PERF: sparse to_csv

Improves to_csv performance for sparse matric by casting to dense
before initializing DataFrameFormatter. Results in many fewer calls to
`to_native_types` which saves time.

Author: rtlee9

asv_bench/benchmarks/io/csv.py

@@ -172,6 +172,26 @@ def time_head_of_multiindex(self):
         self.df_custom_index_then_head.to_csv(self.fname)
 
 
+class ToCSVSparse(BaseIO):
+
+    fname = "__test__.csv"
+
+    def setup(self):
+        from scipy import sparse as sc
+
+        vals = np.random.randint(0, 10, size=(10000, 1000))
+        keep = vals > 3
+        vals[keep] = 0
+        sparse_mtx = sc.coo_matrix(vals)
+        self.data = DataFrame.sparse.from_spmatrix(sparse_mtx)
+
+    def time_sparse_to_csv(self):
+        self.data.to_csv("sparse_pd.csv")
+
+    def time_sparse_to_dense_to_csv(self):
+        self.data.sparse.to_dense().to_csv("sparse_pd.csv")
+
+
 class StringIORewind:
     def data(self, stringio_object):
         stringio_object.seek(0)

doc/source/whatsnew/v2.0.0.rst

@@ -158,6 +158,7 @@ Performance improvements
 - Performance improvements to :func:`read_sas` (:issue:`47403`, :issue:`47405`, :issue:`47656`, :issue:`48502`)
 - Memory improvement in :meth:`RangeIndex.sort_values` (:issue:`48801`)
 - Performance improvement in :class:`DataFrameGroupBy` and :class:`SeriesGroupBy` when ``by`` is a categorical type and ``sort=False`` (:issue:`48976`)
+- Performance improvement for :meth:`NDFrame.to_csv` when data frame is sparse (:issue:`41023`)
 
 .. ---------------------------------------------------------------------------
 .. _whatsnew_200.bug_fixes:

pandas/core/generic.py

@@ -3717,6 +3717,8 @@ def to_csv(
         >>> df.to_csv('folder/subfolder/out.csv')  # doctest: +SKIP
         """
         df = self if isinstance(self, ABCDataFrame) else self.to_frame()
+        if hasattr(df, "sparse"):
+            df = df.sparse.to_dense()  # fixes 41023
 
         formatter = DataFrameFormatter(
             frame=df,