ima-claude 2.20.0 → 2.25.0

package/plugins/ima-claude/skills/py-fp/SKILL.md

@@ -5,158 +5,68 @@ description: "Python FP principles with anti-over-engineering focus - Simple > C
 
 # Python Functional Programming
 
-Core functional programming principles for Python with anti-over-engineering enforcement. Python is a multi-paradigm language with solid FP support — but it's not Haskell. This skill teaches you to be functional in spirit, Pythonic in expression.
+FP in spirit, Pythonic in expression. Python is not Haskell — don't fight the language.
 
-## When to Use This Skill
+## CRITICAL: Anti-Over-Engineering
 
-- Implementing pure, testable Python functions
-- Data transformation pipelines (pandas, polars, ML preprocessing)
-- Need FP architectural guidance for Python
-- Preventing over-engineering and custom FP utility creation
-- Comprehensive testing strategies with pytest
-- Evidence-based performance optimization
-
-## CRITICAL: Anti-Over-Engineering (PRIMARY FOCUS)
-
-**Core Principle**: "Simple > Complex | Evidence > Assumptions"
-
-> **Clarification**: This skill prevents CREATING custom FP utility functions (pipe, compose, curry, monads) to make Python "feel" like Haskell. Using established libraries (toolz, pandas, itertools, etc.) is perfectly fine. FP is a mindset — pure functions, immutability, composition — not a rigid API signature.
-
-### Don't Create Custom FP Utilities
+**Never create custom FP utilities** (`pipe`, `compose`, `curry`, custom monads). Use native patterns or established libraries (`toolz`, `itertools`, `functools`).
 
 ```python
-# DON'T CREATE: pipe() utility
+# DON'T: pipe() utility
 def pipe(value, *functions):
     for fn in functions:
         value = fn(value)
     return value
 
-# INSTEAD: Native function calls with early returns
+# DO: native early returns
 def validate_user(user_data: dict) -> dict:
-    required_check = validate_required(["email", "name"], user_data)
-    if not required_check["valid"]:
-        return required_check
-
-    email_check = validate_email(user_data)
-    if not email_check["valid"]:
-        return email_check
-
+    required = validate_required(["email", "name"], user_data)
+    if not required["valid"]:
+        return required
+    email = validate_email(user_data)
+    if not email["valid"]:
+        return email
     return validate_name_length(user_data)
 
-# DON'T CREATE: compose() utility
-def compose(*fns):
-    def composed(x):
-        for fn in reversed(fns):
-            x = fn(x)
-        return x
-    return composed
-
-# INSTEAD: Direct function calls
-def process_data(raw: dict) -> dict:
-    normalized = normalize(raw)
-    validated = validate(normalized)
-    return transform(validated)
-
-# DON'T CREATE: curry() utility
-# INSTEAD: functools.partial or closures
+# DON'T: compose() or curry() utilities
+# DO: direct calls or functools.partial
 from functools import partial
-
 validate_min_length = partial(validate_length, min_len=3)
 
-# Or closures for more complex cases
-def create_validator(rules: list) -> callable:
-    def validate(value):
-        errors = [r["message"] for r in rules if not r["check"](value)]
-        return {"valid": True} if not errors else {"valid": False, "errors": errors}
-    return validate
-
-# DON'T CREATE: Custom monads (Maybe, Either, IO)
-# INSTEAD: Native error handling with result dicts
+# DON'T: Custom monads (Maybe, Either, IO)
+# DO: result dicts
 def get_user(user_id: int, db) -> dict:
     try:
-        user = db.find(user_id)
-        return {"success": True, "data": user}
+        return {"success": True, "data": db.find(user_id)}
     except Exception as e:
         return {"success": False, "error": str(e)}
 ```
 
-### Context-Appropriate Complexity
-
-```python
-# CLI Script: Simple and direct
-def process_file(file_path: str) -> list[str]:
-    with open(file_path) as f:
-        return [line.upper() for line in f if line.strip()]
+Match complexity to context: CLI script ≠ production service ≠ data pipeline.
 
