Add utility functions to get name of GNU tar executable in OS-agnostic manner

install/install-mac-deps.sh

@@ -24,3 +24,5 @@ brew install awscli;
 
 # pkg-config is required for building the wheel
 brew install pkg-config;
+
+brew install gnu-tar

python/python/bystro/utils/tar.py

@@ -0,0 +1,27 @@
+"""Utility functions for safely invoking tar in cross-OS manner."""
+import shutil
+import sys
+
+
+def _get_executable_name(name: str) -> str:
+    executable_name = shutil.which(name)
+    if executable_name is None:
+        err_msg = f"executable `{name}` not found on system"
+        raise OSError(err_msg)
+    return executable_name
+
+
+def _get_gnu_tar_executable_name() -> str:
+    """Find the name of the GNU tar executable on user's instance."""
+    # Macs use bsdtar by default.  bsdtar is not fully compatible with
+    # GNU tar, the implementation we use.  So if we're on a mac, we'll
+    # want to use 'gtar' instead of 'tar'.
+    if sys.platform == "linux":
+        return _get_executable_name("tar")
+    if sys.platform == "darwin":
+        return _get_executable_name("gtar")
+    err_msg = f"Operating system must be linux or macosx, got {sys.platform} instead."
+    raise OSError(err_msg)
+
+
+GNU_TAR_EXECUTABLE_NAME = _get_gnu_tar_executable_name()

python/python/bystro/utils/tests/test_tar.py

@@ -0,0 +1,70 @@
+"""Test functions for utils module."""
+import shutil
+import subprocess
+import sys
+from subprocess import CalledProcessError
+from typing import Any, Callable, NoReturn
+from unittest.mock import Mock
+
+import pytest
+from bystro.utils.tar import (
+    GNU_TAR_EXECUTABLE_NAME,
+    _get_gnu_tar_executable_name,
+)
+
+
+def test_GNU_TAR_EXECUTABLE_NAME():
+    assert GNU_TAR_EXECUTABLE_NAME in ["/usr/bin/tar", "/opt/homebrew/bin/gtar"]
+
+
+# These tests check that we correctly infer the name of the GNU tar
+# executable regardless of the operating system we find ourselves
+# running on, or that we raise an appropriate error if we can't
+# determine the correct executable.  The
+# _mock_subprocess_run_... methods mocks all interaction with
+# subprocess.run under various OS scenarios and the tests ensure that
+# we return the correct GNU tar executable name or raise the appropriate
+# exception.
+
+
+def _mock_subprocess(stdout_string: str) -> Callable[[Any, Any], Mock]:
+    """Create a subprocess.run mock that returns stdout_string."""
+    mock = Mock()
+    mock.stdout = stdout_string
+
+    def f(*args: list[str], **kwargs: dict[str, Any]) -> Mock:
+        del args, kwargs
+        return mock
+
+    return f
+
+
+def test_get_gnu_tar_executable_macosx_gnu_tar_installed(monkeypatch):
+    monkeypatch.setattr(subprocess, "run", _mock_subprocess("/opt/homebrew/bin/gtar\n"))
+    monkeypatch.setattr(sys, "platform", "darwin")
+    monkeypatch.setattr(shutil, "which", lambda _name: "/opt/homebrew/bin/gtar")
+    assert "/opt/homebrew/bin/gtar" == _get_gnu_tar_executable_name()
+
+
+def test_get_gnu_tar_executable_macosx_gnu_tar_not_installed(monkeypatch):
+    def raise_error(args: list[str], kwargs: dict[str, Any]) -> NoReturn:
+        del args, kwargs
+        raise CalledProcessError(cmd=["which", "gtar"], returncode=1)
+
+    monkeypatch.setattr(shutil, "which", lambda _name: None)
+    monkeypatch.setattr(sys, "platform", "darwin")
+    with pytest.raises(OSError, match="executable `gtar` not found on system"):
+        _get_gnu_tar_executable_name()
+
+
+def test_get_gnu_tar_executable_linux(monkeypatch):
+    monkeypatch.setattr(subprocess, "run", _mock_subprocess("/usr/bin/tar\n"))
+    monkeypatch.setattr(sys, "platform", "linux")
+    monkeypatch.setattr(shutil, "which", lambda _name: "/usr/bin/tar")
+    assert "/usr/bin/tar" == _get_gnu_tar_executable_name()
+
+
+def test_get_gnu_tar_executable_unknown_os(monkeypatch):
+    monkeypatch.setattr(sys, "platform", "some unknown OS")
+    with pytest.raises(OSError, match="got some unknown OS instead"):
+        _get_gnu_tar_executable_name()