【GPT-5-Codex】OpenAI発の最新コーディングモデルの概要・性能・使い方を実例で徹底解説

# shopping_cart.py

class Cart:
    def __init__(self):
        self.items = []  # (price, qty, tax_rate, discount_pct)

    def add(self, price: float, qty: int = 1, tax_rate: float = 0.1, discount_pct: float = 0.0):
        self.items.append((price, qty, tax_rate, discount_pct))

    def total(self) -> float:
        # BUG: 行単位での丸めをせず合算して最後に丸めている
        total = 0.0
        for price, qty, tax, disc in self.items:
            line = price * qty
            line = line * (1 + tax)
            if disc:
                line = line * (1 - disc / 100)
            total += line
        return round(total, 2)

# test_cart.py

def test_line_rounding_then_sum():
    from shopping_cart import Cart
    c = Cart()
    # 値と税率・割引率の組み合わせで「行丸め」と「最後丸め」がズレるケース
    c.add(58.97, 8, 0.05, 0.0)
    c.add(53.78, 2, 0.08, 5.0)
    c.add(69.28, 9, 0.08, 5.0)
    c.add(70.16, 1, 0.10, 0.0)
    assert c.total() == 1322.62  # 期待: 行ごとに2桁丸め→合計

リポジトリに shopping_cart.py と tests/test_cart.py を置きました。
まずテストを実行して失敗を再現してください。
行単位の四捨五入（2桁）→合計、という業務要件に合わせて shopping_cart.py を修正し、テストをグリーンにしてください。
なぜ1セント誤差が生じたかを、丸め順序・小数演算の観点で短く解説してください。
変更点の要約と、わかりやすいコミットメッセージを提案してください。

• Explored
  └ List ls
    List test

• Updated Plan
  └ □ 実装前にテストを実行し失敗を確認する
    □ shopping_cart.py を要件に合わせて修正する
    □ テストを再実行し、結果を確認する
    □ 調査結果・変更点・コミットメッセージ案をまとめる

• Ran pytest

• Explored
  └ Read shopping_cart.py
    Search find . -maxdepth 2 -type f
    Read test_cart.py

• Ran python -c "import shopping_cart; print(shopping_cart.__file__)"

• Ran python -m pytest

• Updated Plan
  └ pytest 実行で 1 セントの丸め誤差による失敗を再現しました。
    ✔ 実装前にテストを実行し失敗を確認する
    □ shopping_cart.py を要件に合わせて修正する
    □ テストを再実行し、結果を確認する
    □ 調査結果・変更点・コミットメッセージ案をまとめる

> Thinking
  I'm figuring out how to modify shopping_cart.py so each line total is rounded
  to two decimals using Decimal for precise rounding before summing. This should
  avoid float binary issues and meet the requirement of per-line rounding with
  typical rounding rules. I'll draft a plan to convert inputs to Decimal, apply
  tax and discount, quantize each line, then sum as Decimal, finally returning a
  float rounded properly. I'll add comments explaining the new rounding behavior
  and carefully handle conversions from floats to Decimals to prevent precision
  errors.

• Edited shopping_cart.py (+17 -6)
    1     +from decimal import Decimal, ROUND_HALF_UP
    2     +
    3     +
    4     +_TWOPLACES = Decimal("0.01")
    5     +
    6     +
    7     +def _to_decimal(value: float) -> Decimal:
    8     +    """Convert float-like inputs via str to avoid binary rounding art
           efacts."""
    9     +    return Decimal(str(value))
    10    +
    11    +
    12     class Cart:
    ⋮
    19         def total(self) -> float:
    9     -        # BUG: 行単位での丸めず合算して最後に丸めている
    10    -        total = 0.0
    20    +        total = Decimal("0")
    21             for price, qty, tax, disc in self.items:
    12    -            line = price * qty
    13    -            line = line * (1 + tax)
    22    +            line = _to_decimal(price) * Decimal(qty)
    23    +            line *= Decimal("1") + _to_decimal(tax)
    24                 if disc:
    15    -                line = line * (1 - disc / 100)
    25    +                line *= Decimal("1") - _to_decimal(disc) / Decimal("1
           00")
    26    +            line = line.quantize(_TWOPLACES, rounding=ROUND_HALF_UP)
    27                 total += line
    17    -        return round(total, 2)
    28    +        return float(total.quantize(_TWOPLACES, rounding=ROUND_HALF_U
           P))