-# Production Service: Appropriate error handling
-def process_file(file_path: str, logger) -> dict:
-    try:
-        with open(file_path) as f:
-            lines = [line.strip() for line in f if line.strip()]
-        logger.info("File processed", extra={"path": file_path, "lines": len(lines)})
-        return {"success": True, "data": [line.upper() for line in lines]}
-    except OSError as e:
-        logger.error("File processing failed", extra={"path": file_path, "error": str(e)})
-        return {"success": False, "error": str(e)}
-
-# Data Pipeline: Composable transformations
-def create_file_processor(transforms: list[callable]):
-    compiled = [compile_transform(t) for t in transforms]
-
-    def process(file_path: str, logger) -> dict:
-        with open(file_path) as f:
-            data = f.read()
-        for transform in compiled:
-            data = transform(data)
-        return {"success": True, "data": data}
-
-    return process
-```
-
-## Core FP Patterns (Error-Preventing Essentials)
+## Core Patterns
 
 ### 1. Purity and Side Effect Isolation
 
-**Rule**: Separate business logic from side effects.
+Separate business logic from side effects.
 
 ```python
-# IMPURE - side effects mixed with logic
-def calculate_total(items: list[dict]) -> float:
-    print("Processing items")  # Side effect
-    global running_total
-    running_total += sum(i["price"] for i in items)  # Mutation
-    return running_total
-
-# PURE: business logic
+# Pure: business logic
 def calculate_total(items: list[dict]) -> float:
     return sum(item["price"] for item in items)
 
-# ISOLATED: side effects separate
+# Side effects separate
 def log_and_calculate(items: list[dict], logger) -> float:
-    total = calculate_total(items)  # Pure calculation
-    logger.info(f"Total: {total}")  # Side effect isolated
+    total = calculate_total(items)
+    logger.info(f"Total: {total}")
     return total
 ```
 
-**Benefits**:
-- 100% testable with all edge cases
-- Predictable behavior and debugging
-- Safe for `multiprocessing` (pure functions serialize trivially)
-- Enables memoization via `@lru_cache`
+Pure functions: 100% testable, `@lru_cache`-able, safe for `multiprocessing`.
 
 ### 2. Composition Over Inheritance
 
-**Rule**: Build complex behavior from simple functions.
-
 ```python
-# Class hierarchy approach (avoid)
-class BaseValidator:
-    def validate(self, value): raise NotImplementedError
-
-class EmailValidator(BaseValidator):
-    def validate(self, value): ...
-
-# Function composition (no utilities needed)
 def validate_required(value) -> bool:
     return value is not None and value != ""
 
@@ -166,7 +76,6 @@ def validate_email(value: str) -> bool:
 def validate_length(min_len: int, max_len: int):
     return lambda value: min_len <= len(value) <= max_len
 
-# Simple composition without pipe() utility
 def validate_user_email(email: str) -> dict:
     if not validate_required(email):
         return {"valid": False, "error": "Required"}
@@ -177,101 +86,67 @@ def validate_user_email(email: str) -> dict:
     return {"valid": True}
 ```
 
-### 3. Dependency Injection Through Parameters
-
-**Rule**: Pass dependencies explicitly, avoid global state.
+### 3. Dependency Injection via Parameters
 
 ```python
-# Hidden dependencies, hard to test
+# Hidden deps — untestable
 def save_user(user_data: dict) -> dict:
-    hashed = bcrypt.hash(user_data["password"])  # Hidden
-    return database.save({**user_data, "password": hashed})  # Hidden
+    hashed = bcrypt.hash(user_data["password"])
+    return database.save({**user_data, "password": hashed})
 
-# Explicit dependencies, fully testable
+# Explicit deps — fully testable
 def save_user(user_data: dict, hasher, database) -> dict:
     hashed = hasher.hash(user_data["password"])
     return database.save({**user_data, "password": hashed})
 
-# Factory for repeated use (closure-based DI)
+# Closure-based DI for repeated use
 def create_user_service(hasher, database):
     def save(user_data: dict) -> dict:
         return save_user(user_data, hasher, database)
-
     def find(user_id: int) -> dict:
         return database.find_by_id(user_id)
-
     return {"save": save, "find": find}
 ```
 
