diff --git a/databricks/koalas/frame.py b/databricks/koalas/frame.py
index 56fd71e19f..1086e1e1aa 100644
--- a/databricks/koalas/frame.py
+++ b/databricks/koalas/frame.py
@@ -10898,6 +10898,74 @@ def from_dict(data, orient="columns", dtype=None, columns=None) -> "DataFrame":
"""
return DataFrame(pd.DataFrame.from_dict(data, orient=orient, dtype=dtype, columns=columns))
+ def lookup(self, row_labels, col_labels) -> np.ndarray:
+ """
+ Label-based "fancy indexing" function for DataFrame.
+
+ Given equal-length arrays of row and column labels, return an
+ array of the values corresponding to each (row, col) pair.
+
+ `row_labels` and `col_labels` are not support the type `Series` and `Index`
+ to prevent performance degradation.
+
+ Parameters
+ ----------
+ row_labels : sequence
+ The row labels to use for lookup.
+ col_labels : sequence
+ The column labels to use for lookup.
+
+ Returns
+ -------
+ numpy.ndarray
+ The found values.
+
+ Examples
+ --------
+ >>> kdf = ks.DataFrame({'A': [3, 4, 5, 6, 7],
+ ... 'B': [10.0, 20.0, 30.0, 40.0, 50.0],
+ ... 'C': ['a', 'b', 'c', 'd', 'e']})
+ >>> kdf
+ A B C
+ 0 3 10.0 a
+ 1 4 20.0 b
+ 2 5 30.0 c
+ 3 6 40.0 d
+ 4 7 50.0 e
+
+ >>> kdf.lookup([0], ["C"])
+ array(['a'], dtype=object)
+
+ >>> kdf.lookup([2, 3], ["A", "B"])
+ array([ 5., 40.])
+ """
+ from databricks.koalas.series import Series
+ from databricks.koalas.indexes import Index, MultiIndex
+
+ if isinstance(row_labels, (Series, Index)):
+ raise TypeError(
+ "'row_labels' doesn't support type '{}'.".format(type(row_labels).__name__)
+ )
+ if isinstance(col_labels, (Series, Index)):
+ raise TypeError(
+ "'col_labels' doesn't support type '{}'.".format(type(col_labels).__name__)
+ )
+
+ if not isinstance(self.index, MultiIndex):
+ return (
+ self.loc[list(set(row_labels)), list(set(col_labels))]
+ .to_pandas()
+ .lookup(row_labels, col_labels)
+ )
+ else:
+ if len(row_labels) != len(col_labels):
+ raise ValueError("Row labels must have same size as column labels")
+ lookups = [
+ self.loc[row_label, col_label]
+ for row_label, col_label in zip(row_labels, col_labels)
+ ]
+ return np.asarray(pd.Series(lookups))
+
def _to_internal_pandas(self):
"""
Return a pandas DataFrame directly from _internal to avoid overhead of copy.
diff --git a/databricks/koalas/missing/frame.py b/databricks/koalas/missing/frame.py
index 5333e664c8..30fcc5b42f 100644
--- a/databricks/koalas/missing/frame.py
+++ b/databricks/koalas/missing/frame.py
@@ -53,7 +53,6 @@ class _MissingPandasLikeDataFrame(object):
insert = _unsupported_function("insert")
interpolate = _unsupported_function("interpolate")
last = _unsupported_function("last")
- lookup = _unsupported_function("lookup")
mode = _unsupported_function("mode")
reorder_levels = _unsupported_function("reorder_levels")
resample = _unsupported_function("resample")
diff --git a/databricks/koalas/tests/test_dataframe.py b/databricks/koalas/tests/test_dataframe.py
index 2482c24caa..02372f38dd 100644
--- a/databricks/koalas/tests/test_dataframe.py
+++ b/databricks/koalas/tests/test_dataframe.py
@@ -5051,6 +5051,63 @@ def test_from_dict(self):
kdf = ks.DataFrame.from_dict(data, orient="index", columns=["A", "B", "C", "D"])
self.assert_eq(pdf, kdf)
+ def test_lookup(self):
+ pdf = pd.DataFrame(
+ {
+ "A": [3, 4, 5, 6, 7],
+ "B": [10.0, 20.0, 30.0, 40.0, 50.0],
+ "C": ["a", "b", "c", "d", "e"],
+ }
+ )
+ kdf = ks.from_pandas(pdf)
+
+ # list
+ self.assert_eq(pdf.lookup([0], ["C"]), kdf.lookup([0], ["C"]))
+ self.assert_list_eq(
+ pdf.lookup([0, 3, 4], ["A", "C", "A"]), kdf.lookup([0, 3, 4], ["A", "C", "A"])
+ )
+
+ # tuple
+ self.assert_eq(pdf.lookup((0,), ("C",)), kdf.lookup((0,), ("C",)))
+ self.assert_list_eq(
+ pdf.lookup((0, 3, 4), ("A", "C", "A")), kdf.lookup((0, 3, 4), ("A", "C", "A"))
+ )
+
+ # dict
+ self.assert_eq(pdf.lookup({0: None}, {"C": None}), kdf.lookup({0: None}, {"C": None}))
+ self.assert_list_eq(
+ pdf.lookup({0: None, 3: None, 4: None}, {"A": None, "C": None, "B": None}),
+ kdf.lookup({0: None, 3: None, 4: None}, {"A": None, "C": None, "B": None}),
+ )
+
+ # MultiIndex
+ pdf.index = pd.MultiIndex.from_tuples(
+ [("a", "v"), ("b", "w"), ("c", "x"), ("d", "y"), ("e", "z")]
+ )
+ kdf = ks.from_pandas(pdf)
+
+ self.assert_eq(pdf.lookup([("a", "v")], ["C"]), kdf.lookup([("a", "v")], ["C"]))
+ self.assert_list_eq(
+ pdf.lookup([("a", "v"), ("d", "y"), ("e", "z")], ["A", "C", "A"]),
+ kdf.lookup([("a", "v"), ("d", "y"), ("e", "z")], ["A", "C", "A"]),
+ )
+
+ err_msg = "Row labels must have same size as column labels"
+ with self.assertRaisesRegex(ValueError, err_msg):
+ kdf.lookup([0, 3, 4], ["A", "C"])
+ err_msg = "'row_labels' doesn't support type 'Index'."
+ with self.assertRaisesRegex(TypeError, err_msg):
+ kdf.lookup(ks.Index([0]), ["C"])
+ err_msg = "'row_labels' doesn't support type 'Series'."
+ with self.assertRaisesRegex(TypeError, err_msg):
+ kdf.lookup(ks.Series([0]), ["C"])
+ err_msg = "'col_labels' doesn't support type 'Index'."
+ with self.assertRaisesRegex(TypeError, err_msg):
+ kdf.lookup([0], ks.Index(["C"]))
+ err_msg = "'col_labels' doesn't support type 'Series'."
+ with self.assertRaisesRegex(TypeError, err_msg):
+ kdf.lookup([0], ks.Series(["C"]))
+
def test_pad(self):
pdf = pd.DataFrame(
{
diff --git a/docs/source/reference/frame.rst b/docs/source/reference/frame.rst
index 613e9cbf1b..a1d3f8f06d 100644
--- a/docs/source/reference/frame.rst
+++ b/docs/source/reference/frame.rst
@@ -62,6 +62,7 @@ Indexing, iteration
DataFrame.items
DataFrame.iteritems
DataFrame.iterrows
+ DataFrame.lookup
DataFrame.itertuples
DataFrame.keys
DataFrame.pop