fix xxhash 4 string hashing, skip numpy-2-incompatible pytables tests

alphabase/peptide/precursor.py

@@ -15,6 +15,11 @@
 from alphabase.peptide.mass_calc import calc_peptide_masses_for_same_len_seqs
 
 
+def _xxh64_str_intdigest(text: str, *, seed: int) -> int:
+    """xxhash>=4 no longer encodes strings implicitly; utf-8 keeps hash values identical to xxhash<4."""
+    return xxh64_intdigest(text.encode(), seed=seed)
+
+
 def refine_precursor_df(
     df: pd.DataFrame,
     drop_frag_idx=False,
@@ -178,9 +183,9 @@ def get_mod_seq_hash(
     """
     return np.array(
         [
-            xxh64_intdigest(sequence, seed=seed),
-            xxh64_intdigest(mods, seed=seed),
-            xxh64_intdigest(mod_sites, seed=seed),
+            _xxh64_str_intdigest(sequence, seed=seed),
+            _xxh64_str_intdigest(mods, seed=seed),
+            _xxh64_str_intdigest(mod_sites, seed=seed),
         ],
         dtype=np.uint64,
     ).sum()  # use np.sum to prevent overflow
@@ -233,15 +238,15 @@ def get_mod_seq_charge_hash(
 def hash_mod_seq_df(precursor_df: pd.DataFrame, *, seed=0):
     """Internal function"""
     hash_vals = precursor_df.sequence.apply(
-        lambda x: xxh64_intdigest(x, seed=seed)
+        lambda x: _xxh64_str_intdigest(x, seed=seed)
     ).to_numpy(copy=True, dtype=np.uint64)
     hash_vals += (
-        precursor_df.mods.apply(lambda x: xxh64_intdigest(x, seed=seed))
+        precursor_df.mods.apply(lambda x: _xxh64_str_intdigest(x, seed=seed))
         .astype(np.uint64)
         .values
     )
     hash_vals += (
-        precursor_df.mod_sites.apply(lambda x: xxh64_intdigest(x, seed=seed))
+        precursor_df.mod_sites.apply(lambda x: _xxh64_str_intdigest(x, seed=seed))
         .astype(np.uint64)
         .values
     )

tests/integration/conftest.py

@@ -2,22 +2,38 @@
 
 import importlib.util
 import os
+from importlib.metadata import version
 from pathlib import Path
 
+import numpy as np
 import pandas as pd
 import pytest
+from packaging.version import Version
 
 from alphabase.tools.data_downloader import DataShareDownloader
 
 TABLES_PACKAGE_UNAVAILABLE = importlib.util.find_spec("tables") is None
 NUMBA_UNAVAILABLE = importlib.util.find_spec("numba") is None
 
+FIRST_NUMPY2_COMPATIBLE_TABLES_VERSION = Version("3.10")
+FIRST_ABI_BREAKING_NUMPY_VERSION = Version("2.0")
+
+PYTABLES_NUMPY_ABI_BROKEN = not TABLES_PACKAGE_UNAVAILABLE and (
+    Version(version("tables")) < FIRST_NUMPY2_COMPATIBLE_TABLES_VERSION
+    and Version(np.__version__) >= FIRST_ABI_BREAKING_NUMPY_VERSION
+)
+
 
 pytest.mark.optional_pytables_dependency = pytest.mark.skipif(
     TABLES_PACKAGE_UNAVAILABLE,
     reason="pytables package not installed. Install with `pip install alphabase[hdf]`",
 )
 
+pytest.mark.broken_pytables_numpy_abi = pytest.mark.skipif(
+    PYTABLES_NUMPY_ABI_BROKEN,
+    reason="tables<3.10 is built against numpy<2 and raises a dtype-size ValueError on import with numpy>=2",
+)
+
 pytest.mark.requires_numba = pytest.mark.skipif(
     NUMBA_UNAVAILABLE,
     reason="numba package not installed",

tests/integration/test_pg_readers.py

@@ -53,6 +53,7 @@ def test_import_csv_file_equivalent(
 
         pd.testing.assert_frame_equal(result_df, reference)
 
+    @pytest.mark.broken_pytables_numpy_abi
     @pytest.mark.optional_pytables_dependency
     def test_import_hdf_file_equivalent(
         self, example_alphapept_hdf: tuple[str, pd.DataFrame]
@@ -108,6 +109,7 @@ def test_import_csv_file(
             PGCols.DECOY_INDICATOR,
         ]
 
+    @pytest.mark.broken_pytables_numpy_abi
     @pytest.mark.optional_pytables_dependency
     @pytest.mark.parametrize(
         ("measurement_regex", "expected_shape", "expected_colums"),