-### 4. Immutability Patterns
-
-**Rule**: Don't mutate input data. Create new structures.
+### 4. Immutability
 
 ```python
 from dataclasses import dataclass, replace, field
 from typing import NamedTuple
 
-# Frozen dataclasses — the primary immutable record
 @dataclass(frozen=True)
 class User:
     name: str
     email: str
     settings: dict = field(default_factory=dict)
 
-# "Update" via replace() — returns new instance
 def update_email(user: User, new_email: str) -> User:
-    return replace(user, email=new_email)
-
-# NamedTuple — lighter weight alternative
-class Point(NamedTuple):
-    x: float
-    y: float
-
-def translate(point: Point, dx: float, dy: float) -> Point:
-    return Point(point.x + dx, point.y + dy)
+    return replace(user, email=new_email)  # new instance
 
-# Dict immutability — spread with {**d}
+# Dict/list — spread, don't mutate
 def update_settings(user: dict, settings: dict) -> dict:
     return {**user, "settings": {**user.get("settings", {}), **settings}}
 
-# List operations without mutation
-def add_item(items: list, new_item) -> list:
-    return [*items, new_item]
-
-def remove_item(items: list[dict], item_id: int) -> list[dict]:
-    return [i for i in items if i["id"] != item_id]
-
 def update_item(items: list[dict], item_id: int, updates: dict) -> list[dict]:
     return [{**i, **updates} if i["id"] == item_id else i for i in items]
 ```
 
-> **Important**: `frozen=True` is shallow — it prevents attribute reassignment but doesn't freeze mutable attribute values (like dicts or lists inside). For deep immutability, use `pyrsistent` or keep nested values as tuples/frozensets.
+> `frozen=True` is shallow — nested dicts/lists remain mutable. Use `pyrsistent` or `MappingProxyType` for deep immutability.
 
 ## Python-Specific Patterns
 
-### Comprehensions Over map/filter (Pythonic FP)
+### Comprehensions (Prefer Over map/filter)
 
 ```python
-# map/filter style (less Pythonic)
-active_names = list(map(
-    lambda u: u["name"],
-    filter(lambda u: u["active"], users)
-))
-
-# Comprehension style (Pythonic FP)
+# Pythonic FP
 active_names = [u["name"] for u in users if u["active"]]
-
-# Dict comprehension for transformations
 name_by_id = {u["id"]: u["name"] for u in users}
-
-# Generator expression for lazy evaluation (large data)
 totals = sum(item["price"] for item in items if item["taxable"])
 ```
 
 ### Generators for Lazy Pipelines
 
 ```python
-# Generators are Python's answer to lazy evaluation
 def read_records(path: str):
     with open(path) as f:
         for line in f:
@@ -288,47 +163,34 @@ def transform(records):
 
 # Compose lazily — no intermediate lists
 pipeline = transform(filter_valid(read_records("data.csv")))
-results = list(pipeline)  # Materializes only when needed
 ```
 
-### itertools and functools (The FP Standard Library)
+### itertools / functools
 
 ```python
 from functools import partial, lru_cache, reduce
-from itertools import chain, groupby, islice, takewhile
-from operator import itemgetter
+from itertools import groupby
+from operator import itemgetter, mul
 
-# partial — the Pythonic way to "curry"
+# partial — Pythonic curry
 multiply_by_tax = partial(lambda rate, price: price * (1 + rate), 0.08)
 
-# lru_cache — memoize pure functions (MUST be pure!)
+# lru_cache — memoize pure functions only
 @lru_cache(maxsize=256)
 def fibonacci(n: int) -> int:
-    if n < 2:
-        return n
-    return fibonacci(n - 1) + fibonacci(n - 2)
-
-# operator module — eliminate trivial lambdas
-from operator import itemgetter, attrgetter
+    return n if n < 2 else fibonacci(n - 1) + fibonacci(n - 2)
 
+# operator — eliminate trivial lambdas
 sorted_by_name = sorted(users, key=itemgetter("name"))
-# vs: sorted(users, key=lambda u: u["name"])
-
-# itertools for complex transformations
-from itertools import groupby
 
 def group_by_category(items: list[dict]) -> dict:
     sorted_items = sorted(items, key=itemgetter("category"))
-    return {
-        k: list(v)
-        for k, v in groupby(sorted_items, key=itemgetter("category"))
-    }
+    return {k: list(v) for k, v in groupby(sorted_items, key=itemgetter("category"))}
 ```
 
