Change tests

scripts/make_resources.py

@@ -1,7 +1,14 @@
+import argparse
+
 import pandas as pd
 import seaborn as sns
 
-if __name__ == "__main__":
+
+class Inputs(argparse.Namespace):
+    make_benchmark_files: bool
+
+
+def main(args: Inputs):
     df = sns.load_dataset("titanic")
     X = df.select_dtypes("number").drop(columns=["survived"]).astype(float)
     y = df["survived"].astype(float)
@@ -29,24 +36,33 @@
         index=False,
         header=False,
     )
+    if args.make_benchmark_files:
+        dfb = df.sample(
+            100_000,
+            random_state=0,
+            replace=True,
+        ).reset_index(drop=True)
 
-    dfb = df.sample(
-        100_000,
-        random_state=0,
-        replace=True,
-    ).reset_index(drop=True)
+        Xb = dfb.select_dtypes("number").drop(columns=["survived"]).astype(float)
+        Xb = pd.concat([Xb] * 10, axis=0)
+        yb = dfb["survived"].astype(float)
+        print("benmark files sizes: {Xb.shape}, {y.shape}")
 
-    Xb = dfb.select_dtypes("number").drop(columns=["survived"]).astype(float)
-    yb = dfb["survived"].astype(float)
+        pd.Series(Xb.fillna(0).to_numpy().ravel(order="F")).to_csv(
+            "resources/contiguous_no_missing_100k_samp_seed0.csv",
+            index=False,
+            header=False,
+        )
 
-    pd.Series(Xb.fillna(0).to_numpy().ravel(order="F")).to_csv(
-        "resources/contiguous_no_missing_100k_samp_seed0.csv",
-        index=False,
-        header=False,
-    )
+        yb.to_csv(
+            "resources/performance_100k_samp_seed0.csv",
+            index=False,
+            header=False,
+        )
 
-    yb.to_csv(
-        "resources/performance_100k_samp_seed0.csv",
-        index=False,
-        header=False,
-    )
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--make-benchmark-files", "-mbf", action="store_true")
+    args = parser.parse_args(namespace=Inputs())
+    main(args)

src/histogram.rs

@@ -89,16 +89,8 @@ pub fn create_feature_histogram(
     sorted_hess: &[f32],
     index: &[usize],
 ) -> Vec<Bin<f32>> {
-    let mut histogram: Vec<Bin<f32>> = Vec::with_capacity(cuts.len());
-
-    let mut gradient_sums: Vec<f64> = Vec::with_capacity(cuts.len());
-    let mut hessian_sums: Vec<f64> = Vec::with_capacity(cuts.len());
-
-    histogram.push(Bin::new_f32(f64::NAN));
-    // The last cut value is simply the maximum possible value, so we don't need it.
-    // This value is needed initially for binning, but we don't need to count it as
-    // a histogram bin.
-    histogram.extend(cuts[..(cuts.len() - 1)].iter().map(|c| Bin::new_f32(*c)));
+    let mut gradient_sums: Vec<f64> = vec![0.; cuts.len()];
+    let mut hessian_sums: Vec<f64> = vec![0.; cuts.len()];
 
     index
         .iter()
@@ -112,14 +104,21 @@ pub fn create_feature_histogram(
                 *v += f64::from(*h);
             }
         });
+
+    // The first value si reserved for missing.
     // The last cut value is simply the maximum possible value, so we don't need it.
     // This value is needed initially for binning, but we don't need to count it as
     // a histogram bin.
-    histogram
-        .iter_mut()
+    Some(&f64::NAN)
+        .into_iter()
+        .chain(cuts[..(cuts.len() - 1)].iter())
         .zip(gradient_sums)
         .zip(hessian_sums)
-        .map(|((hist, g), h)| hist.update(g, h))
+        .map(|((c, g), h)| Bin {
+            gradient_sum: g as f32,
+            hessian_sum: h as f32,
+            cut_value: *c,
+        })
         .collect()
 }