How I Parsed Your Data – CSV Transaction Parser

A smart and robust CSV parser designed to automatically extract banking/financial transactions from CSV files of any bank or institution, even when the format is unknown or non-standard.

The class name (HowIParsedYourDataAlgorithm) is a tribute to Ted Mosby's iconic line from How I Met Your Mother:
"Kids, I'm going to tell you an incredible story — the story of how I parsed your data."

Purpose

Many users receive bank statements or transaction lists in CSV format with varying columns, names in different languages, and inconsistent date/amount formats. This parser uses advanced heuristics to:

• Automatically detect the start of the data table (skipping headers, notes, or legends)
• Map the correct columns for date, amount, currency, and description/title
• Handle multiple fallbacks and global search when primary fields are ambiguous or empty
• Support dozens of date formats, amount formats (with . or , separators and thousands), and currency symbols
• Automatically determine whether a transaction is an income or expense

Key Features

• Automatic detection of the header row in the data table
• Intelligent scoring based on:
  • Keywords in over 5 languages (Italian, English, German, French, Spanish)
  • Regex patterns for dates, amounts, and currencies
  • Analysis of sample values from the first rows
• Fallback management (up to 3 alternatives per field)
• Robust date parsing (ISO, DD/MM/YYYY, MM-DD-YYYY, DD.MM.YYYY, Italian textual months like "15 gen 2025")
• Amount normalization (removing currency symbols, handling comma/dot, thousands separators)
• Currency recognition (EUR, USD, GBP, JPY, etc.) from code or symbol
• Structured output with typed transactions and totals summary

Installation

npm install papaparse

Usage

importHowIParsedYourDataAlgorithmfrom"./src/TransactionParser";// Read the CSV file content as a stringconstcsvContent=awaitfetch("/path/to/bank-statement.csv").then((r)=>r.text());constparser=newHowIParsedYourDataAlgorithm();constresult=awaitparser.parseCSVFile(csvContent);console.log("Parsed transactions:",result.transactions.length);console.log("Column mapping:",result.mapping);console.log("Confidence:",result.mapping.confidence);console.log("Total income:",result.summary.totalIncome);console.log("Total expenses:",result.summary.totalExpenses);console.log("Errors:",result.errors);

Example Output (IValueReturned)

{
transactions: IReadedTransaction[];
mapping: IFieldMapping;
errors: string[];
detectedStartRow?: number;
summary: {
totalTransactions: number;
totalIncome: number;
totalExpenses: number;}}

Transaction Structure (IReadedTransaction)

{
date: string;// ISO string
amount: number;// absolute value
action: TransactionAction;
currency: Currencies;
title: string;// description
originalRow: Record<string,any>;// original row}

Supported Languages & Formats

Dates: ISO, DD/MM/YYYY, MM/DD/YYYY, DD.MM.YYYY, DD Mon YYYY (Italian: "15 gen 2025")
Amounts: with/without currency symbol, decimal separator . or ,, thousands separator
Currencies: EUR, USD, GBP, JPY, CHF, CAD, AUD + symbols (€ $ £ ¥ ₹ ₽)
Descriptions: over 100 keywords in Italian, English, German, French, Spanish

Contributing

Contributions are welcome! You can:

• Add new date/amount patterns
• Expand keyword lists for additional languages
• Improve scoring or fallback logic
• Add tests with real (anonymized) CSV samples

License

MIT