Add register_fields method

doc/source/analyzing/fields.rst

@@ -122,6 +122,28 @@ field, like its default units or the source code for it.
    print(ds.field_info["gas", "pressure"].get_units())
    print(ds.field_info["gas", "pressure"].get_source())
 
+Defining properties for on-disk fields
+--------------------------------------
+
+Occasionally a dataset will have extra fields that are not defined in its
+associated frontend. When this is the case, those fields will not be
+assigned units, aliases, or a properly formatted display name. These properties
+can be defined manually by calling
+:meth:`~yt.data_objects.static_output.Dataset.register_field`.
+
+.. code-block:: python
+
+   ds = yt.load("DD1234/DD1234")
+   ds.register_field(("enzo", "DinosaurDensity"), units="code_mass/code_length**3")
+
+This can also be used to override properties defined by the frontend. Note, this
+must be done immediately after loading the dataset (i.e., before the ``field_list``
+has been populated).
+
+To do this within a configuration file, see :ref:`per-field-config`. However, note
+that the configuration file method will only operate when the field is not defined
+by the frontend; it will not override.
+
 Using fields to access data
 ---------------------------
 

yt/data_objects/static_output.py

@@ -180,6 +180,7 @@ class Dataset(abc.ABC):
     _ionization_label_format = "roman_numeral"
     _determined_fields: dict[str, list[FieldKey]] | None = None
     fields_detected = False
+    _registered_fields = None
 
     # these are set in self._parse_parameter_file()
     domain_left_edge = MutableAttribute(True)
@@ -1716,6 +1717,47 @@ def quan(self):
         self._quan = functools.partial(YTQuantity, registry=self.unit_registry)
         return self._quan
 
+    def register_field(self, name, units=None, aliases=None, display_name=None):
+        """
+        Register properties for an on-disk field.
+
+        Register or override units, aliases, or the display name for an
+        on-disk field.
+
+        Note, this must be called immediately after yt.load (i.e., before
+        the list of fields is generated).
+
+        Parameters
+        ----------
+
+        name : str
+           name of the field
+        units : str
+           units of the field
+        aliases : list
+           a list of alias names
+        display_name: str
+           the name used in plots
+
+        """
+        if self._instantiated_index:
+            raise RuntimeError(
+                "Cannot call register_field after field_list is populated. "
+                "This must be called immediately after load."
+            )
+        if self._registered_fields is None:
+            self._registered_fields = {}
+
+        entry = {}
+        if units is not None:
+            entry["units"] = units
+        if aliases is not None:
+            entry["aliases"] = aliases
+        if display_name is not None:
+            entry["display_name"] = display_name
+
+        self._registered_fields[name] = entry
+
     def add_field(
         self, name, function, sampling_type, *, force_override=False, **kwargs
     ):

yt/fields/field_info_container.py

@@ -150,7 +150,13 @@ def setup_particle_fields(self, ptype, ftype="gas", num_neighbors=64):
                 raise RuntimeError
             if field[0] not in self.ds.particle_types:
                 continue
+
             units = self.ds.field_units.get(field, None)
+            rfields = self.ds._registered_fields
+            if rfields is not None:
+                if entry := rfields.get(field):
+                    units = entry.get("units", units)
+
             if units is None:
                 try:
                     units = ytcfg.get("fields", *field, "units")
@@ -256,6 +262,7 @@ def setup_fluid_aliases(self, ftype: FieldType = "gas") -> None:
                 raise RuntimeError
             if field[0] in self.ds.particle_types:
                 continue
+
             args = known_other_fields.get(field[1], None)
             if args is not None:
                 units, aliases, display_name = args
@@ -273,6 +280,15 @@ def setup_fluid_aliases(self, ftype: FieldType = "gas") -> None:
             # field *name* is in there, then the field *tuple*.
             units = self.ds.field_units.get(field[1], units)
             units = self.ds.field_units.get(field, units)
+
+            # allow user to override with call to ds.register_fields
+            rfields = self.ds._registered_fields
+            if rfields is not None:
+                if entry := rfields.get(field):
+                    units = entry.get("units", units)
+                    aliases = entry.get("aliases", aliases)
+                    display_name = entry.get("display_name", display_name)
+
             self.add_output_field(
                 field, sampling_type="cell", units=units, display_name=display_name
             )

yt/fields/tests/test_fields.py

@@ -504,6 +504,18 @@ def test_deposit_amr():
         assert_allclose_units(gpm, dpm)
 
 
+@requires_module("h5py")
+@requires_file(ISOGAL)
+def test_register_field():
+    field = ("enzo", "Phi_pField")
+    # this is made up
+    unit_string = "Msun*code_length/kg"
+    ds = load(ISOGAL)
+    ds.register_field(field, units=unit_string)
+    ds.field_list
+    assert ds.field_info[field].units == unit_string
+
+
 def test_ion_field_labels():
     fields = [
         "O_p1_number_density",