-### Pattern Matching (Python 3.10+)
+### Pattern Matching (3.10+)
 
 ```python
-# Structural pattern matching for clean dispatch
 def process_event(event: dict) -> dict:
     match event:
         case {"type": "click", "target": target}:
@@ -339,48 +201,31 @@ def process_event(event: dict) -> dict:
             return handle_error(code, msg)
         case _:
             return {"error": f"Unknown event type: {event.get('type')}"}
-
-# Pattern matching with guards
-def categorize_score(score: int) -> str:
-    match score:
-        case n if n >= 90: return "A"
-        case n if n >= 80: return "B"
-        case n if n >= 70: return "C"
-        case _: return "F"
 ```
 
-### Type Hints for FP Clarity
+### Type Hints
 
 ```python
 from typing import TypeVar, Callable, Optional
-from collections.abc import Iterable, Iterator
+from collections.abc import Iterable
 
 T = TypeVar("T")
 U = TypeVar("U")
 
-# Type hints make pure function contracts explicit
 def transform_all(items: Iterable[T], fn: Callable[[T], U]) -> list[U]:
     return [fn(item) for item in items]
 
-# Union types (3.10+) for result patterns
-def divide(a: float, b: float) -> dict:
-    if b == 0:
-        return {"success": False, "error": "Division by zero"}
-    return {"success": True, "data": a / b}
-
-# Optional signals nullable return
 def find_user(users: list[dict], user_id: int) -> Optional[dict]:
     return next((u for u in users if u["id"] == user_id), None)
 ```
 
-## Data Science / ML Patterns
+## Data Science / ML
 
-### pandas pipe() for Transformation Chains
+### pandas pipe()
 
 ```python
 import pandas as pd
 
-# Each function: DataFrame in, DataFrame out (pure transforms)
 def clean_nulls(df: pd.DataFrame) -> pd.DataFrame:
     return df.dropna(subset=["email", "name"])
 
@@ -392,272 +237,121 @@ def add_age_group(df: pd.DataFrame) -> pd.DataFrame:
     labels = ["youth", "young_adult", "adult", "senior"]
     return df.assign(age_group=pd.cut(df["age"], bins=bins, labels=labels))
 
-# Compose via pipe() — FP pipeline, Pythonic syntax
-result = (
-    raw_df
-    .pipe(clean_nulls)
-    .pipe(normalize_names)
-    .pipe(add_age_group)
-)
+result = raw_df.pipe(clean_nulls).pipe(normalize_names).pipe(add_age_group)
 ```
 
-### Pure Feature Engineering
-
-```python
-# Pure transformation functions for ML pipelines
-def extract_features(df: pd.DataFrame) -> pd.DataFrame:
-    return df.assign(
-        word_count=df["text"].str.split().str.len(),
-        has_url=df["text"].str.contains(r"https?://", regex=True),
-        text_length=df["text"].str.len(),
-    )
-
-# scikit-learn FunctionTransformer for pipeline integration
-from sklearn.preprocessing import FunctionTransformer
-from sklearn.pipeline import Pipeline
-
-feature_extractor = FunctionTransformer(extract_features)
-
-pipeline = Pipeline([
-    ("features", feature_extractor),
-    ("scaler", StandardScaler()),
-    ("model", LogisticRegression()),
-])
-```
-
-### Polars (FP-Friendly Alternative to pandas)
+### Polars (FP-native)
 
 ```python
 import polars as pl
 
-# Polars expressions are lazy and composable by design
 result = (
-    pl.scan_csv("data.csv")  # Lazy — nothing runs yet
+    pl.scan_csv("data.csv")
     .filter(pl.col("status") == "active")
     .with_columns(
         full_name=pl.col("first_name") + " " + pl.col("last_name"),
-        age_group=pl.when(pl.col("age") < 30).then(pl.lit("young"))
-                    .otherwise(pl.lit("senior")),
+        age_group=pl.when(pl.col("age") < 30).then(pl.lit("young")).otherwise(pl.lit("senior")),
     )
     .group_by("department")
     .agg(pl.col("salary").mean().alias("avg_salary"))
-    .collect()  # Execute the entire plan at once
+    .collect()
 )
 ```
 
