QuanTAlib/lib/feeds/IFeed.md

IFeed Interface

Property	Value
Category	Feed
Inputs	OHLCV bar (TBar)
Parameters	None
Outputs	Single series (IFeed)
Output range	Varies (see docs)
Warmup	1 bar

TL;DR

• IFeed defines the standard contract for all data feeds in QuanTAlib, ensuring consistent behavior across different data sources (synthetic, file-...
• No configurable parameters; computation is stateless per bar.
• Output range: Varies (see docs).
• Requires 1 bar of warmup before first valid output (IsHot = true).
• Validated against TA-Lib, Skender, and Tulip reference implementations where available.

IFeed defines the standard contract for all data feeds in QuanTAlib, ensuring consistent behavior across different data sources (synthetic, file-based, or live API).

Key Concepts

• Bidirectional Control: The Next(ref bool isNew) method allows the consumer to request a new bar (isNew = true) or an update to the current bar (isNew = false).
• Streaming: Designed for bar-by-bar processing, simulating real-time data flow.
• Batching: Supports fetching historical data ranges via Fetch().

Interface Definition

public interface IFeed
{
    /// <summary>
    /// Gets the next bar with full control over new/update state.
    /// </summary>
    TBar Next(ref bool isNew);

    /// <summary>
    /// Convenience overload for simple next-bar requests.
    /// </summary>
    TBar Next(bool isNew = true);

    /// <summary>
    /// Retrieves a batch of historical bars.
    /// </summary>
    TBarSeries Fetch(int count, long startTime, TimeSpan interval);
}

Implementation Guidelines

When implementing IFeed:

1. State Management: Maintain the current position in the data source.
2. End of Data: When data is exhausted, Next should return the last valid bar and set isNew to false.
3. Intra-bar Updates: If the source supports it (e.g., live ticks), Next(isNew: false) should return the updated state of the current bar. If not supported (e.g., CSV), it should return the current bar unchanged.
4. Thread Safety: Implementations are generally not required to be thread-safe unless specified.

Implementations

• GBM: Geometric Brownian Motion generator (Synthetic).
• CsvFeed: Reads OHLCV data from CSV files (Historical).