handle patching with NaT values (#5675)

panel/tests/widgets/test_tables.py

@@ -18,6 +18,7 @@
 from panel.io.state import set_curdoc
 from panel.models.tabulator import CellClickEvent, TableEditEvent
 from panel.tests.util import serve_and_request, wait_until
+from panel.util import BOKEH_JS_NAT
 from panel.widgets import Button, TextInput
 from panel.widgets.tables import DataFrame, Tabulator
 
@@ -1295,6 +1296,28 @@ def test_tabulator_patch_with_timestamp(document, comm):
         if col != 'index':
             np.testing.assert_array_equal(table.value[col].values, expected[col])
 
+def test_tabulator_patch_with_NaT(document, comm):
+    df = pd.DataFrame(dict(A=pd.to_datetime(['1980-01-01', np.nan])))
+    assert df.loc[1, 'A'] is pd.NaT
+    table = Tabulator(df)
+
+    model = table.get_root(document, comm)
+
+    table.patch({'A': [(0, pd.NaT)]})
+
+    # We're also checking that the NaT value that was in the original table
+    # at .loc[1, 'A'] is converted in the model as BOKEH_JS_NAT.
+    expected = {
+        'index': np.array([0, 1]),
+        'A': np.array([BOKEH_JS_NAT, BOKEH_JS_NAT])
+    }
+    for col, values in model.source.data.items():
+        expected_array = expected[col]
+        np.testing.assert_array_equal(values, expected_array)
+        # Not checking that the data in table.value is the same as expected
+        # In table.value we have NaT values, in expected the BOKEH_JS_NAT constant.
+
+
 def test_tabulator_stream_series_paginated_not_follow(document, comm):
     df = makeMixedDataFrame()
     table = Tabulator(df, pagination='remote', page_size=2)

panel/widgets/tables.py

@@ -29,7 +29,8 @@
 from ..io.state import state
 from ..reactive import Reactive, ReactiveData
 from ..util import (
-    clone_model, datetime_as_utctimestamp, isdatetime, lazy_load, updating,
+    BOKEH_JS_NAT, clone_model, datetime_as_utctimestamp, isdatetime, lazy_load,
+    updating,
 )
 from .base import Widget
 from .button import Button
@@ -824,6 +825,8 @@ def patch(self, patch_value, as_index=True):
                         self.value.iloc[data_ind, columns.index(k)] = value
                     if isinstance(value, pd.Timestamp):
                         value = datetime_as_utctimestamp(value)
+                    elif value is pd.NaT:
+                        value = BOKEH_JS_NAT
                     values.append((patch_ind, value))
                 patches[k] = values
             self._patch(patches)