TickerTensor

Multi-Ticker CSV Ingestion → Aligned N×M Price Matrix + Moving-Window Stats

PROJECT OVERVIEW

TickerTensor is a small, production-friendly Node.js tool that ingests multiple equity CSVs and converts them into a clean, chronologically aligned N×M matrix of closing prices—great for quant research and backtests. It includes a functional, loop-free moving-window statistics (mean, standard deviation, peak/trough counts) over overlapping and non-overlapping windows and prints concise summaries.

KEY FEATURES 🔑

• Robust CSV ingestion.
  • Handles comma/semicolon/tab delimiters, header aliases (Close/Last, Adj Close, etc.), $ and thousands separators, and date normalization to ISO (YYYY-MM-DD).
• Deterministic alignment
  • Intersects trading dates across all tickers to produce a rectangular N×M matrix (rows = dates, columns = tickers).
• Fast assembly (O(N×M))
  • Uses per-ticker Map(date → close) for O(1) lookups while building the matrix.
• Moving-window statistics
  • Functional JS (only map, reduce, filter, slice, Array.from; no for/while).
  • Computes mean, standard deviation (population by default; sample optional), and strict local peaks/troughs (endpoints ignored).
  • Overlapping windows (step = 1 or 5) and non-overlapping (step = window size).
  • Window sizes: 50, 200, 2000 (or custom via CLI for small samples).
• Readable CLI output
  • Prints a concise recap (start/end dates, companies, points per series) and a formatted table preview (first/last rows).

TECHNICAL STACK 🧱

• Language/Runtime: JavaScript (ESM), Node.js

• Core Modules: fs, path, url

• Testing: Jasmine 5

• Tooling: ESLint (flat config), VS Code, Git/GitHub

• CI (optional): GitHub Actions

ROBUSTNESS & EDGE CASES

• Delimiters: comma / semicolon / tab

• Headers: Close, Adj Close, Close/Last, Adjusted Close, Last

• Prices: strip $ and , safely before Number()

• Dates: handle YYYY-MM-DD, M/D/YYYY, and other JS-parsable formats → ISO

TESTING STRATEGY 🧪

Run with npm test.

Highlights (Jasmine):

• Matrix shape checks: matrix.length === dates.length, each row matches tickers.length

• Date ordering: every dates[i] >= dates[i-1].

• Real-data sanity test (if data/real available): non-empty intersection and rectangularity.

PERFORMANCE NOTES 🗒️

• Avoids quadratic .find lookups by pre-building Map(date → close) per ticker.

• Memory footprint tied to dataset size; parsing filters invalid rows early.

• Deterministic sorting ensures repeatable results.

DATA & PRIVACY ⚙️

• Ship small sample CSVs only.

• Keep large/third-party datasets local: add data/real/ to .gitignore.

• No external network calls; everything runs on local CSV files.

Contributors

• Kelvin Ihezue