Added polars_performance.py

benchmarks/dataframe_performance.cc

@@ -29,53 +29,54 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include <DataFrame/DataFrameStatsVisitors.h>
 #include <DataFrame/RandGen.h>
 
+#include <chrono>
 #include <iostream>
 
 using namespace hmdf;
 
-constexpr std::size_t   ALIGNMENT = 256;
+constexpr std::size_t   ALIGNMENT = 64;
+constexpr std::size_t   SIZE = 100000000;
 
-typedef StdDataFrame256<time_t> MyDataFrame;
+typedef StdDataFrame64<time_t> MyDataFrame;
 
 // -----------------------------------------------------------------------------
 
-int main(int, char *[]) {
+using namespace std::chrono;
 
-    std::cout << "Starting ... " << std::endl;
+int main(int, char *[]) {
 
-    const auto  first = time(nullptr);
-    auto        index_vec =
-        MyDataFrame::gen_datetime_index("01/01/1970", "08/15/2019",
-                                        time_frequency::secondly, 1);
-    const auto  index_sz = index_vec.size();
+    const auto  first = high_resolution_clock::now();
     MyDataFrame df;
 
     df.load_data(
-        std::move(index_vec),
-        std::make_pair("normal", gen_normal_dist<double, ALIGNMENT>(index_sz)),
-        std::make_pair("log_normal", gen_lognormal_dist<double, ALIGNMENT>(index_sz)),
-        std::make_pair("exponential", gen_exponential_dist<double, ALIGNMENT>(index_sz)));
+        MyDataFrame::gen_sequence_index(0, SIZE, 1),
+        std::make_pair("normal", gen_normal_dist<double, ALIGNMENT>(SIZE)),
+        std::make_pair("log_normal", gen_lognormal_dist<double, ALIGNMENT>(SIZE)),
+        std::make_pair("exponential", gen_exponential_dist<double, ALIGNMENT>(SIZE)));
 
-    const auto  second = time(nullptr);
+    const auto  second = high_resolution_clock::now();
 
     std::cout << "All data loadings are done. Calculating means ... "
-              << second - first << std::endl;
+              << double(duration_cast<microseconds>(second - first).count()) / 1000000.0
+              << std::endl;
 
     MeanVisitor<double, time_t> n_mv;
-    MeanVisitor<double, time_t> ln_mv;
-    MeanVisitor<double, time_t> e_mv;
+    VarVisitor<double, time_t>  ln_vv;
+    CorrVisitor<double, time_t> e_ln_cv;
 
     auto    fut1 = df.visit_async<double>("normal", n_mv);
-    auto    fut2 = df.visit_async<double>("log_normal", ln_mv);
-    auto    fut3 = df.visit_async<double>("exponential", e_mv);
+    auto    fut2 = df.visit_async<double>("log_normal", ln_vv);
+    auto    fut3 = df.visit_async<double, double>("exponential", "log_normal", e_ln_cv);
 
     std::cout << fut1.get().get_result() << ", "
               << fut2.get().get_result() << ", "
               << fut3.get().get_result() << std::endl;
 
-    const auto  third = time(nullptr);
+    const auto  third = high_resolution_clock::now();
 
-    std::cout << third - second << ", " << third - first
+    std::cout << double(duration_cast<microseconds>(third - second).count()) / 1000000.0
+              << ", "
+              << double(duration_cast<microseconds>(third - first).count()) / 1000000.0
               << " All done" << std::endl;
     return (0);
 }

benchmarks/polars_performance.py

@@ -0,0 +1,35 @@
+import datetime
+import numpy as np
+import polars as pl
+
+# ------------------------------------------------------------------------------
+
+SIZE: int = 100000000
+
+first = datetime.datetime.now()
+df = pl.DataFrame({"normal": np.random.normal(size=SIZE),
+                   "log_normal": np.random.lognormal(size=SIZE),
+                   "exponential": np.random.exponential(size=SIZE),
+                   })
+second = datetime.datetime.now()
+print(f"All data loadings are done. Calculating means ... "
+      f"{(second - first).seconds}.{(second - first).microseconds}")
+
+m1: float = df["normal"].mean()
+m2: float = df["log_normal"].var()
+m3: float = df.select(pl.corr("exponential", "log_normal"))
+
+print(f"{m1}, {m2}, {m3}")
+third = datetime.datetime.now()
+
+
+print(f"{(third - second).seconds}.{(third - second).microseconds}, "
+      f"{(third - first).seconds}.{(third - first).microseconds} All done");
+
+# ------------------------------------------------------------------------------
+
+# Local Variables:
+# mode:Python
+# tab-width:4
+# c-basic-offset:4
+# End: