Deprecations and FutureWarnings Fixes

setup.py

@@ -30,7 +30,7 @@ def about_package(init_posixpath: pathlib.Path) -> dict:
 # Dependencies for the package itself
 DEPENDENCIES = [
     "numpy>=1.19.0",
-    "pandas>=1.0",
+    "pandas>=2.1",
 ]
 
 # Extra dependencies
@@ -53,7 +53,7 @@ def about_package(init_posixpath: pathlib.Path) -> dict:
     url=ABOUT_TFS["__url__"],
     packages=setuptools.find_packages(include=(MODULE_NAME,)),
     include_package_data=True,
-    python_requires=">=3.7",
+    python_requires=">=3.9",
     license=ABOUT_TFS["__license__"],
     classifiers=[
         "Development Status :: 5 - Production/Stable",

tfs/__init__.py

@@ -10,7 +10,7 @@
 __title__ = "tfs-pandas"
 __description__ = "Read and write tfs files."
 __url__ = "https://github.com/pylhc/tfs"
-__version__ = "3.7.3"
+__version__ = "3.8.0"
 __author__ = "pylhc"
 __author_email__ = "pylhc@github.com"
 __license__ = "MIT"
@@ -19,4 +19,4 @@
 read = read_tfs
 write = write_tfs
 
-__all__ = [concat, read, write, read_hdf, write_hdf, TfsDataFrame, __version__]
+__all__ = ["concat", "read", "write", "read_hdf", "write_hdf", "TfsDataFrame", "TfsFormatError", "__version__"]

tfs/collection.py

@@ -4,7 +4,7 @@
 
 Advanced **TFS** files reading and writing functionality.
 """
-from typing import Tuple, List, Dict
+from typing import Tuple, Dict
 
 import pathlib
 

tfs/frame.py

@@ -263,8 +263,13 @@ def _element_is_list(element):
         raise TfsFormatError("Lists or tuple elements are not accepted in a TfsDataFrame")
 
     # -----  Check that no element is non-physical value in the dataframe ----- #
-    with pd.option_context('mode.use_inf_as_na', True):
-        inf_or_nan_bool_df = data_frame.isna()
+    # The pd.option_context('mode.use_inf_as_na', True) context manager raises FutureWarning
+    # and will likely disappear in pandas 3.0 so we replace 'inf' values by NaNs before calling
+    # .isna(). Additionally, the downcasting behaviour of .replace() is deprecated and raises a
+    # FutureWarning, so we use .infer_objects() first to attemps soft conversion to a better dtype
+    # for object-dtype columns (which strings can be). Since .infer_objects() and .replace() return
+    # (lazy for the former) copies we're not modifying the original dataframe during validation :)
+    inf_or_nan_bool_df = data_frame.infer_objects().replace([np.inf, -np.inf], np.nan).isna()
 
     if inf_or_nan_bool_df.to_numpy().any():
         LOGGER.warning(

tfs/reader.py

@@ -135,8 +135,7 @@ def read_tfs(
         tfs_file_path,
         engine="c",  # faster, and we do not need the features of the python engine
         skiprows=metadata.non_data_lines,  # no need to read these lines again
-        delim_whitespace=True,  # understands ' ' is our delimiter
-        skipinitialspace=True,  # understands ' ' and '     ' are both valid delimiters
+        sep=r"\s+",  # understands ' ' as delimiter | replaced deprecated 'delim_whitespace' in tfs-pandas 3.8.0
         quotechar='"',  # elements surrounded by " are one entry -> correct parsing of strings with spaces
         names=metadata.column_names,  # column names we have determined, avoids using first read row for columns
         dtype=dict(

tfs/writer.py

@@ -241,7 +241,7 @@ def quote_strings(s):
                 return f'"{s}"'
         return s
 
-    data_frame = data_frame.applymap(quote_strings)
+    data_frame = data_frame.map(quote_strings)
     return data_frame