• Updated Plan
  └ 小数演算を Decimal 化し、行ごとに 2 桁丸めた後に合計するように修正しました。
    ✔ 実装前にテストを実行し失敗を確認する
    ✔ shopping_cart.py を要件に合わせて修正する
    □ テストを再実行し、結果を確認する
    □ 調査結果・変更点・コミットメッセージ案をまとめる

• Ran python -m pytest

• Updated Plan
  └ pytest の実行でテストがグリーンになりました。
    ✔ 実装前にテストを実行し失敗を確認する
    ✔ shopping_cart.py を要件に合わせて修正する
    ✔ テストを再実行し、結果を確認する
    □ 調査結果・変更点・コミットメッセージ案をまとめる

• Ran git diff

• Updated Plan
  └ 修正内容と原因の整理ができました。
    ✔ 実装前にテストを実行し失敗を確認する
    ✔ shopping_cart.py を要件に合わせて修正する
    ✔ テストを再実行し、結果を確認する
    ✔ 調査結果・変更点・コミットメッセージ案をまとめる

> Thinking
  I'm putting together a clear summary of changes made to shopping_cart.py,
  including line references for imports and helper functions, explaining the
  1-cent rounding discrepancy due to binary float accumulation versus per-
  line Decimal rounding, noting tests run with pytest, and suggesting a concise
  commit message. I'll format this with concise headers and bullet points to
  keep everything clear and organized.

> Changes

  - Reworked totals to convert inputs to Decimal and round each line to two
  decimals with ROUND_HALF_UP before summing, ensuring the returned total is
  quantized as required (shopping_cart.py:1, shopping_cart.py:14).

  Cause

  - The original code added unrounded float line totals and only rounded once at
  the end; accumulating binary floating-point artifacts meant the final rounding
  lost 1 cent compared with rounding each line first.

  Tests

  - python -m pytest

  Commit

  - Round cart totals with per-line Decimal rounding

# feature.md
# Email Normalization Spec
- 入力は任意の大文字小文字/全角半角/前後空白を許す文字列
- 出力は小文字、前後空白除去、Unicode正規化(NFKC)
- ドメインはIDNA（punycode）を正規化
- Gmailのみ:
  - ローカル部の '.' は無視
  - '+' 以降のタグは除去
- ただし "gmail.com" と "googlemail.com" は等価として扱う
- 不正な形式は ValueError

# email_utils.py
# 実装は未完成なのでCodexに書かせる
def normalize_email(s: str) -> str:
    """
    要件:
      - 前後空白除去・小文字化・Unicode正規化(NFKC)
      - ドメインのIDNA正規化
      - Gmail系は '.' 無視と '+' タグ除去
      - 形式不正は ValueError
    """
    raise NotImplementedError

仕様は feature.md、コードの骨子は email_utils.py にあります。
仕様を満たすように email_utils.py を実装してください。
pytest のユニットテストを作成
GitHub ActionsのCI定義
Gmail/Outlook/Yahooの挙動差について注意書きを追加
仕上げに、テストで固定したい条件を3つ列挙してください。

• Explored
  └ List ls
    List ls
    Read feature.md, email_utils.py

• Updated Plan
  └ Outlining multi-part work: implement function, create tests, docs, CI, and
    invariants list.
    □ Understand spec and plan implementation and tests for email normalization
    □ Implement normalize_email with required behavior and documentation
    □ Add pytest unit tests (including boundary/property tests) and Hypothesis if
      feasible
    □ Add README fragment covering usage/design notes and provider behavior
      differences
    □ Add GitHub Actions workflow for pytest on Python 3.11/3.12
    □ List invariants to enforce via tests

