perf(streamlit): migrate from pandas to polars

machine-learning/hm-streamlit/applications/live-line-chart/pyproject.toml

@@ -4,7 +4,7 @@ version = "1.0.0"
 requires-python = "~=3.13.0"
 dependencies = [
   "numpy==2.2.1",
-  "pandas==2.2.3",
+  "polars==1.18.0",
   "streamlit==1.41.1",
 ]
 

machine-learning/hm-streamlit/applications/live-line-chart/src/main.py

@@ -2,39 +2,56 @@
 from datetime import datetime
 
 import numpy as np
-import pandas as pd
+import polars as pl
 import streamlit as st
 
 
 @st.cache_data
-def get_data() -> pd.DataFrame:
-    return pd.DataFrame(columns=["value1", "value2"])
+def get_data() -> pl.DataFrame:
+    return pl.DataFrame(
+        {"timestamp": [], "value1": [], "value2": []},
+        schema={"timestamp": pl.Datetime, "value1": pl.Float64, "value2": pl.Float64},
+    )
 
 
 def main() -> None:
     st.title("Live Line Chart")
     generator = np.random.default_rng(42)
     max_data_points = 100
-    prev_time = None
+    prev_values = None
     df = get_data()
     placeholder = st.empty()
+
     while True:
         current_time = datetime.now()
         new_data_point = (
-            generator.standard_normal(2) / 10.0 + df.loc[prev_time]
-            if prev_time
+            generator.standard_normal(2) / 10.0 + prev_values
+            if prev_values is not None
             else generator.standard_normal(2)
         )
-        df.loc[current_time] = new_data_point
-        prev_time = current_time
+        prev_values = new_data_point
+
+        # Add new row to DataFrame
+        df = pl.concat(
+            [
+                df,
+                pl.DataFrame(
+                    {
+                        "timestamp": [current_time],
+                        "value1": [new_data_point[0]],
+                        "value2": [new_data_point[1]],
+                    }
+                ),
+            ]
+        )
 
         # Remove old timestamps if the DataFrame exceeds the maximum size
         if len(df) > max_data_points:
-            df = df.iloc[1:]
+            df = df.slice(1, len(df))
 
         with placeholder.container():
             st.header("Chart")
-            st.line_chart(df, height=200)
+            st.line_chart(df.select(["value1", "value2"]), height=200)
 
             st.header("Table")
             st.dataframe(df)