 ## Result Type Pattern
 
 ```python
-# Standard result shape — consistent across the codebase
 def success(data=None) -> dict:
     return {"success": True, "data": data}
 
 def failure(error: str) -> dict:
     return {"success": False, "error": error}
 
-# Chain results with early return
 def process_order(order: dict) -> dict:
     validated = validate_order(order)
     if not validated["success"]:
         return validated
-
     priced = calculate_pricing(validated["data"])
     if not priced["success"]:
         return priced
-
     return submit_order(priced["data"])
 ```
 
-## Testing Essentials (Enabled by Purity)
-
-**Philosophy**: Pure functions enable testing all edge cases systematically.
+## Testing
 
 ```python
 import pytest
 
-# Parametrized testing — the FP way
 @pytest.mark.parametrize("price,rate,expected", [
     (100.0, 0.1, 10.0),
     (50.0, 0.2, 10.0),
     (0.0, 0.1, 0.0),
-    (100.0, 0.0, 0.0),
 ])
 def test_calculate_discount(price, rate, expected):
     assert calculate_discount(price, rate) == expected
 
-# Edge case testing — pure functions handle all inputs
-@pytest.mark.parametrize("invalid_input", [None, "", [], {}, 0])
-def test_validate_required_rejects_invalid(invalid_input):
-    assert not validate_required(invalid_input)
-
-# Testing result chains
-def test_process_order_validation_failure():
-    result = process_order({"items": []})
-    assert not result["success"]
-    assert "empty" in result["error"].lower()
-
-# Frozen dataclass testability
 def test_update_email_returns_new_instance():
     user = User(name="Alice", email="old@test.com")
     updated = update_email(user, "new@test.com")
     assert updated.email == "new@test.com"
-    assert user.email == "old@test.com"  # Original unchanged
+    assert user.email == "old@test.com"  # original unchanged
 ```
 
-## Performance Patterns (Evidence-Based)
-
-**IMPORTANT**: Optimize only when needed with evidence.
-
-### Memoization for Expensive Pure Functions
+## Performance (Evidence-Based Only)
 
 ```python
-from functools import lru_cache
-
-# ONLY memoize pure functions with hashable arguments
+# Memoize pure functions with hashable args
 @lru_cache(maxsize=128)
 def parse_template(template: str) -> dict:
-    # Expensive parsing — cached after first call
     return heavy_parse(template)
 
-# For unhashable args, manually cache
-_cache = {}
-def get_config(key: str) -> dict:
-    if key not in _cache:
-        _cache[key] = load_config(key)
-    return _cache[key]
-```
-
-### Generator Pipelines for Large Data
-
-```python
-# Memory-efficient: processes one record at a time
+# Batch generator for large data
 def process_large_file(path: str):
-    records = read_records(path)       # Generator
-    valid = filter_valid(records)       # Generator
-    transformed = transform(valid)      # Generator
-
-    # Only materializes in batches
+    pipeline = transform(filter_valid(read_records(path)))
     batch = []
-    for record in transformed:
+    for record in pipeline:
         batch.append(record)
         if len(batch) >= 1000:
             yield batch
             batch = []
     if batch:
         yield batch
-```
-
-### multiprocessing with Pure Functions
-
-```python
-from multiprocessing import Pool
 
 # Pure functions parallelize trivially
-def process_item(item: dict) -> dict:
-    return {**item, "score": compute_score(item)}
-
-# No shared state, no locks, no bugs
+from multiprocessing import Pool
 with Pool(4) as pool:
     results = pool.map(process_item, items)
 ```
 
