TabularData: Add dropna method.

This method mimics pandas `dropna`.

dvc/compare.py

@@ -12,6 +12,7 @@ from typing import (
     Mapping,
     MutableSequence,
     Sequence,
+    Set,
     Tuple,
     Union,
     overload,
@@ -181,6 +182,35 @@ class TabularData(MutableSequence[Sequence["CellT"]]):
             {k: self._columns[k][i] for k in keys} for i in range(len(self))
         ]
 
+    def dropna(self, axis: str = "rows"):
+        if axis not in ["rows", "cols"]:
+            raise ValueError(
+                f"Invalid 'axis' value {axis}."
+                "Choose one of ['rows', 'cols']"
+            )
+
+        to_drop: Set = set()
+        for n_row, row in enumerate(self):
+            for n_col, col in enumerate(row):
+                if col == self._fill_value:
+                    if axis == "rows":
+                        to_drop.add(n_row)
+                        break
+                    else:
+                        to_drop.add(self.keys()[n_col])
+
+        if axis == "rows":
+            for name in self.keys():
+                self._columns[name] = Column(
+                    [
+                        x
+                        for n, x in enumerate(self._columns[name])
+                        if n not in to_drop
+                    ]
+                )
+        else:
+            self.drop(*to_drop)
+
 
 def _normalize_float(val: float, precision: int):
     return f"{val:.{precision}g}"

tests/unit/test_tabular_data.py

@@ -177,3 +177,36 @@ def test_row_from_dict():
         ["", "", "value3", "value4", "", ""],
         ["", "", "value3", "", "value5", "value6"],
     ]
+
+
+@pytest.mark.parametrize(
+    "axis,expected",
+    [
+        (
+            "rows",
+            [
+                ["foo", "bar", "foobar"],
+            ],
+        ),
+        ("cols", [["foo"], ["foo"], ["foo"]]),
+    ],
+)
+def test_dropna(axis, expected):
+    td = TabularData(["col-1", "col-2", "col-3"])
+    td.extend([["foo"], ["foo", "bar"], ["foo", "bar", "foobar"]])
+    assert list(td) == [
+        ["foo", "", ""],
+        ["foo", "bar", ""],
+        ["foo", "bar", "foobar"],
+    ]
+
+    td.dropna(axis)
+
+    assert list(td) == expected
+
+
+def test_dropna_invalid_axis():
+    td = TabularData(["col-1", "col-2", "col-3"])
+
+    with pytest.raises(ValueError, match="Invalid 'axis' value foo."):
+        td.dropna("foo")