentLineHasChars = false;
|
|
let totalRows = 0;
|
|
|
|
return {
|
|
pushChunk(chunk: Buffer | string) {
|
|
const str = typeof chunk === 'string' ? chunk : chunk.toString('utf8');
|
|
for (let i = 0; i < str.length; i += 1) {
|
|
const ch = str[i];
|
|
if (pendingQuote) {
|
|
pendingQuote = false;
|
|
if (ch === '"') {
|
|
// Escaped quote "" across chunk boundary or inline
|
|
continue;
|
|
}
|
|
inQuotes = !inQuotes;
|
|
}
|
|
|
|
if (ch === '"') {
|
|
currentLineHasChars = true;
|
|
if (inQuotes) {
|
|
if (i + 1 < str.length) {
|
|
if (str[i + 1] === '"') {
|
|
i += 1;
|
|
} else {
|
|
inQuotes = false;
|
|
}
|
|
} else {
|
|
pendingQuote = true;
|
|
}
|
|
} else {
|
|
inQuotes = true;
|
|
}
|
|
} else if (ch === '\n' && !inQuotes) {
|
|
if (currentLineHasChars) {
|
|
totalRows += 1;
|
|
currentLineHasChars = false;
|
|
}
|
|
} else if (ch !== '\r' && ch !== ' ' && ch !== '\t') {
|
|
currentLineHasChars = true;
|
|
}
|
|
}
|
|
},
|
|
finish(): number {
|
|
if (currentLineHasChars) {
|
|
totalRows += 1;
|
|
}
|
|
return Math.max(0, totalRows - 1);
|
|
},
|
|
};
|
|
}
|
|
|
|
private static countCsvDataRowsChunkedSync(filePath: string): number {
|
|
const counter = this.createChunkRowCounter();
|
|
const buf = Buffer.allocUnsafe(64 * 1024);
|
|
const fd = fs.openSync(filePath, 'r');
|
|
try {
|
|
let bytesRead = 0;
|
|
while ((bytesRead = fs.readSync(fd, buf, 0, buf.length, null)) > 0) {
|
|
counter.pushChunk(buf.subarray(0, bytesRead));
|
|
}
|
|
return counter.finish();
|
|
} finally {
|
|
fs.closeSync(fd);
|
|
}
|
|
}
|
|
|
|
private static async countCsvDataRowsStreamAsync(
|
|
filePath: string,
|
|
): Promise<number> {
|
|
const counter = this.createChunkRowCounter();
|
|
return new Promise<number>((resolve, reject) => {
|
|
const stream = fs.createReadStream(filePath, {
|
|
highWaterMark: 64 * 1024,
|
|
});
|
|
stream.on('data', (chunk) => counter.pushChunk(chunk));
|
|
stream.on('end', () => resolve(counter.finish()));
|
|
stream.on('error', (err) => reject(err));
|
|
});
|
|
}
|
|
|
|
/**
|
|
* Resolves target CSV file path with daily and configured MB rotation rules
|
|
*/
|
|
public static getTargetFilePath(clientId: string, dateStr?: string): string {
|
|
const clientDir = this.ensureClientDir(clientId);
|
|
const date = this.extractSafePartitionDate(dateStr);
|
|
let baseFileName = `${date}.csv`;
|
|
let targetPath = path.join(clientDir, baseFileName);
|
|
|
|
// Check for file rotation
|
|
let partIndex = 1;
|
|
while (fs.existsSync(targetPath)) {
|
|
try {
|
|
const stats = fs.statSync(targetPath);
|
|
if (stats.size >= this.maxFileSizeBytes) {
|
|
partIndex += 1;
|
|
baseFileName = `${date}-part${partIndex}.csv`;
|
|
targetPath = path.join(clientDir, baseFileName);
|
|
} else {
|
|
break;
|
|
}
|
|
} catch {
|
|
break;
|
|
}
|
|
}
|
|
|
|
return targetPath;
|
|
}
|
|
|
|
/**
|
|
* Batch append events into target CSV file under cooperative partition lock.