• Updated Plan
  └ ✔ Understand spec and plan implementation and tests for email normalization
    □ Implement normalize_email with required behavior and documentation
    □ Add pytest unit tests (including boundary/property tests) and Hypothesis if
      feasible
    □ Add README fragment covering usage/design notes and provider behavior
      differences
    □ Add GitHub Actions workflow for pytest on Python 3.11/3.12
    □ List invariants to enforce via tests

• Edited email_utils.py (+59 -9)
    1     -# 実装は未完成なのでCodexに書かせる
    2     -def normalize_email(s: str) -> str:
    1     +"""Utility helpers for email-address normalization."""
    2     +
    3     +from __future__ import annotations
    4     +
    5     +import re
    6     +import unicodedata
    7     +from typing import Final
    8     +
    9     +_ADDRESS_PATTERN: Final[re.Pattern[str]] = re.compile(r"^[^@\s]+@[^@\s]+$
           ")
    10    +_GMAIL_DOMAINS: Final[frozenset[str]] = frozenset({"gmail.com", "googlema
           il.com"})
    11    +
    12    +
    13    +def normalize_email(address: str) -> str:
    14    +    """Normalize an email address according to the project specification.
    15    +
    16    +    The normalization pipeline performs Unicode normalization (NFKC), tri
           ms
    17    +    surrounding whitespace, lower-cases the address, converts the domain
           to its
    18    +    IDNA/ASCII representation, and applies Gmail-specific canonicalizatio
           n.
    19    +
    20    +    Args:
    21    +        address: User-provided email address that may contain arbitrary c
           asing
    22    +            and spacing.
    23    +
    24    +    Returns:
    25    +        The canonicalized email address string.
    26    +
    27    +    Raises:
    28    +        TypeError: If *address* is not a string.
    29    +        ValueError: If the input cannot be parsed as an email address or
           if the
    30    +            resulting local part is empty after Gmail canonicalization.
    31         """
    4     -    要件:
    5     -      - 前後空白除去・小文字化・Unicode正規化(NFKC)
    6     -      - ドメインのIDNA正規化
    7     -      - Gmail系は '.' 無視と '+' タグ除去
    8     -      - 形式不正は ValueError
    9     -    """
    10    -    raise NotImplementedError
    32    +
    33    +    if not isinstance(address, str):
    34    +        raise TypeError("address must be a str")
    35    +
    36    +    # Normalize Unicode before trimming; normalization can introduce ASCI
           I
    37    +    # whitespace that should be stripped.
    38    +    normalized = unicodedata.normalize("NFKC", address).strip()
    39    +    if not normalized:
    40    +        raise ValueError("email address is empty after trimming")
    41    +
    42    +    normalized = normalized.lower()
    43    +    if not _ADDRESS_PATTERN.match(normalized):
    44    +        raise ValueError("invalid email address format")
    45    +
    46    +    local_part, domain = normalized.rsplit("@", 1)
    47    +    try:
    48    +        domain_ascii = domain.encode("idna").decode("ascii")
    49    +    except UnicodeError as exc:  # pragma: no cover - defensive branch
    50    +        raise ValueError("invalid domain name") from exc
    51    +
    52    +    is_gmail = domain_ascii in _GMAIL_DOMAINS
    53    +    if is_gmail:
    54    +        # Gmail treats dots as insignificant and strips tags starting wit
           h '+'.
    55    +        local_part = local_part.split("+", 1)[0].replace(".", "")
    56    +        if not local_part:
    57    +            raise ValueError("gmail local part empty after normalization"
           )
    58    +        domain_ascii = "gmail.com"
    59    +
    60    +    return f"{local_part}@{domain_ascii}"

• Updated Plan
  └ ✔ Understand spec and plan implementation and tests for email normalization
    ✔ Implement normalize_email with required behavior and documentation
    □ Add pytest unit tests (including boundary/property tests) and Hypothesis if
      feasible
    □ Add README fragment covering usage/design notes and provider behavior
      differences
    □ Add GitHub Actions workflow for pytest on Python 3.11/3.12
    □ List invariants to enforce via tests

> Thinking
  I'm planning a new pytest-based test suite with Hypothesis to cover many cases
  for normalize_email, including Unicode normalization, Gmail-specific local-part
  tweaks, invalid inputs, and property-based tests ensuring idempotence with various
  generated local and domain parts. I'll carefully build string strategies to
  generate valid test emails reflecting real-world quirks, making sure assumptions
  avoid pathological cases that would break normalization. This should give robust
  validation of email normalization behavior!

