Confirm parity between Python and Rust indicators

[cjdsellers]

In preparation for cutting over to the Rust implementations of indicators, a valuable task would be to go through each indicator one by one with an LLM and check (an LLM would be ideal for picking up potential discrepancies):

• Rust impl matches the Python impl
• Rust test coverage at least matches the Python test module

The Python/Cython implementations can be considered the source of truth.

[v9ai]

hey @cjdsellers may I look into this?

[v9ai]

Adaptive Moving Average (AMA) – Rust vs. Python/Cython

#	Topic	Rust implementation	Python / Cython implementation	Parity	Key notes & actions
1	Constructor arguments	period_efficiency_ratio, period_fast, period_slow, price_type (optional)	period_er, period_alpha_fast, period_alpha_slow, price_type (optional)	✅	Only the parameter names differ.
2	Parameter validation	None	Positive-int checks + period_alpha_slow > period_alpha_fast	⚠️	Add identical checks in Rust.
3	Doc comment for period_slow	Comment implies slow < fast	Docstring + runtime check require slow > fast	⚠️	Fix Rust comment to match intent.
4	Alpha constants	α_fast = 2/(fast+1) α_slow = 2/(slow+1) α_diff = α_fast – α_slow	Same maths	✅	Identical computation.
5	First-input handling	Initializes then returns	Initializes then continues (Δ = 0, so result is the same)	✅	Numerical behaviour matches.
6	Update counter	Field count, never incremented	_increment_count() each update	⚠️	Increment count in Rust or drop the field.
7	EfficiencyRatio price-type	Passes caller’s price_type to ER	ER always uses PriceType.LAST	⚠️	Choose one approach and align both.
8	Smoothing constant	sc = (ER·α_diff + α_slow)²	Same formula	✅	Identical.
9	initialized semantics	True once ER is ready	Same via base-class counter	✅	Equivalent readiness logic.
10	Reset methods	Two paths: trait reset() and extra const fn reset()	Single reset path	⚠️	Remove / merge duplicate Rust reset.
11	Unit-test breadth	Mirrors Python tests except count	Covers count	⚠️	Add Rust test for count and new validation.

---

TODO for Full Parity

1. Input validation in Rust (positive integers & slow > fast).
2. Correct Rust comment for period_slow.
3. Increment count (or deprecate) and test it.
4. Align ER price-type behaviour across both languages.
5. Consolidate reset logic in Rust.
6. Extend Rust tests for the above changes.

[v9ai]

Double Exponential Moving Average (DEMA) – Rust vs. Python/Cython

#	Topic	Rust implementation	Python / Cython implementation	Parity	Key notes & actions
1	Constructor arguments & defaults	period, price_type (optional, default = Last)	period, price_type (optional, default = PriceType.LAST)	✅	Same two parameters, only the enum path differs.
2	Parameter validation	None	Condition.positive_int(period) guard	⚠️	Add the same positive-integer check in Rust.
3	Inner EMA creation	ema1/ema2 = ExponentialMovingAverage::new(period, price_type)	ExponentialMovingAverage(period) (no price_type)	⚠️	Decide whether price_type matters for inner EMAs and align both sides.
4	First-input handling	Explicit branch initializes value and sets has_inputs = true	No branch—formula collapses to the raw value on the first update	✅	Numerical results are identical; stylistic difference only.
5	Update counter	Field count incremented every update; used for initialized	No public counter; uses inner EMA readiness	⚠️	Either expose _count in Python or drop count from Rust.
6	Initialization rule	initialized = count ≥ period	initialized when second EMA is ready	✅	Both trigger after period samples.
7	Reset behaviour	Resets own fields but not ema1/ema2	_reset_ma() resets both inner EMAs and own fields	⚠️	Call ema1.reset() / ema2.reset() inside Rust reset().
8	has_inputs semantics	Manual has_inputs boolean	Uses base-class _set_has_inputs(True)	✅	Equivalent semantics; just implemented differently.
9	Trade-tick price type	Ignores price_type (trade.price always used)	Same behaviour	✅	Consistent across languages.
10	Unit-test breadth	Thorough tests but missing reset-inner-EMA case	Relies on MovingAverage suite	⚠️	Extend Rust tests for (i) invalid period, (ii) inner-EMA reset, (iii) count parity.

---

TODO for Full Parity

1. Add runtime validation in Rust (period > 0).
2. Decide on inner-EMA price_type and apply consistently.
3. Reset inner EMAs inside the Rust reset() method.
4. Unify the count concept (either expose it in Python or remove it from Rust).
5. Expand Rust tests to cover the new validation, inner-EMA reset, and chosen count semantics.

[v9ai]

Exponential Moving Average (EMA) – Rust vs. Python/Cython

#	Topic	Rust implementation	Python / Cython implementation	Parity	Key notes & actions
1	Constructor arguments & defaults	new(period, price_type=None) – price_type defaults to PriceType::Last	__init__(period, price_type=PriceType.LAST) – identical signature & default	✅	Same public API, only the enum path differs (PriceType::Last vs. PriceType.LAST).
2	Parameter validation	None – any period accepted	Condition.positive_int(period) guard enforces period > 0	⚠️	Add the same positive‐integer check in Rust.
3	Alpha constant	α = 2 / (period + 1) computed once in new	Same formula in __init__	✅	Identical maths.
4	First-input handling	Branch sets has_inputs = true and seeds value before EMA formula	Branch seeds value; base class owns has_inputs flag	✅	Numerical behaviour matches.
5	Update counter	Explicit count field incremented every update	_increment_count() in base class	⚠️	Decide: expose count in Python or drop count from Rust for symmetry.
6	Initialisation rule	initialized = (count ≥ period) in update_raw	Base class uses the same rule	✅	Equivalent readiness logic.
7	has_inputs semantics	Manual boolean toggled after first sample	Base class sets flag inside _increment_count()	✅	Implementation differs but external semantics are the same.
8	Quote/price-type handling	handle_quote passes self.price_type to extract_price	Same behaviour in handle_quote_tick	✅	Consistent treatment of PriceType.
9	Reset behaviour	reset() zeros value, count, flags	Inherits MovingAverage.reset() which resets the same state	✅	Functionally equivalent.
10	Unit-test breadth	Dedicated tests cover ctor, update path, all handlers, reset, display	Relies on generic MovingAverage suite – no EMA-specific edge-case tests	⚠️	Add Python tests for (i) invalid period, (ii) first-tick seeding, (iii) parity with Rust values.

---

TODO for Full Parity

1. Runtime validation in Rust – reject period ≤ 0 with a clear error, mirroring Condition.positive_int.
2. Counter symmetry – either expose _count in Python (public read-only) or remove count from Rust if it is purely internal.
3. Extra Python tests
   • invalid period raises ValueError;
   • first-tick seeding vs. formula collapse;
   • exact numeric parity with Rust for a known input series.