Skip to content

Batch & Streaming Examples

QuantWave allows you to use the same logic for both batch processing and streaming.

Batch Processing (Polars)

import polars as pl
import quantwave  # registers pl.col().ta (bundled in PyPI wheel)

df = pl.read_parquet("data.parquet")
df = df.lazy().with_columns(
    pl.col("close").ta.rsi(timeperiod=14).alias("rsi"),
).collect()

use polars::prelude::*;
use quantwave_polars::TA;

let df = df.lazy().ta().rsi("close", 14).collect()?;

Streaming Processing

QuantWave indicators implement the universal Next<T> trait. This is the single source of mathematical truth for every indicator — the exact same logic used by the Polars expressions. As a result, streaming and batch results are guaranteed to be bit-identical.

import quantwave as qw

# Simple streaming example
Rsi = qw.streaming_class("rsi")
rsi = qw.wrap_streaming(Rsi(14), name="rsi")
for price in prices:
    val = rsi.next(price)
    if rsi.is_ready:
        print(val)

# SuperTrend streaming example
St = qw.streaming_class("supertrend")
st = qw.wrap_streaming(St(period=10, multiplier=3.0), name="supertrend")
for high, low, close in ohlcv_data:
    signal = st.next((high, low, close))
    if st.is_ready:
        print(signal)

use quantwave_core::indicators::{RSI, SuperTrend};
use quantwave_core::traits::Next;

// Simple streaming example
let mut rsi = RSI::new(14);
for price in prices {
    println!("{:?}", rsi.next(price));
}

// SuperTrend streaming example
let mut st = SuperTrend::new(10, 3.0);
for (high, low, close) in ohlcv_data {
    let signal = st.next((high, low, close));
    println!("{:?}", signal);
}

ML Features + Rich Backtest (Cross-Epic Canonical Example)

The complete demonstration of the 4ps/gwx integration contract (locked .ta().features() surface for batch signal prep + FeatureToSignal adapter implementing Next<&Bar, Output=StrategySignal> with rich metadata for run_streaming_simulation + verified parity + metadata in Trade.entry_metadata) is the runnable notebook:

  • docs/examples/notebooks/ml_feature_backtest_parity.py

It is the primary evidence artifact for closure of both epics. Run instructions and full source provenance are inside the notebook.