• Added tests/test_email_utils.py (+82 -0)
    1     +"""Tests for the email normalization utilities."""
    2     +
    3     +from __future__ import annotations
    4     +
    5     +import string
    6     +
    7     +import pytest
    8     +from hypothesis import given, strategies as st
    9     +
    10    +from email_utils import normalize_email
    11    +
    12    +
    13    +def test_basic_normalization() -> None:
    14    +    assert normalize_email(" Foo.Bar@Example.COM ") == "foo.bar@example.c
           om"
    15    +
    16    +
    17    +def test_nfkc_and_fullwidth_handling() -> None:
    18    +    # Full-width characters should collapse to ASCII equivalents and be l
           ower-cased.
    19    +    assert normalize_email(" Ｆｏｏ＠ｅｘａｍｐｌｅ．ｃｏｍ ") ==
"foo@example.com"
    20    +
    21    +
    22    +def test_idna_domain_normalization() -> None:
    23    +    expected = "ユーザー@xn--r8jz45g.xn--zckzah"
    24    +    assert normalize_email("ユーザー@例え.テスト") == expected
    25    +
    26    +
    27    +def test_gmail_specific_rules() -> None:
    28    +    assert (
    29    +        normalize_email("Test.Email+spam@Gmail.Com") == "testemail@gmail.
           com"
    30    +    )
    31    +
    32    +
    33    +def test_googlemail_equivalence() -> None:
    34    +    assert (
    35    +        normalize_email("Test.Email+spam@GoogleMail.com") == "testemail@g
           mail.com"
    36    +    )
    37    +
    38    +
    39    +def test_invalid_format_raises() -> None:
    40    +    with pytest.raises(ValueError):
    41    +        normalize_email("invalid@@example.com")
    42    +
    43    +
    44    +def test_gmail_empty_after_cleanup_raises() -> None:
    45    +    with pytest.raises(ValueError):
    46    +        normalize_email("...+tag@gmail.com")
    47    +
    48    +
    49    +def test_non_string_input_raises() -> None:
    50    +    with pytest.raises(TypeError):
    51    +        normalize_email(123)  # type: ignore[arg-type]
    52    +
    53    +
    54    +LOCAL_ALPHABET = string.ascii_letters + string.digits + "._+-"
    55    +LABEL_ALPHABET = string.ascii_lowercase + string.digits + "-"
    56    +
    57    +
    58    +def _domain_strategy() -> st.SearchStrategy[str]:
    59    +    label = st.text(LABEL_ALPHABET, min_size=1, max_size=10).filter(
    60    +        lambda s: s[0] != "-" and s[-1] != "-"
    61    +    )
    62    +    generic = st.lists(label, min_size=2, max_size=3).map(lambda parts: "
           .".join(parts))
    63    +    return st.one_of(generic, st.sampled_from(["gmail.com", "googlemail.c
           om"]))
    64    +
    65    +
    66    +@given(
    67    +    local=st.text(LOCAL_ALPHABET, min_size=1, max_size=20),
    68    +    domain=_domain_strategy(),
    69    +)
    70    +def test_normalization_idempotent(local: str, domain: str) -> None:
    71    +    canonical_candidate = local
    72    +    if domain in {"gmail.com", "googlemail.com"}:
    73    +        canonical_candidate = canonical_candidate.split("+", 1)[0].replac
           e(".", "")
    74    +        assume_msg = ""
    75    +        if not canonical_candidate:
    76    +            pytest.skip("hypothesis generated gmail local without signifi
           cant characters")
    77    +
    78    +    address = f" {local}@{domain} "
    79    +    first = normalize_email(address)
    80    +    second = normalize_email(first)
    81    +    assert first == second
    82    +

