fix[py]: remove mimalloc in library (#7826)

Author: joseph-isaacs

Cargo.lock

@@ -28,10 +28,6 @@ default = ["extension-module", "tui"]
 # duplicate PyInit__lib symbols.
 extension-module = []
 tui = ["dep:vortex-tui"]
-# Set the default allocator to mimalloc, this is enabled by default
-# for the maturin wheels, but it is not a default feature to not conflict
-# with users that set a different global allocator.
-mimalloc = ["dep:mimalloc"]
 
 [dependencies]
 arrow-array = { workspace = true }
@@ -40,7 +36,6 @@ arrow-schema = { workspace = true }
 bytes = { workspace = true }
 itertools = { workspace = true }
 log = { workspace = true }
-mimalloc = { workspace = true, optional = true }
 object_store = { workspace = true, features = [
     "fs",
     "aws",

vortex-python/Cargo.toml

@@ -61,7 +61,7 @@ profile = "release"
 editable-profile = "dev"
 python-source = "python"
 module-name = "vortex._lib"
-features = ["pyo3/extension-module", "mimalloc"]
+features = ["pyo3/extension-module"]
 compatibility = "manylinux2014"
 include = [
     { path = "rust-toolchain.toml", format = "sdist" },

vortex-python/pyproject.toml

@@ -1,10 +1,6 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright the Vortex contributors
 
-#[cfg(feature = "mimalloc")]
-#[global_allocator]
-static GLOBAL: mimalloc::MiMalloc = mimalloc::MiMalloc;
-
 use std::ops::Deref;
 use std::sync::LazyLock;
 

vortex-python/src/lib.rs

@@ -0,0 +1,37 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright the Vortex contributors
+
+"""Regression test for https://github.com/vortex-data/vortex/issues/7760.
+
+Importing vortex after pyarrow.dataset must not corrupt pyarrow's runtime.
+This test runs in a subprocess because import-order bugs and allocator
+conflicts only manifest in a fresh process with specific load ordering.
+"""
+
+import subprocess
+import sys
+
+import pytest
+
+
+@pytest.mark.parametrize(
+    "script",
+    [
+        # Issue #7760: pyarrow.dataset before vortex
+        "import pyarrow.dataset; import vortex; import pyarrow as pa; pa.array([1])",
+        # Reverse order should also be fine
+        "import vortex; import pyarrow.dataset; import pyarrow as pa; pa.array([1])",
+    ],
+    ids=["pyarrow_first", "vortex_first"],
+)
+def test_import_order_no_crash(script: str) -> None:
+    result = subprocess.run(
+        [sys.executable, "-c", script],
+        capture_output=True,
+        timeout=30,
+    )
+    assert result.returncode == 0, (
+        f"Process crashed (exit code {result.returncode}).\n"
+        f"stdout: {result.stdout.decode()}\n"
+        f"stderr: {result.stderr.decode()}"
+    )