GMT_DATASET: Add workaround for None values in trailing text (#3174)

Co-authored-by: Wei Ji <[email protected]>

Author: seisman

.github/workflows/cache_data.yaml

@@ -80,7 +80,7 @@ jobs:
         with:
           name: gmt-cache
           path: |
-              ~/.gmt/cache
-              ~/.gmt/server
-              ~/.gmt/gmt_data_server.txt
-              ~/.gmt/gmt_hash_server.txt
+            ~/.gmt/cache
+            ~/.gmt/server
+            ~/.gmt/gmt_data_server.txt
+            ~/.gmt/gmt_hash_server.txt

pygmt/datatypes/dataset.py

@@ -3,6 +3,7 @@
 """
 
 import ctypes as ctp
+import warnings
 from collections.abc import Mapping
 from typing import Any, ClassVar
 
@@ -153,6 +154,17 @@ def to_strings(self) -> np.ndarray[Any, np.dtype[np.str_]]:
             for segment in table.contents.segment[: table.contents.n_segments]:
                 if segment.contents.text:
                     textvector.extend(segment.contents.text[: segment.contents.n_rows])
+        if None in textvector:
+            # Workaround for upstream GMT bug reported in
+            # https://github.com/GenericMappingTools/pygmt/issues/3170.
+            msg = (
+                "The trailing text column contains `None' values and has been replaced"
+                "with empty strings to avoid TypeError exceptions. "
+                "It's likely caused by an upstream GMT API bug. "
+                "Please consider reporting to us."
+            )
+            warnings.warn(msg, category=RuntimeWarning, stacklevel=1)
+            textvector = [item if item is not None else b"" for item in textvector]
         return np.char.decode(textvector) if textvector else np.array([], dtype=str)
 
     def to_dataframe(

pygmt/tests/test_datatypes_dataset.py

@@ -6,6 +6,7 @@
 
 import pandas as pd
 import pytest
+from pygmt import which
 from pygmt.clib import Session
 from pygmt.helpers import GMTTempFile
 
@@ -81,3 +82,28 @@ def test_dataset_empty():
         assert df.empty  # Empty DataFrame
         expected_df = dataframe_from_pandas(tmpfile.name)
         pd.testing.assert_frame_equal(df, expected_df)
+
+
+def test_dataset_to_strings_with_none_values():
+    """
+    Test that None values in the trailing text doesn't raise an excetion.
+
+    Due to a likely upstream bug, the trailing texts sometimes can be ``None`` when
+    downloading tiled grids. The temporary workaround is to replace any None values with
+    an empty string.
+
+    See the bug report at https://github.com/GenericMappingTools/pygmt/issues/3170.
+    """
+    tiles = ["@N30W120.earth_relief_15s_p.nc", "@N00E000.earth_relief_15s_p.nc"]
+    paths = which(fname=tiles, download="a")
+    assert len(paths) == 2
+    # 'paths' may contain an empty string or not, depending on if the tiles are cached.
+    if "" not in paths:  # Contains two valid paths.
+        # Delete the cached tiles and try again.
+        for path in paths:
+            Path(path).unlink()
+        with pytest.warns(expected_warning=RuntimeWarning) as record:
+            paths = which(fname=tiles, download="a")
+            assert len(record) == 1
+        assert len(paths) == 2
+        assert "" in paths