• Edited tests/test_email_utils.py (+2 -7)
    7      import pytest
    8     -from hypothesis import given, strategies as st
    8     +from hypothesis import assume, given, strategies as st
    9
    ⋮
    70     def test_normalization_idempotent(local: str, domain: str) -> None:
    71    -    canonical_candidate = local
    71         if domain in {"gmail.com", "googlemail.com"}:
    73    -        canonical_candidate = canonical_candidate.split("+", 1)[0].replac
           e(".", "")
    74    -        assume_msg = ""
    75    -        if not canonical_candidate:
    76    -            pytest.skip("hypothesis generated gmail local without signifi
           cant characters")
    72    +        assume(local.split("+", 1)[0].replace(".", "") != "")
    73
    ⋮
    77         assert first == second
    82    -

• Ran pytest

> Thinking
  I’m working out how to keep property-based tests running even without Hypothesis
  installed, since network restrictions prevent easy installation. My plan is to try
  importing Hypothesis and, if unavailable, skip those tests or provide a fallback
  using deterministic, randomized tests with pytest, ensuring the property tests
  still cover core checks without breaking. This way I balance following the user's
  request with environment constraints.

• Edited tests/test_email_utils.py (+51 -17)
    4
    5     +import random
    6      import string
    ⋮
    8      import pytest
    8     -from hypothesis import assume, given, strategies as st
    9
    ⋮
    11
    12    +try:
    13    +    from hypothesis import assume, given, strategies as st
    14    +except ModuleNotFoundError:  # pragma: no cover - environment guard
    15    +    HYPOTHESIS_AVAILABLE = False
    16    +else:  # pragma: no cover - import success path exercised via tests
    17    +    HYPOTHESIS_AVAILABLE = True
    18    +
    19
    ⋮
    64
    58    -def _domain_strategy() -> st.SearchStrategy[str]:
    59    -    label = st.text(LABEL_ALPHABET, min_size=1, max_size=10).filter(
    60    -        lambda s: s[0] != "-" and s[-1] != "-"
    65    +if HYPOTHESIS_AVAILABLE:
    66    +
    67    +    def _domain_strategy() -> st.SearchStrategy[str]:
    68    +        label = st.text(LABEL_ALPHABET, min_size=1, max_size=10).filter(
    69    +            lambda s: s[0] != "-" and s[-1] != "-"
    70    +        )
    71    +        generic = (
    72    +            st.lists(label, min_size=2, max_size=3).map(lambda parts: "."
           .join(parts))
    73    +        )
    74    +        return st.one_of(generic, st.sampled_from(["gmail.com", "googlema
           il.com"]))
    75    +
    76    +
    77    +    @given(
    78    +        local=st.text(LOCAL_ALPHABET, min_size=1, max_size=20),
    79    +        domain=_domain_strategy(),
    80         )
    62    -    generic = st.lists(label, min_size=2, max_size=3).map(lambda parts: "
           .".join(parts))
    63    -    return st.one_of(generic, st.sampled_from(["gmail.com", "googlemail.c
           om"]))
    81    +    def test_normalization_idempotent(local: str, domain: str) -> None:
    82    +        if domain in {"gmail.com", "googlemail.com"}:
    83    +            assume(local.split("+", 1)[0].replace(".", "") != "")
    84    +
    85    +        address = f" {local}@{domain} "
    86    +        first = normalize_email(address)
    87    +        second = normalize_email(first)
    88    +        assert first == second
    89
    90    +else:
    91
    66    -@given(
    67    -    local=st.text(LOCAL_ALPHABET, min_size=1, max_size=20),
    68    -    domain=_domain_strategy(),
    69    -)
    70    -def test_normalization_idempotent(local: str, domain: str) -> None:
    71    -    if domain in {"gmail.com", "googlemail.com"}:
    72    -        assume(local.split("+", 1)[0].replace(".", "") != "")
    92    +    def test_normalization_idempotent_fallback() -> None:
    93    +        random.seed(0)
    94    +        domains = [
    95    +            "example.com",
    96    +            "sub.example.net",
    97    +            "gmail.com",
    98    +            "googlemail.com",
    99    +        ]
    100   +        for _ in range(200):
    101   +            domain = random.choice(domains)
    102   +            local_length = random.randint(1, 20)
    103   +            local = "".join(random.choice(LOCAL_ALPHABET) for _ in range(
           local_length))
    104   +            if domain in {"gmail.com", "googlemail.com"}:
    105   +                if local.split("+", 1)[0].replace(".", "") == "":
    106   +                    continue
    107
    74    -    address = f" {local}@{domain} "
    75    -    first = normalize_email(address)
    76    -    second = normalize_email(first)
    77    -    assert first == second
    108   +            address = f" {local}@{domain} "
    109   +            first = normalize_email(address)
    110   +            second = normalize_email(first)
    111   +            assert first == second

• Ran pytest

> Thinking
  I’m figuring out why pytest can’t find the top-level email_utils module despite
  running from root. It looks like pytest isn’t adding root to sys.path as expected,
  so I’m planning to tweak the test files to insert the repo root manually into
  sys.path before imports to fix this.

• Edited tests/test_email_utils.py (+7 -0)
    4
    5     +import pathlib
    6      import random
    ⋮
    10
    11    +import sys
    12    +
    13    +PROJECT_ROOT = pathlib.Path(__file__).resolve().parents[1]
    14    +if str(PROJECT_ROOT) not in sys.path:
    15    +    sys.path.insert(0, str(PROJECT_ROOT))
    16    +
    17     from email_utils import normalize_email

• Ran pytest

• Updated Plan
  └ ✔ Understand spec and plan implementation and tests for email normalization
    ✔ Implement normalize_email with required behavior and documentation
    ✔ Add pytest unit tests (including boundary/property tests) and Hypothesis if
      feasible
    □ Add README fragment covering usage/design notes and provider behavior
      differences
    □ Add GitHub Actions workflow for pytest on Python 3.11/3.12
    □ List invariants to enforce via tests

• Added README.md (+24 -0)
    1     +# Email Normalization Utility
    2     +
    3     +## 使い方
    4     +```python
    5     +from email_utils import normalize_email
    6     +
    7     +canonical = normalize_email(" Test.Email+spam@Gmail.Com ")
    8     +print(canonical)  # => "testemail@gmail.com"
    9     +```
    10    +
    11    +- 例外 `ValueError` は入力の形式が不正な場合に発生します。
    12    +- 非文字列を渡した場合は `TypeError` が送出されます。
    13    +
    14    +## 設計上の注意
    15    +- Unicode は常に NFKC に正規化した上で小文字化し、末端の空白を除去し
ます。
    16    +- ドメイン部分は IDNA 正規化（punycode 化）して ASCII ドメインに統一し
ます。
    17    +- Gmail 系ドメインではローカル部のドットと `+` タグを無視し、`gmail.com`
に揃えます。
    18    +- 期待される入力は単一の `@` を含むメールアドレス文字列のみです。内部空白
や制御文字はサポート対象外です。
    19    +
    20    +## プロバイダごとの注意
    21    +- **Gmail/Googlemail**: ドットや `+tag` を無視する仕様のため、正規化後に
別の実アドレスへ集約されます。
    22    +- **Outlook (Hotmail/Live)**: ドットは重要で、`+` タグも通常のローカル部
として扱われるため変換は行いません。
    23    +- **Yahoo! Mail**: ドットの扱いはローカル部の一部であり、`-` などを使った
エイリアスに注意が必要です。本ライブラリでは Ya
           hoo 固有の別名仕様（`-base` 等）までは正規化しません。
    24    +

• Updated Plan
  └ ✔ Understand spec and plan implementation and tests for email normalization
    ✔ Implement normalize_email with required behavior and documentation
    ✔ Add pytest unit tests (including boundary/property tests) and Hypothesis if
      feasible
    ✔ Add README fragment covering usage/design notes and provider behavior
      differences
    □ Add GitHub Actions workflow for pytest on Python 3.11/3.12
    □ List invariants to enforce via tests

> Thinking
  I'll set up a CI workflow in .github/workflows/ci.yml that runs pytest across
  Python 3.11 and 3.12 using a matrix strategy. I'll include the Hypothesis library
  in the install step to support property-based testing if present. I'll check for
  the directory and create it if missing, then add this workflow file.