|
|
* If file doesn't exist, writes header: clientId,eventId,timestamp,payload
|
|
*/
|
|
public static async appendBatch(
|
|
events: DataEvent[],
|
|
): Promise<{ written: number; filesUpdated: string[] }> {
|
|
if (!events || events.length === 0) {
|
|
return { written: 0, filesUpdated: [] };
|
|
}
|
|
|
|
// Group events by client + date
|
|
const grouped = new Map<string, DataEvent[]>();
|
|
for (const evt of events) {
|
|
const date = this.extractSafePartitionDate(evt.timestamp);
|
|
const key = `${evt.clientId}:::${date}`;
|
|
if (!grouped.has(key)) {
|
|
grouped.set(key, []);
|
|
}
|
|
grouped.get(key)!.push(evt);
|
|
}
|
|
|
|
const filesUpdated: string[] = [];
|
|
const fields = ['clientId', 'eventId', 'timestamp', 'payload'];
|
|
|
|
for (const [key, batch] of grouped.entries()) {
|
|
const [clientId, date] = key.split(':::');
|
|
const targetPath = await this.withPartitionLock(clientId, date, () => {
|
|
const resolvedPath = this.getTargetFilePath(clientId, date);
|
|
let hasContent = false;
|
|
try {
|
|
hasContent =
|
|
fs.existsSync(resolvedPath) && fs.statSync(resolvedPath).size > 0;
|
|
} catch {
|
|
hasContent = false;
|
|
}
|
|
|
|
const cleanBatch = batch.map((item) => ({
|
|
clientId: item.clientId,
|
|
eventId: item.eventId,
|
|
timestamp: item.timestamp,
|
|
payload:
|
|
typeof item.payload === 'string'
|
|
? item.payload
|
|
: JSON.stringify(item.payload),
|
|
}));
|
|
|
|
const parser = new Parser({ fields, header: !hasContent, eol: '\n' });
|
|
const csvData = parser.parse(cleanBatch);
|
|
|
|
if (hasContent) {
|
|
fs.appendFileSync(resolvedPath, '\n' + csvData, 'utf8');
|
|
} else {
|
|
fs.writeFileSync(resolvedPath, csvData, 'utf8');
|
|
}
|
|
|
|
// Invalidate row cache for modified file immediately
|
|
this.rowCountCache.delete(resolvedPath);
|
|
return resolvedPath;
|
|
});
|
|
|
|
filesUpdated.push(targetPath);
|
|
}
|
|
|
|
return { written: events.length, filesUpdated };
|
|
}
|
|
|
|
/**
|
|
* List all stored CSV files with cached row counts (invalidated on mtime/size changes)
|
|
*/
|
|
public static listAllFiles(): StorageFileMeta[] {
|
|
const results: StorageFileMeta[] = [];
|
|
if (!fs.existsSync(this.baseStorageDir)) {
|
|
this.rowCountCache.clear();
|
|
return results;
|
|
}
|
|
|
|
const activePaths = new Set<string>();
|
|
try {
|
|
const clientDirs = fs.readdirSync(this.baseStorageDir);
|
|
for (const clientDir of clientDirs) {
|
|
const fullClientPath = path.join(this.baseStorageDir, clientDir);
|
|
if (!fs.statSync(fullClientPath).isDirectory()) continue;
|
|
|
|
const files = fs.readdirSync(fullClientPath);
|
|
for (const file of files) {
|
|
if (!file.endsWith('.csv')) continue;
|
|
const fullFilePath = path.join(fullClientPath, file);
|
|
const stats = fs.statSync(fullFilePath);
|
|
activePaths.add(fullFilePath);
|
|
|
|
const cached = this.rowCountCache.get(fullFilePath);
|
|
let linesCount = 0;
|
|
if (
|
|
cached &&
|
|
cached.sizeBytes === stats.size &&
|
|
cached.mtimeMs === stats.mtimeMs
|
|
) {
|
|
linesCount = cached.linesCount;
|
|
} else {
|
