# Pandas UUID dtype

Pandas ExtensionArray / ExtensionDType for UUID

.gitignore

/.venv/

pandas types.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Official docs: https://pandas.pydata.org/docs/development/extending.html\n",
    "- StackOverflow: https://stackoverflow.com/a/68972163/247482\n",
    "- Arrow integration: https://arrow.apache.org/docs/python/extending_types.html#defining-extension-types-user-defined-types\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from __future__ import annotations\n",
    "\n",
    "import builtins\n",
    "from uuid import UUID\n",
    "from typing import TYPE_CHECKING, ClassVar, Self, Never, get_args\n",
    "from collections.abc import Sequence, Iterable\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from numpy.typing import NDArray\n",
    "from pandas.api.indexers import check_array_indexer\n",
    "from pandas.api.extensions import ExtensionDtype, ExtensionArray\n",
    "from pandas.core.ops.common import unpack_zerodim_and_defer\n",
    "from pandas.core.algorithms import take\n",
    "\n",
    "if TYPE_CHECKING:\n",
    "    import pyarrow\n",
    "    from pandas.core.arrays import BooleanArray"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "UuidLike = UUID | bytes | int | str\n",
    "\n",
    "# 16 void bytes: 128 bit, every pattern valid, no funky behavior like 0 stripping.\n",
    "_UuidNumpyDtype = np.dtype(\"V16\")\n",
    "_UuidScalar = _UuidNumpyDtype.type\n",
    "\n",
    "\n",
    "def _to_uuid(v: UuidLike) -> UUID:\n",
    "    match v:\n",
    "        case UUID():\n",
    "            return v\n",
    "        case bytes():\n",
    "            return UUID(bytes=v)\n",
    "        case int():\n",
    "            return UUID(int=v)\n",
    "        case str():\n",
    "            return UUID(v)\n",
    "    msg = f\"Unknown type for Uuid: {type(v)} is not {get_args(UuidLike)}\"\n",
    "    raise TypeError(msg)\n",
    "\n",
    "\n",
    "class UuidDtype(ExtensionDtype):\n",
    "    # ExtensionDtype essential API (3 class attrs and methods)\n",
    "\n",
    "    name: ClassVar[str] = \"uuid\"\n",
    "    type: ClassVar[builtins.type[UUID]] = UUID\n",
    "\n",
    "    @classmethod\n",
    "    def construct_array_type(cls) -> type[UuidExtensionArray]:\n",
    "        return UuidExtensionArray\n",
    "\n",
    "    # ExtensionDtype overrides\n",
    "\n",
    "    kind: ClassVar[str] = _UuidNumpyDtype.kind\n",
    "    # index_class: ClassVar[type[pd.Index]] = pd.Index\n",
    "\n",
    "    @property\n",
    "    def na_value(self) -> Never:\n",
    "        # TODO: figure this out\n",
    "        raise NotImplementedError()\n",
    "\n",
    "    # IO\n",
    "\n",
    "    def __from_arrow__(self, array: pyarrow.Array) -> ExtensionArray:\n",
    "        ...\n",
    "\n",
    "\n",
    "class UuidExtensionArray(ExtensionArray):\n",
    "    # Implementation details and convenience\n",
    "\n",
    "    _data: NDArray[_UuidScalar]\n",
    "\n",
    "    def __init__(self, values: Iterable[UuidLike], *, copy: bool = False) -> None:\n",
    "        if isinstance(values, np.ndarray):\n",
    "            self._data = values.astype(_UuidNumpyDtype, copy=copy)\n",
    "        else:\n",
    "            # TODO: more efficient\n",
    "            self._data = np.array(\n",
    "                [_to_uuid(x).bytes for x in values], dtype=_UuidNumpyDtype\n",
    "            )\n",
    "\n",
    "        if self._data.ndim != 1:\n",
    "            raise ValueError(\"Array only supports 1-d arrays\")\n",
    "\n",
    "    # ExtensionArray essential API (11 class attrs and methods)\n",
    "\n",
    "    dtype: ClassVar[UuidDtype] = UuidDtype()\n",
    "\n",
    "    @classmethod\n",
    "    def _from_sequence(\n",
    "        cls,\n",
    "        data: Iterable[UuidLike],\n",
    "        dtype: UuidDtype | None = None,\n",
    "        copy: bool = False,\n",
    "    ) -> Self:\n",
    "        if dtype is None:\n",
    "            dtype = UuidDtype()\n",
    "\n",
    "        if not isinstance(dtype, UuidDtype):\n",
    "            msg = f\"'{cls.__name__}' only supports 'UuidDtype' dtype\"\n",
    "            raise TypeError(msg)\n",
    "        return cls(data, copy=copy)\n",
    "\n",
    "    def __getitem__(self, index) -> Self | UUID:\n",
    "        if isinstance(index, int):\n",
    "            return UUID(bytes=self._data[index].tobytes())\n",
    "        index = check_array_indexer(self, index)\n",
    "        return self._simple_new(self._data[index])\n",
    "\n",
    "    # def __setitem__(self, index, value):\n",
    "\n",
    "    def __len__(self) -> int:\n",
    "        return len(self._data)\n",
    "\n",
    "    @unpack_zerodim_and_defer(\"__eq__\")\n",
    "    def __eq__(self, other):\n",
    "        return self._cmp(\"eq\", other)\n",
    "\n",
    "    def nbytes(self) -> int:\n",
