Encoder Module Architecture

This document describes the structure, responsibilities, and usage of the Encoder module in the communications framework.

---

🧱 Overview

The encoder module enables a clean, extensible, and efficient way to implement multiple channel coding schemes in the framework. It separates interface, implementation, and selection logic using a three-layer architecture:

[BaseEncoder] ⇨ [Encoder] ⇨ [DedicatedEncoder]

---

🧩 1. BaseEncoder (Abstract Interface)

Location: framework/base/encoder.py
Type: Abstract base class (ABC)

Purpose:

Defines the standard encoder interface, shared functionality, and enforces implementation of the encoding logic.

Responsibilities:

• Store input/output lengths (A, G)

• Validate inputs

• Provide the public .encode() method

• Require subclasses to implement _encode_np(...)

Signature:

class BaseEncoder(ABC):
    def __init__(A: int, G: int, name: str = "")
    def encode(info_bits: List[int]) -> List[int]
    def _encode_np(info_bits: np.ndarray) -> np.ndarray  # abstract
    def check_input_length(...)
    def reset()  # optional

---

🧩 2. Encoder (Factory Wrapper)

Location: tx/encoders/encoder.py
Type: Lightweight instantiator and delegate

Purpose:

Abstracts the logic for selecting the correct encoder based on code.type (e.g., "polar", "uncoded", "nr5g").

Responsibilities:

• Initialize the correct dedicated encoder

• Forward .encode() calls to the selected encoder

• Expose shared properties like A and G

Signature:

class Encoder:
    def __init__(code)
    def encode(info_bits: List[int]) -> List[int]
    @property def A
    @property def G

---

🧩 3. Dedicated Encoder Implementations

Examples:

• PolarEncoder → tx/encoders/polar_encoder.py

• UncodedEncoder → tx/encoders/uncoded_encoder.py

• PolarNR5GEncoder → schemes/nr5g/polar_nr5g_encoder.py

Purpose:

Each class contains the actual encoding algorithm.

Responsibilities:

• Inherit from BaseEncoder

• Implement _encode_np() for core logic

• May contain additional helper methods (e.g., CRC, polar matrix generation)

Template:

class MyEncoder(BaseEncoder):
    def __init__(self, code): ...
    def _encode_np(self, info_bits: np.ndarray) -> np.ndarray:
        # Do the actual encoding here

---

✅ Usage Example

from tx.encoders.encoder import Encoder

encoder = Encoder(code)  # Code object with .type, .len_k, .len_n, etc.
info_bits = [1, 0, 1, 1, 0, 0, 1]
encoded_bits = encoder.encode(info_bits)  # Output: [0, 1, 0, 1, ...]

---

✨ Design Highlights

Feature	Benefit
Abstract base class	Enforces consistent interface
Clean factory wrapper	No repeated “if code.type == …”
Easy to extend	Add new encoders without touching pipelines
Efficient in production	_encode_np() allows in-place optimization
Readable and testable	Each encoder is isolated and modular

---