|
try {
|
|
linesCount = this.countCsvDataRowsChunkedSync(fullFilePath);
|
|
this.rowCountCache.set(fullFilePath, {
|
|
sizeBytes: stats.size,
|
|
mtimeMs: stats.mtimeMs,
|
|
linesCount,
|
|
});
|
|
} catch {
|
|
linesCount = 0;
|
|
}
|
|
}
|
|
|
|
const datePart = file.replace('.csv', '').split('-part')[0];
|
|
|
|
results.push({
|
|
clientId: clientDir,
|
|
date: datePart,
|
|
filename: file,
|
|
filePath: fullFilePath,
|
|
sizeBytes: stats.size,
|
|
linesCount,
|
|
modifiedAt: stats.mtime,
|
|
});
|
|
}
|
|
}
|
|
|
|
for (const key of this.rowCountCache.keys()) {
|
|
if (!activePaths.has(key)) {
|
|
this.rowCountCache.delete(key);
|
|
}
|
|
}
|
|
} catch {
|
|
// Return whatever collected
|
|
}
|
|
|
|
return results.sort((a, b) => b.modifiedAt.getTime() - a.modifiedAt.getTime());
|
|
}
|
|
|
|
/**
|
|
* Non-blocking async listing using fs.promises + streaming row counting on cache misses
|
|
*/
|
|
public static async listAllFilesAsync(): Promise<StorageFileMeta[]> {
|
|
const results: StorageFileMeta[] = [];
|
|
if (!fs.existsSync(this.baseStorageDir)) {
|
|
this.rowCountCache.clear();
|
|
return results;
|
|
}
|
|
|
|
const activePaths = new Set<string>();
|
|
try {
|
|
const clientDirs = await fs.promises.readdir(this.baseStorageDir);
|
|
for (const clientDir of clientDirs) {
|
|
const fullClientPath = path.join(this.baseStorageDir, clientDir);
|
|
const dirStat = await fs.promises.stat(fullClientPath).catch(() => null);
|
|
if (!dirStat || !dirStat.isDirectory()) continue;
|
|
|
|
const files = await fs.promises.readdir(fullClientPath);
|
|
for (const file of files) {
|
|
if (!file.endsWith('.csv')) continue;
|
|
const fullFilePath = path.join(fullClientPath, file);
|
|
const stats = await fs.promises.stat(fullFilePath).catch(() => null);
|
|
if (!stats) continue;
|
|
activePaths.add(fullFilePath);
|
|
|
|
const cached = this.rowCountCache.get(fullFilePath);
|
|
let linesCount = 0;
|
|
if (
|
|
cached &&
|
|
cached.sizeBytes === stats.size &&
|
|
cached.mtimeMs === stats.mtimeMs
|
|
) {
|
|
linesCount = cached.linesCount;
|
|
} else {
|
|
try {
|
|
linesCount = await this.countCsvDataRowsStreamAsync(fullFilePath);
|
|
this.rowCountCache.set(fullFilePath, {
|
|
sizeBytes: stats.size,
|
|
mtimeMs: stats.mtimeMs,
|
|
linesCount,
|
|
});
|
|
} catch {
|
|
linesCount = 0;
|
|
}
|
|
}
|
|
|
|
const datePart = file.replace('.csv', '').split('-part')[0];
|
|
|
|
results.push({
|
|
clientId: clientDir,
|
|
date: datePart,
|
|
filename: file,
|
|
filePath: fullFilePath,
|
|
sizeBytes: stats.size,
|
|
linesCount,
|
|
modifiedAt: stats.mtime,
|
|
});
|
|
}
|
|
}
|
|
|
|
for (const key of this.rowCountCache.keys()) {
|
|
if (!activePaths.has(key)) {
|
|
this.rowCountCache.delete(key);
|
|
}
|
|
}
|
|
} catch {
|
|
// Return whatever collected
|
|
}
|
|
|
|
return results.sort((a, b) => b.modifiedAt.getTime() - a.modifiedAt.getTime());
|
|
}
|
|
|
|
/**
|
|
* Prune CSV partitions older than autoPruneDays without counting row contents
|
|
*/
|
|
public static pruneExpiredPartitions(autoPruneDays: number): number {