-## Python-Specific Gotchas
-
-### Recursion Limit
+## Gotchas
 
+**Recursion limit** — no tail call optimization. Use `reduce` for large inputs:
 ```python
-# Python has a 1000-call recursion limit. No tail call optimization.
-# DON'T: recursive algorithms for large inputs
-def factorial(n):
-    return 1 if n <= 1 else n * factorial(n - 1)  # Blows up at n > 1000
-
-# DO: iterative or reduce
 from functools import reduce
 from operator import mul
-
 def factorial(n: int) -> int:
     return reduce(mul, range(1, n + 1), 1)
 ```
 
-### Mutable Default Arguments
-
+**Mutable defaults** — never use mutable default args:
 ```python
-# NEVER use mutable defaults
-def add_item(item, items=[]):  # BUG: shared mutable default
-    items.append(item)
-    return items
-
-# INSTEAD: None sentinel
 def add_item(item, items=None):
-    items = items if items is not None else []
-    return [*items, item]  # Return new list, don't mutate
-```
-
-### Shallow vs Deep Copy
-
-```python
-# frozen=True is SHALLOW
-@dataclass(frozen=True)
-class Config:
-    settings: dict  # This dict is still mutable!
-
-config = Config(settings={"debug": True})
-# config.settings = {}  # FrozenInstanceError
-config.settings["debug"] = False  # This WORKS — shallow freeze
-
-# For deep immutability: use tuples, frozensets, or pyrsistent
-from types import MappingProxyType
-
-def freeze_dict(d: dict):
-    return MappingProxyType(d)  # Read-only view
+    return [*(items or []), item]
 ```
 
-## Quality Gates (Pre-Implementation Checklist)
-
-1. **"Can this be pure?"** - Separate business logic from side effects
-2. **"Can this use native patterns?"** - Comprehensions, generators, functools, itertools
-3. **"Can this be simplified?"** - Choose simple solution over complex abstraction
-4. **"Is this complexity justified?"** - Evidence-based complexity decisions
-5. **"Is this testable?"** - Pure functions enable comprehensive testing
-6. **"Are type hints used?"** - Type hints on all public function signatures
-
-## When to Load Reference Files
-
-### Deep Principles and Explanations
-**File**: `references/core-principles.md`
-**Load when**:
-- Learning mode or explaining WHY behind patterns
-- Making architectural decisions
-- Need complete Result Type patterns
-- Anti-pattern recognition details
-- Python-specific FP philosophy deep-dive
-
-### Testing Methodology
-**File**: `references/testing-patterns.md`
-**Load when**:
-- Building comprehensive test suites with pytest
-- Improving test coverage
-- Edge case analysis and boundary testing
-- Testing data pipelines and ML code
-- Property-based testing with Hypothesis
-
-### Working Examples
-**Directory**: `examples/`
-**Load when**:
-- Need complete working code
-- Integration examples
-- Learning implementation patterns
-
-## Integration with Domain Skills
-
-This core skill provides the foundation for:
+**Shallow freeze** — `frozen=True` doesn't freeze nested dicts. Use `MappingProxyType` or `pyrsistent` for deep immutability.
 
-- **py-fp-django**: Django patterns with FP principles (future)
-- **py-fp-fastapi**: FastAPI patterns with FP principles (future)
-- **py-fp-ml**: ML/data science patterns with FP principles (future)
+## Quality Gates
 
-Each domain skill references this core and adds domain-specific patterns.
+1. Can this be pure? — separate logic from side effects
+2. Can this use native patterns? — comprehensions, generators, functools, itertools
+3. Is this complexity justified? — evidence-based decisions only
+4. Are type hints used? — all public function signatures
 
-## Philosophy
+## References
 
-*"Pure functions, Pythonic patterns, type hints, appropriate complexity, and comprehensive testing for maintainable, predictable code. Be functional in spirit, Pythonic in expression — don't fight the language, work with it."*
+| Resource | Load when |
+|----------|-----------|
+| `references/core-principles.md` | Learning mode, architecture decisions, anti-pattern recognition |
+| `references/testing-patterns.md` | Building test suites, Hypothesis property-based testing |
+| `examples/` | Complete working code, integration examples |