Pipeline

Pipeline

This page describes the end-to-end flow from CLI arguments to files on disk.

High-level flow

Parse CLI → Fetch OHLCV → Normalize schema → Compute indicators → Write outputs

Deep dives

• CLI Parsing
• Data Source (Yahoo Finance)
• Source Data Deep Dive
• Schema Normalization
• Data Shape Invariants
• Config Resolution
• Polars Engine
• Output Writing
• Write Output Deep Dive

1) Parse CLI

Inputs come from CLI Reference:

• Tickers: positional args or a .txt file
• Period: a single value or a comma-separated list
• Timeframe: a single interval or a JSON mapping periods→intervals
• Optional overrides: config JSON, output naming, output format, engine

2) Fetch OHLCV (Yahoo Finance)

Data is fetched using yfinance as a combined history request and then split per-ticker.

Key behaviors:

• Missing tickers are skipped if no data is returned
• The tool expects typical Yahoo Finance columns like Date/Open/High/Low/Close/Volume

See Data Source (Yahoo Finance).

3) Normalize schema

The tool lowercases column names (e.g. Close → close) to make indicator expressions consistent.

4) Compute indicators

Indicators are computed with rolling windows and exponential moving means.

Important output naming notes (current implementation):

• ATR is uppercase (not atr)
• Stochastic columns are uppercase K and D

See Indicators (Overview).

5) Write outputs

The tool writes one output file per computed dataset, either to:

• an explicit --output name (or outputs list), or
• an auto-generated name: <ticker>_<period>_<timeframe>.<ext>

See Output Formats and Advanced Usage.

Common failure modes

• Yahoo Finance returns no data for a symbol → ticker skipped
• Permission issues writing to directory → writer error
• Very fine-grained intervals with large periods → large downloads / memory pressure

Related pages:

• Troubleshooting
• Performance