BUG: fix to_latex() when using MultiIndex with NaN in (#20797)

closes #14249
closes #19981
closes #18326
closes #18669

Author: tomneep

doc/source/whatsnew/v0.23.0.txt

@@ -1156,6 +1156,10 @@ I/O
 - Bug in :func:`read_csv` causing heap corruption on 32-bit, big-endian architectures (:issue:`20785`)
 - Bug in :func:`read_sas` where a file with 0 variables gave an ``AttributeError`` incorrectly. Now it gives an ``EmptyDataError`` (:issue:`18184`)
 - Bug in :func:`DataFrame.to_latex()` where pairs of braces meant to serve as invisible placeholders were escaped (:issue:`18667`)
+- Bug in :func:`DataFrame.to_latex()` where a ``NaN`` in a ``MultiIndex`` would cause an ``IndexError`` or incorrect output (:issue:`14249`)
+- Bug in :func:`DataFrame.to_latex()` where a non-string index-level name would result in an ``AttributeError`` (:issue:`19981`)
+- Bug in :func:`DataFrame.to_latex()` where the combination of an index name and the `index_names=False` option would result in incorrect output (:issue:`18326`)
+- Bug in :func:`DataFrame.to_latex()` where a ``MultiIndex`` with an empty string as its name would result in incorrect output (:issue:`18669`)
 - Bug in :func:`read_json` where large numeric values were causing an ``OverflowError`` (:issue:`18842`)
 - Bug in :func:`DataFrame.to_parquet` where an exception was raised if the write destination is S3 (:issue:`19134`)
 - :class:`Interval` now supported in :func:`DataFrame.to_excel` for all Excel file types (:issue:`19242`)

pandas/io/formats/latex.py

@@ -64,35 +64,32 @@ def get_col_type(dtype):
 
         # reestablish the MultiIndex that has been joined by _to_str_column
         if self.fmt.index and isinstance(self.frame.index, MultiIndex):
+            out = self.frame.index.format(
+                adjoin=False, sparsify=self.fmt.sparsify,
+                names=self.fmt.has_index_names, na_rep=self.fmt.na_rep
+            )
+
+            # index.format will sparsify repeated entries with empty strings
+            # so pad these with some empty space
+            def pad_empties(x):
+                for pad in reversed(x):
+                    if pad:
+                        break
+                return [x[0]] + [i if i else ' ' * len(pad) for i in x[1:]]
+            out = (pad_empties(i) for i in out)
+
+            # Add empty spaces for each column level
             clevels = self.frame.columns.nlevels
-            strcols.pop(0)
-            name = any(self.frame.index.names)
-            cname = any(self.frame.columns.names)
-            lastcol = self.frame.index.nlevels - 1
-            previous_lev3 = None
-            for i, lev in enumerate(self.frame.index.levels):
-                lev2 = lev.format()
-                blank = ' ' * len(lev2[0])
-                # display column names in last index-column
-                if cname and i == lastcol:
-                    lev3 = [x if x else '{}' for x in self.frame.columns.names]
-                else:
-                    lev3 = [blank] * clevels
-                if name:
-                    lev3.append(lev.name)
-                current_idx_val = None
-                for level_idx in self.frame.index.labels[i]:
-                    if ((previous_lev3 is None or
-                        previous_lev3[len(lev3)].isspace()) and
-                            lev2[level_idx] == current_idx_val):
-                        # same index as above row and left index was the same
-                        lev3.append(blank)
-                    else:
-                        # different value than above or left index different
-                        lev3.append(lev2[level_idx])
-                        current_idx_val = lev2[level_idx]
-                strcols.insert(i, lev3)
-                previous_lev3 = lev3
+            out = [[' ' * len(i[-1])] * clevels + i for i in out]
+
+            # Add the column names to the last index column
+            cnames = self.frame.columns.names
+            if any(cnames):
+                new_names = [i if i else '{}' for i in cnames]
+                out[self.frame.index.nlevels - 1][:clevels] = new_names
+
+            # Get rid of old multiindex column and add new ones
+            strcols = out + strcols[1:]
 
         column_format = self.column_format
         if column_format is None:
@@ -118,7 +115,7 @@ def get_col_type(dtype):
         ilevels = self.frame.index.nlevels
         clevels = self.frame.columns.nlevels
         nlevels = clevels
-        if any(self.frame.index.names):
+        if self.fmt.has_index_names and self.fmt.show_index_names:
             nlevels += 1
         strrows = list(zip(*strcols))
         self.clinebuf = []

pandas/tests/io/formats/test_to_latex.py

@@ -621,3 +621,74 @@ def test_to_latex_multiindex_names(self, name0, name1, axes):
 \end{tabular}
 """ % tuple(list(col_names) + [idx_names_row])
         assert observed == expected
+
+    @pytest.mark.parametrize('one_row', [True, False])
+    def test_to_latex_multiindex_nans(self, one_row):
+        # GH 14249
+        df = pd.DataFrame({'a': [None, 1], 'b': [2, 3], 'c': [4, 5]})
+        if one_row:
+            df = df.iloc[[0]]
+        observed = df.set_index(['a', 'b']).to_latex()
+        expected = r"""\begin{tabular}{llr}
+\toprule
+    &   &  c \\
+a & b &    \\
+\midrule
+NaN & 2 &  4 \\
+"""
+        if not one_row:
+            expected += r"""1.0 & 3 &  5 \\
+"""
+        expected += r"""\bottomrule
+\end{tabular}
+"""
+        assert observed == expected
+
+    def test_to_latex_non_string_index(self):
+        # GH 19981
+        observed = pd.DataFrame([[1, 2, 3]] * 2).set_index([0, 1]).to_latex()
+        expected = r"""\begin{tabular}{llr}
+\toprule
+  &   &  2 \\
+0 & 1 &    \\
+\midrule
+1 & 2 &  3 \\
+  & 2 &  3 \\
+\bottomrule
+\end{tabular}
+"""
+        assert observed == expected
+
+    def test_to_latex_midrule_location(self):
+        # GH 18326
+        df = pd.DataFrame({'a': [1, 2]})
+        df.index.name = 'foo'
+        observed = df.to_latex(index_names=False)
+        expected = r"""\begin{tabular}{lr}
+\toprule
+{} &  a \\
+\midrule
+0 &  1 \\
+1 &  2 \\
+\bottomrule
+\end{tabular}
+"""
+
+        assert observed == expected
+
+    def test_to_latex_multiindex_empty_name(self):
+        # GH 18669
+        mi = pd.MultiIndex.from_product([[1, 2]], names=[''])
+        df = pd.DataFrame(-1, index=mi, columns=range(4))
+        observed = df.to_latex()
+        expected = r"""\begin{tabular}{lrrrr}
+\toprule
+  &  0 &  1 &  2 &  3 \\
+{} &    &    &    &    \\
+\midrule
+1 & -1 & -1 & -1 & -1 \\
+2 & -1 & -1 & -1 & -1 \\
+\bottomrule
+\end{tabular}
+"""
+        assert observed == expected