• Ran mkdir -p .github/workflows

• Added .github/workflows/ci.yml (+30 -0)
    1     +name: CI
    2     +
    3     +on:
    4     +  push:
    5     +    branches: [ main ]
    6     +  pull_request:
    7     +
    8     +jobs:
    9     +  tests:
    10    +    runs-on: ubuntu-latest
    11    +    strategy:
    12    +      matrix:
    13    +        python-version: ["3.11", "3.12"]
    14    +    steps:
    15    +      - name: Checkout source
    16    +        uses: actions/checkout@v4
    17    +
    18    +      - name: Set up Python
    19    +        uses: actions/setup-python@v5
    20    +        with:
    21    +          python-version: ${{ matrix.python-version }}
    22    +
    23    +      - name: Install dependencies
    24    +        run: |
    25    +          python -m pip install --upgrade pip
    26    +          python -m pip install pytest hypothesis
    27    +
    28    +      - name: Run pytest
    29    +        run: python -m pytest
    30    +

• Updated Plan
  └ ✔ Understand spec and plan implementation and tests for email normalization
    ✔ Implement normalize_email with required behavior and documentation
    ✔ Add pytest unit tests (including boundary/property tests) and Hypothesis if
      feasible
    ✔ Add README fragment covering usage/design notes and provider behavior
      differences
    ✔ Add GitHub Actions workflow for pytest on Python 3.11/3.12
    □ List invariants to enforce via tests

• Ran git status --short

• Updated Plan
  └ Prepared invariants to describe alongside tests; remaining work is reporting
    back.
    ✔ Understand spec and plan implementation and tests for email normalization
    ✔ Implement normalize_email with required behavior and documentation
    ✔ Add pytest unit tests (including boundary/property tests) and Hypothesis if
      feasible
    ✔ Add README fragment covering usage/design notes and provider behavior
      differences
    ✔ Add GitHub Actions workflow for pytest on Python 3.11/3.12
    ✔ List invariants to enforce via tests

• Explored
  └ Read email_utils.py, test_email_utils.py, README.md, ci.yml

> Thinking
  I'm preparing a final summary that highlights the pytest run and the key
  invariants enforced through tests, like normalization idempotence, Gmail-specific
  canonicalization, and domain ASCII enforcement. I'll also mention the Hypothesis
  fallback mechanism ensuring coverage even when optional dependencies are missing,
  along with referencing code locations and changes consistently.

> Changes

  - Implemented a fully spec-compliant normalize_email with NFKC trimming, IDNA
  domain handling, and Gmail-specific canonicalization plus validation and docstring
  in email_utils.py:1-60.
  - Added boundary/unit/property coverage (Hypothesis when available, deterministic
  fallback otherwise) and ensured repository importability inside tests/
  test_email_utils.py:1-118.
  - Documented usage, design considerations, and Gmail/Outlook/Yahoo behavior
  differences in README.md:1-24.
  - Set up GitHub Actions CI matrix to run python -m pytest on Python 3.11/3.12 and
  install pytest/Hypothesis in .github/workflows/ci.yml:1-29.

  Tests

  - pytest

  Invariants

  - Normalization is idempotent: normalize_email(normalize_email(x)) always equals
  normalize_email(x) for valid addresses.
  - Gmail-family inputs canonicalize to gmail.com with a non-empty local part free
  of dots and tags after normalization.
  - Every normalized address contains exactly one @, has no surrounding whitespace,
  and uses an ASCII domain (IDNA applied).