我是streamlit的新手。我正在通过阅读电子表格（~ 200万行）来创建一个web服务。
我创建了两个下拉选择使用列“PRODUCTID”，“PRODUCTNDC”从表中.然而，当用户选择的标识符从下拉，它是阅读一个字符串再次，这使得它非常慢.我如何才能提高这个服务的速度？
下面是代码：

#import libraries
import pandas as pd
import numpy as np
import openpyxl
import streamlit as st
from streamlit_extras.app_logo import add_logo

#page settings
st.set_page_config(page_title = 'FDA-NDC Code Service', layout="wide")
st.header('FDA-NDC Code Service')
#st.subheader('beta version')

#data reading & pre-processing
df = pd.read_excel(io='package.xlsx', engine='openpyxl', sheet_name= 'package')

#removing leading spaces and more than 1 space in column names
df = df[["PRODUCTID", "PRODUCTNDC", "NDCPACKAGECODE", "PACKAGEDESCRIPTION",  "STARTMARKETINGDATE", "ENDMARKETINGDATE", "NDC_EXCLUDE_FLAG", "SAMPLE_PACKAGE"]]
df['PRODUCTID'] = df['PRODUCTID'].str.strip()

product_list = list(df['PRODUCTID'].unique())

df6 = df.copy()

options_pid = df6['PRODUCTID'].unique().tolist()
selected_pid = st.sidebar.multiselect('Search product id',options_pid)

options_nid = df6['PRODUCTNDC'].unique().tolist()
selected_nid = st.sidebar.multiselect('Search product ndc',options_nid)

if selected_pid:
    df6 = df6[df6["PRODUCTID"].isin(selected_pid)]
    
elif selected_nid:
    df6 = df6[df6["PRODUCTNDC"].isin(selected_nid)]
    
else:
    st.markdown(
              f'<p class="header_title"> {str(df6.shape[0])} </p>',
              unsafe_allow_html=True,
          )

if selected_pid or selected_nid:
    st.dataframe(df6.reset_index(drop = True))
else:
    st.dataframe(df)

字符串
任何帮助都非常感谢。