Relax varnames

lib/ugrid_checks/check.py

@@ -49,8 +49,15 @@
     "trajectory_id",
 ]
 
-# Valid cf varname regex : copied from iris.common.metadata code.
-_VALID_NAME_REGEX = re.compile(r"""^[a-zA-Z][a-zA-Z0-9]*[\w.+\-@]*$""")
+# Valid varname regex.
+# It is not possible to use the common 'variable name' patterns,
+# as we *can* legally have variables which start with a digit, etc.
+# In principle, in netcdf anything but forward-slash is allowed
+#   - c.f. https://github.com/cf-convention/cf-conventions/issues/307
+# However, clearly, we can't really handle spaces in names.
+# FOR NOW: allow a Python "normal word character", followed by anything except
+# the forward-slash and space.
+_VALID_NAME_REGEX = re.compile(r"^\w[^ /]*$")
 
 
 class Checker:
@@ -152,7 +159,7 @@ def check_mesh_attr_is_varlist(
                         "Mesh",
                         meshvar.name,
                         f'has {attrname}="{value}", '
-                        "which is not a valid list of netcdf variable names.",
+                        "which is not a valid list of variable names.",
                     )
                     success = False
             if success:
@@ -165,7 +172,7 @@ def check_mesh_attr_is_varlist(
                             "Mesh",
                             meshvar.name,
                             f'has {attrname}="{varname}", '
-                            "which is not a valid netcdf variable name.",
+                            "which is not a valid variable name.",
                         )
                         success = False
                     elif varname not in self._all_vars:
@@ -202,7 +209,7 @@ def var_ref_problem(self, attr_value: np.ndarray) -> str:
         if succeed:
             var_name = property_as_single_name(attr_value)
             if not _VALID_NAME_REGEX.match(var_name):
-                result = "is not a valid netcdf variable name"
+                result = "is not a valid variable name"
                 succeed = False
         if succeed:
             bounds_var = self._all_vars.get(var_name)

lib/ugrid_checks/tests/check/test_check_dataset__checks.py

@@ -432,7 +432,7 @@ def test_r105_r108_mesh_badcoordattr_empty(self, scan_2d_and_meshvar):
         meshvar.attributes["node_coordinates"] = ""
         msg = (
             '"topology" has node_coordinates="".*'
-            "is not a valid list of netcdf variable"
+            "is not a valid list of variable names"
         )
         self.check(
             scan,
@@ -449,7 +449,7 @@ def test_r105_r108_mesh_badcoordattr_invalidname(
         meshvar.attributes["node_coordinates"] = "$123"
         msg = (
             r'"topology" has node_coordinates="\$123"'
-            ".*not a valid netcdf variable name"
+            ".*not a valid variable name"
         )
         self.check(
             scan,
@@ -505,7 +505,7 @@ def test_r109_mesh_badconn_empty(self, scan_2d_and_meshvar):
                     "R105",
                     (
                         '"topology" has face_node_connectivity="".*'
-                        "not a valid list of netcdf variable"
+                        "not a valid list of variable names"
                     ),
                 ),
                 (
@@ -921,8 +921,8 @@ def test_r502_datavar_empty_mesh(self, scan_2d_and_datavar):
 
     def test_r502_datavar_invalid_mesh_name(self, scan_2d_and_datavar):
         scan, datavar = scan_2d_and_datavar
-        datavar.attributes["mesh"] = "$123"
-        msg = r'mesh="\$123", which is not a valid netcdf variable name\.'
+        datavar.attributes["mesh"] = "/bad/"
+        msg = r'mesh="/bad/", which is not a valid variable name\.'
         self.check(scan, "R502", msg)
 
     def test_r502_datavar_bad_mesh_dtype(self, scan_2d_and_datavar):
@@ -999,10 +999,10 @@ def test_r507_datavar_lis_with_location(self, scan_0d_with_lis_datavar):
     def test_r508_datavar_lis_invalid(self, scan_0d_with_lis_datavar):
         # The lis attribute should be a valid variable reference.
         scan, data_var = scan_0d_with_lis_datavar
-        data_var.attributes["location_index_set"] = "$123"
+        data_var.attributes["location_index_set"] = "/bad/"
         msg = (
-            r'location_index_set="\$123", '
-            r"which is not a valid netcdf variable name\."
+            r'location_index_set="/bad/", '
+            r"which is not a valid variable name\."
         )
         self.check(scan, "R508", msg)
 
@@ -1095,10 +1095,10 @@ def test_r203_coord_bounds_bad_attrdtype(self, scan_2d_and_coordvar):
     def test_r203_coord_bounds_bad_name(self, scan_2d_and_coordvar):
         scan, coord = scan_2d_and_coordvar
         # Add an invalid bounds attribute to the node_lon coord.
-        coord.attributes["bounds"] = "$123"
+        coord.attributes["bounds"] = "/bad/"
         msg = (
-            r'"node_lon" within topology:node_coordinates has bounds="\$123", '
-            "which is not a valid netcdf variable name"
+            r'"node_lon" within topology:node_coordinates has bounds="/bad/", '
+            r"which is not a valid variable name\."
         )
         self.check(scan, "R203", msg)
 

lib/ugrid_checks/tests/check/test_name_regex.py

@@ -0,0 +1,26 @@
+from ...check import _VALID_NAME_REGEX as var_re
+
+
+def test_valid_varnames():
+    # Just check a few things that should be accepted, and not
+    assert var_re.match("abc")
+    assert var_re.match("x")  # single char
+
+    # various unusual chars, cannot appear at start
+    nonstart_chars = r"#$£+-*^%?!.:;,\()[]{}"  # almost anything !!
+    for nonstart_char in nonstart_chars:
+        assert not var_re.match(nonstart_char)
+    # But these are all OK *after* the start position
+    assert var_re.match("x" + nonstart_chars)
+
+    # Examples of characters which *are* allowed at start
+    start_chars = "_10"  # NB includes digits.
+    for start_char in start_chars:
+        assert var_re.match(start_char)
+
+    # not empty
+    assert not var_re.match("")
+    # no spaces
+    assert not var_re.match("space in name")
+    # no forward-slash
+    assert not var_re.match("forward/slash")