    "        return self._data.nbytes\n",
    "\n",
    "    def isna(self) -> NDArray[np.bool_]:\n",
    "        return pd.isna(self._data)\n",
    "\n",
    "    def take(\n",
    "        self, indexer, *, allow_fill: bool = False, fill_value: UUID | None = None\n",
    "    ) -> Self:\n",
    "        if allow_fill and fill_value is None:\n",
    "            fill_value = self.dtype.na_value\n",
    "\n",
    "        result = take(self._data, indexer, allow_fill=allow_fill, fill_value=fill_value)\n",
    "        return self._simple_new(result)\n",
    "\n",
    "    def copy(self) -> Self:\n",
    "        return self._simple_new(self._data.copy())\n",
    "\n",
    "    @classmethod\n",
    "    def _concat_same_type(cls, to_concat: Sequence[Self]) -> Self:\n",
    "        return cls._simple_new(np.concatenate([x._data for x in to_concat]))\n",
    "\n",
    "    # Helpers\n",
    "\n",
    "    @classmethod\n",
    "    def _simple_new(cls, values: NDArray[_UuidScalar]) -> Self:\n",
    "        result = UuidExtensionArray.__new__(cls)\n",
    "        result._data = values\n",
    "        return result\n",
    "\n",
    "    def _cmp(self, op: str, other) -> BooleanArray:\n",
    "        if isinstance(other, UuidExtensionArray):\n",
    "            other = other._data\n",
    "        elif isinstance(other, Sequence):\n",
    "            other = np.asarray(other)\n",
    "            if other.ndim > 1:\n",
    "                raise NotImplementedError(\"can only perform ops with 1-d structures\")\n",
    "            if len(self) != len(other):\n",
    "                raise ValueError(\"Lengths must match to compare\")\n",
    "\n",
    "        method = getattr(self._data, f\"__{op}__\")\n",
    "        result = method(other)\n",
    "\n",
    "        # if result is NotImplemented:\n",
    "        #     result = invalid_comparison(self._data, other, op)\n",
    "\n",
    "        rv: BooleanArray = pd.array(result, dtype=\"boolean\")  # type: ignore\n",
    "        return rv\n",
    "\n",
    "    # IO\n",
    "\n",
    "    def __arrow_array__(self, type=None):\n",
    "        \"\"\"convert the underlying array values to a pyarrow Array\"\"\"\n",
    "        import pyarrow\n",
    "\n",
    "        return pyarrow.array(..., type=type)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pandas has a bug around void dtypes: https://github.com/pandas-dev/pandas/issues/54810\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from contextlib import contextmanager\n",
    "\n",
    "\n",
    "@contextmanager\n",
    "def patch_pandas_constructors():\n",
    "    @classmethod\n",
    "    def _validate_dtype(\n",
    "        cls, dtype: np.dtype | ExtensionDtype\n",
    "    ) -> np.dtype | ExtensionDtype | None:\n",
    "        if dtype is None:\n",
    "            return None\n",
    "\n",
    "        from pandas.core.dtypes.common import pandas_dtype\n",
    "\n",
    "        dtype = pandas_dtype(dtype)\n",
    "        # a compound dtype\n",
    "        if getattr(dtype, \"fields\", None) is not None:\n",
    "            raise NotImplementedError(\n",
    "                \"compound dtypes are not implemented \"\n",
    "                f\"in the {cls.__name__} constructor\"\n",
    "            )\n",
    "\n",
    "        return dtype\n",
    "\n",
    "    from unittest.mock import patch\n",
    "    from pandas.core.generic import NDFrame\n",
    "\n",
    "    with patch.object(NDFrame, \"_validate_dtype\", _validate_dtype):\n",
    "        yield"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<UuidExtensionArray>\n",
       "[UUID('00000000-0000-0000-0000-000000000000'), UUID('5748c75a-4418-4bc2-81b8-59703a6ba0cd')]\n",
       "Length: 2, dtype: uuid"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from uuid import uuid4\n",
    "\n",
    "UuidExtensionArray([0, uuid4()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    8be4f607-afaa-4f34-868d-727bc9088999\n",
       "Name: s, dtype: uuid"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "with patch_pandas_constructors():\n",
    "    s = pd.Series([uuid4()], dtype=UuidDtype(), name=\"s\")\n",
    "s"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>s</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>8be4f607-afaa-4f34-868d-727bc9088999</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                      s\n",
       "0  8be4f607-afaa-4f34-868d-727bc9088999"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame(s)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Numeric Python",
   "language": "python",
   "name": "numeric"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.5"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}

pyproject.toml

[project]
name = "numeric-notebooks"
version = "0.1.0"
dependencies = ["numpy", "scipy", "pandas", "pyarrow"]