|
|
if (!Number.isFinite(autoPruneDays) || autoPruneDays <= 0) {
|
|
return 0;
|
|
}
|
|
if (!fs.existsSync(this.baseStorageDir)) {
|
|
return 0;
|
|
}
|
|
|
|
const cutoffDate = new Date(
|
|
Date.now() - Math.floor(autoPruneDays) * 86_400_000,
|
|
)
|
|
.toISOString()
|
|
.split('T')[0];
|
|
|
|
let prunedCount = 0;
|
|
try {
|
|
const clientDirs = fs.readdirSync(this.baseStorageDir);
|
|
for (const clientDir of clientDirs) {
|
|
const fullClientPath = path.join(this.baseStorageDir, clientDir);
|
|
if (!fs.statSync(fullClientPath).isDirectory()) continue;
|
|
|
|
const files = fs.readdirSync(fullClientPath);
|
|
for (const file of files) {
|
|
if (!file.endsWith('.csv')) continue;
|
|
const datePart = file.replace('.csv', '').split('-part')[0];
|
|
if (/^\d{4}-\d{2}-\d{2}$/.test(datePart) && datePart < cutoffDate) {
|
|
const fullPath = path.join(fullClientPath, file);
|
|
try {
|
|
if (fs.existsSync(fullPath)) {
|
|
fs.unlinkSync(fullPath);
|
|
this.rowCountCache.delete(fullPath);
|
|
prunedCount += 1;
|
|
}
|
|
} catch {
|
|
// Ignore file removal error
|
|
}
|
|
}
|
|
}
|
|
}
|
|
} catch {
|
|
// Ignore directory scan failure
|
|
}
|
|
|
|
return prunedCount;
|
|
}
|
|
|
|
private static formatStorageStats(files: StorageFileMeta[]) {
|
|
const totalBytes = files.reduce((acc, f) => acc + f.sizeBytes, 0);
|
|
const totalRows = files.reduce((acc, f) => acc + f.linesCount, 0);
|
|
const uniqueClients = new Set(files.map((f) => f.clientId)).size;
|
|
const rotationMb = Math.max(
|
|
1,
|
|
Math.round(this.maxFileSizeBytes / (1024 * 1024)),
|
|
);
|
|
|
|
return {
|
|
totalFiles: files.length,
|
|
totalBytes,
|
|
totalMB: (totalBytes / (1024 * 1024)).toFixed(2),
|
|
totalRows,
|
|
uniqueClients,
|
|
storageDirectory: this.baseStorageDir,
|
|
rotationMb,
|
|
rotationRule: `${rotationMb}MB / Daily`,
|
|
};
|
|
}
|
|
|
|
/**
|
|
* Get aggregate storage statistics (sync cached)
|
|
*/
|
|
public static getStorageStats() {
|
|
return this.formatStorageStats(this.listAllFiles());
|
|
}
|
|
|
|
/**
|
|
* Get aggregate storage statistics (non-blocking async)
|
|
*/
|
|
public static async getStorageStatsAsync() {
|
|
const files = await this.listAllFilesAsync();
|
|
return this.formatStorageStats(files);
|
|
}
|
|
|
|
/**
|
|
* Reads file content for export / download with path traversal protection
|
|
*/
|
|
public static getFileContent(clientId: string, filename: string): string | null {
|
|
const safeClientId = String(clientId || '').replace(/[^a-zA-Z0-9_-]/g, '_');
|
|
const safeFilename = path.basename(String(filename || ''));
|
|
if (!safeClientId || !safeFilename || !safeFilename.endsWith('.csv')) {
|
|
return null;
|
|
}
|
|
const fullPath = path.resolve(this.baseStorageDir, safeClientId, safeFilename);
|
|
const resolvedBase = path.resolve(this.baseStorageDir);
|
|
if (!fullPath.startsWith(resolvedBase + path.sep)) {
|
|
return null;
|
|
}
|
|
if (fs.existsSync(fullPath)) {
|
|
return fs.readFileSync(fullPath, 'utf8');
|
|
}
|
|
return null;
|
|
}
|
|
}
|
|
|
|
export default CSVStorageManager;
|