From 6f4f3756012f56efb26bf22a8b8548199edf0cf9 Mon Sep 17 00:00:00 2001
From: Patrick Hoefler <61934744+phofl@users.noreply.github.com>
Date: Mon, 6 May 2024 21:21:21 +0200
Subject: [PATCH] Fix pandas internals changes (#78)

---
 .github/workflows/ci.yml |  2 +-
 partd/pandas.py          | 57 +++++++++++++++++++++++-----------------
 pyproject.toml           |  9 +++----
 3 files changed, 38 insertions(+), 30 deletions(-)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 8d58bce..db8cdce 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -9,7 +9,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ["3.7", "3.8", "3.9", "3.10", "3.11"]
+        python-version: ["3.9", "3.10", "3.11", "3.12"]
 
     steps:
       - name: Checkout source
diff --git a/partd/pandas.py b/partd/pandas.py
index 5a25c26..880558c 100644
--- a/partd/pandas.py
+++ b/partd/pandas.py
@@ -1,38 +1,40 @@
 from functools import partial
 import pickle
 
-import numpy as np
 import pandas as pd
-from pandas.core.internals import create_block_manager_from_blocks, make_block
+from packaging.version import Version
+
+PANDAS_GE_210 = Version(pd.__version__).release >= (2, 1, 0)
+PANDAS_GE_300 =  Version(pd.__version__).major >= 3
+
+if PANDAS_GE_300:
+    from pandas.api.internals import create_dataframe_from_blocks
+    create_block_manager_from_blocks = None
+    make_block = None
+else:
+    create_dataframe_from_blocks = None
+    try:
+        from pandas.core.internals.managers import create_block_manager_from_blocks
+    except ImportError:
+        from pandas.core.internals import create_block_manager_from_blocks
+
+    from pandas.core.internals import make_block
 
 from . import numpy as pnp
 from .core import Interface
 from .encode import Encode
 from .utils import extend, framesplit, frame
+from pandas.api.types import is_extension_array_dtype
+from pandas.api.extensions import ExtensionArray
 
-try:
-    # pandas >= 0.24.0
-    from pandas.api.types import is_extension_array_dtype
-except ImportError:
-    def is_extension_array_dtype(dtype):
-        return False
-
-try:
-    # Some `ExtensionArray`s can have a `.dtype` which is not a `ExtensionDtype`
-    # (e.g. they can be backed by a NumPy dtype). For these cases we check
-    # whether the instance is a `ExtensionArray`.
-    # https://github.com/dask/partd/issues/48
-    from pandas.api.extensions import ExtensionArray
-    def is_extension_array(x):
-        return isinstance(x, ExtensionArray)
-except ImportError:
-    def is_extension_array(x):
-        return False
+def is_extension_array(x):
+    return isinstance(x, ExtensionArray)
 
 
 dumps = partial(pickle.dumps, protocol=pickle.HIGHEST_PROTOCOL)
 
 
+
 class PandasColumns(Interface):
     def __init__(self, partd=None):
         self.partd = pnp.Numpy(partd)
@@ -148,7 +150,7 @@ def block_to_header_bytes(block):
     return header, bytes
 
 
-def block_from_header_bytes(header, bytes):
+def block_from_header_bytes(header, bytes, create_block: bool):
     placement, dtype, shape, (extension_type, extension_values) = header
 
     if extension_type == "other":
@@ -164,7 +166,9 @@ def block_from_header_bytes(header, bytes):
         tz_info = extension_values[0]
         values = pd.DatetimeIndex(values).tz_localize('utc').tz_convert(
             tz_info)
-    return make_block(values, placement=placement)
+    if create_block:
+        return make_block(values, placement=placement)
+    return values, placement
 
 
 def serialize(df):
@@ -193,9 +197,14 @@ def deserialize(bytes):
     bytes = frames[1:]
     axes = [index_from_header_bytes(headers[0], bytes[0]),
             index_from_header_bytes(headers[1], bytes[1])]
-    blocks = [block_from_header_bytes(h, b)
+    blocks = [block_from_header_bytes(h, b, create_block=not PANDAS_GE_300)
               for (h, b) in zip(headers[2:], bytes[2:])]
-    return pd.DataFrame(create_block_manager_from_blocks(blocks, axes))
+    if PANDAS_GE_300:
+        return pd.api.internals.create_dataframe_from_blocks(blocks, axes[1], axes[0])
+    elif PANDAS_GE_210:
+        return pd.DataFrame._from_mgr(create_block_manager_from_blocks(blocks, axes), axes=axes)
+    else:
+        return pd.DataFrame(create_block_manager_from_blocks(blocks, axes))
 
 
 def join(dfs):
diff --git a/pyproject.toml b/pyproject.toml
index 051a159..1eca1aa 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -10,15 +10,14 @@ license = {text = "BSD"}
 keywords = []
 classifiers = [
     "Programming Language :: Python :: 3",
-    "Programming Language :: Python :: 3.7",
-    "Programming Language :: Python :: 3.8",
     "Programming Language :: Python :: 3.9",
     "Programming Language :: Python :: 3.10",
     "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
 ]
 readme = "README.rst"
 urls = {Homepage = "http://github.com/dask/partd/"}
-requires-python = ">=3.7"
+requires-python = ">=3.9"
 dynamic = ["version"]
 dependencies = [
     "locket",
@@ -27,8 +26,8 @@ dependencies = [
 
 [project.optional-dependencies]
 complete = [
-    "numpy >= 1.9.0",
-    "pandas >=0.19.0",
+    "numpy >= 1.20.0",
+    "pandas >=1.3",
     "pyzmq",
     "blosc",
 ]