MacdExt

MACD Extended (MACDEXT) — MACD with an independently selectable moving-average type (SMA / EMA / WMA / DEMA / TEMA / TRIMA) for each of the fast, slow and signal lines.

Quick reference

Field	Value
Family	Trend & Directional
Input type	f64 (single close)
Output type	MacdOutput { macd, signal, histogram }
Output range	unbounded; histogram = macd − signal
Default parameters	none — all three periods and all three MA types are required
Warmup period	slow.warmup_period() + signal.warmup_period() (>= 26 at the classic lengths)
Interpretation	A generalised MACD: swap the EMA for any period-only moving average per line.

Formula

fast_ma   = MA_fast(close, fast_period)      // type = fast_type
slow_ma   = MA_slow(close, slow_period)      // type = slow_type
macd      = fast_ma − slow_ma
signal    = MA_signal(macd, signal_period)   // type = signal_type
histogram = macd − signal

Each of the three lines uses a separately chosen moving-average type. The selectable types are the period-only moving averages, matching TA-Lib's MA_Type codes 0..=5:

MaType	TA-Lib code	Moving average
Sma	0	Simple
Ema	1	Exponential
Wma	2	Weighted
Dema	3	Double exponential
Tema	4	Triple exponential
Trima	5	Triangular

(TA-Lib's KAMA / MAMA / T3 take extra shape parameters and are not selectable here; codes 6..=8 are rejected.) With all three types set to Ema and periods 12, 26, 9, MacdExt reproduces the classic MACD. See crates/wickra-core/src/indicators/macd_ext.rs.

Parameters

Name	Type	Default	Valid range	Description	Source
fast / slow / signal	usize	none	>= 1, fast < slow	The three periods. Any zero errors with Error::PeriodZero; fast >= slow errors with Error::InvalidPeriod.	macd_ext.rs:158
fast_type / slow_type / signal_type	MaType	none	Sma..=Trima	Moving-average type per line. MaType::from_code rejects codes outside 0..=5 with Error::InvalidPeriod.	macd_ext.rs:40

Inputs / Outputs

From crates/wickra-core/src/indicators/macd_ext.rs:

use wickra::{Indicator, MacdExt, MaType};
use wickra::MacdOutput;
// MacdExt: Input = f64, Output = MacdOutput
const _: fn(&mut MacdExt, f64) -> Option<MacdOutput> = <MacdExt as Indicator>::update;

In Python the constructor takes the three periods interleaved with their TA-Lib MA-type codes — MacdExt(fast, fast_matype, slow, slow_matype, signal, signal_matype); update returns a (macd, signal, histogram) tuple and batch an (n, 3) numpy.ndarray. Node mirrors this and returns a { macd, signal, histogram } object from update.

Warmup

MacdExt::new(...).warmup_period() is slow.warmup_period() + signal.warmup_period(), so it depends on the chosen periods and types and is at least 26 at the classic lengths. The unit test accessors_and_metadata pins warmup_period() >= 26.

Edge cases

• Histogram identity holds for every MA type. For all six types the emitted histogram equals macd − signal to floating-point tolerance. The unit test every_ma_type_produces_a_consistent_histogram pins this.
• Mixed types per line. Each line can use a different MA type (e.g. WMA fast, SMA slow, EMA signal). The unit test mixed_ma_types_per_line pins this.
• Code mapping. MaType::from_code(0..=5) maps to the six types; 6 and above error. The unit test from_code_maps_all_supported_types pins this.
• Invalid periods. A zero period or fast >= slow is rejected at construction. The unit test rejects_invalid_periods pins this.
• Constant series. Once warmed, a flat price gives (0, 0, 0) for any MA type. See the example below.

Examples

Rust

use wickra::{BatchExt, Indicator, MacdExt, MaType};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Classic 12/26/9 but with a WMA fast line and an SMA signal line.
    let mut m = MacdExt::new(12, MaType::Wma, 26, MaType::Ema, 9, MaType::Sma)?;
    let out = m.batch(&[100.0; 60]);
    println!("{:?}", out.last().unwrap());
    Ok(())
}

Output:

Some(MacdOutput { macd: 0.0, signal: 0.0, histogram: 0.0 })

On a constant series every moving average converges to the constant, so macd = 0, signal = 0, histogram = 0 regardless of the chosen types. On real data the lines diverge; histogram = macd − signal always holds.

Python

import numpy as np
import wickra as ta

# fast=12 WMA(2), slow=26 EMA(1), signal=9 SMA(0)
m = ta.MacdExt(12, 2, 26, 1, 9, 0)
out = m.batch(np.full(60, 100.0))   # (60, 3) array
print(out[-1])

Output:

[0. 0. 0.]

Node

const ta = require('wickra');
// fast=12 WMA(2), slow=26 EMA(1), signal=9 SMA(0)
const m = new ta.MacdExt(12, 2, 26, 1, 9, 0);
let last = null;
for (let i = 0; i < 60; i++) last = m.update(100);
console.log(last);

Output:

{ macd: 0, signal: 0, histogram: 0 }

Interpretation

MacdExt is the MACD for traders who want to control its smoothing character. Choosing a faster MA type (WMA, DEMA, TEMA) on the fast/slow lines reduces lag at the cost of more whipsaw; a TRIMA or SMA signal line smooths the trigger. Otherwise read it exactly like the classic MACD — macd/signal crossovers, zero-line crossovers, and histogram momentum. Set all three types to Ema with 12, 26, 9 to recover MacdFix / the standard MACD.

Common pitfalls

• Unsupported MA codes. Only codes 0..=5 (the period-only averages) are valid; KAMA / MAMA / T3 (6..=8) are rejected because they need extra parameters.
• fast >= slow. The fast period must be strictly shorter than the slow period, or construction fails.
• Comparing warmups across configs. Warmup depends on the slow and signal types (DEMA/TEMA need longer seeding than SMA), so two MacdExt instances can start emitting at different bars.

References

Gerald Appel's MACD (1979); the selectable-MA generalisation matches TA-Lib's MACDEXT and its MA_Type codes.

See also

• Indicator-MacdIndicator — the fully configurable EMA MACD.
• Indicator-MacdFix — the fixed 12/26 MACD.
• Indicators-Overview — the full taxonomy.