Add datasets comparison

change_visualizer.py

@@ -1,6 +1,7 @@
 import gradio as gr
 
 import generate_annotated_diffs
+import statistics
 
 df_manual = generate_annotated_diffs.manual_data_with_annotated_diffs()
 n_diffs_manual = len(df_manual)
@@ -8,6 +9,9 @@ n_diffs_manual = len(df_manual)
 df_synthetic = generate_annotated_diffs.synthetic_data_with_annotated_diffs()
 n_diffs_synthetic = len(df_synthetic)
 
+STATISTICS = {"manual": statistics.get_statistics_for_manual_df(df_manual),
+              "synthetic": statistics.get_statistics_for_synthetic_df(df_synthetic)}
+
 
 def update_manual_view(diff_idx):
     diff_idx -= 1
@@ -74,6 +78,24 @@ if __name__ == '__main__':
 
             slider_synthetic.change(update_synthetic_view, inputs=slider_synthetic,
                                     outputs=view_synthetic)
+        with gr.Tab("Compare"):
+            def layout_for_statistics(statistics_group_name):
+                gr.Markdown(f"### {statistics_group_name}")
+                stats = STATISTICS[statistics_group_name]
+                gr.Number(label="Average deletions number (rel to the result length)", interactive=False,
+                          value=stats['deletions'].mean().item(), precision=3)
+                gr.Number(label="Average insertions number (rel to the result length)", interactive=False,
+                          value=stats['insertions'].mean().item(), precision=3)
+                gr.Number(label="Average changes number (rel to the result length)", interactive=False,
+                          value=stats['changes'].mean().item(), precision=3)
+
+
+            with gr.Row():
+                with gr.Column(scale=1):
+                    layout_for_statistics("manual")
+
+                with gr.Column(scale=1):
+                    layout_for_statistics("synthetic")
 
         application.load(update_manual_view, inputs=slider_manual,
                          outputs=view_manual)

statistics.py

@@ -0,0 +1,38 @@
+import numpy as np
+import pandas as pd
+
+
+def get_statistics_for_df(df: pd.DataFrame, end_col, annotated_col):
+    relative_deletions = []
+    relative_insertions = []
+    relative_changes = []
+
+    for _, row in df.iterrows():
+        sum_deletions = 0
+        sum_insertions = 0
+        for text, change_type in row[annotated_col]:
+            if change_type == '-':
+                sum_deletions += len(text)
+            elif change_type == '+':
+                sum_insertions += len(text)
+
+        sum_changes = sum_deletions + sum_insertions
+        end_length = len(row[end_col])
+
+        relative_deletions.append(sum_deletions / end_length)
+        relative_insertions.append(sum_insertions / end_length)
+        relative_changes.append(sum_changes / end_length)
+
+    return {
+        "deletions": np.asarray(relative_deletions),
+        "insertions": np.asarray(relative_insertions),
+        "changes": np.asarray(relative_changes)
+    }
+
+
+def get_statistics_for_manual_df(df):
+    return get_statistics_for_df(df, end_col='commit_msg_end', annotated_col='annotated_diff')
+
+
+def get_statistics_for_synthetic_df(df):
+    return get_statistics_for_df(df, end_col='reference', annotated_col='annotated_diff')