import streamlit as st
import pandas as pd
import re
import os
import base64
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
import math

# Realistic placeholder dataframe (added Abstract field)
data = {
    "Title": [
        "The impact of climate change on biodiversity",
        "Deep learning algorithms for image classification",
        "Quantum computing and its applications in cryptography",
        "Machine learning approaches for natural language processing",
        "Modeling the effects of climate change on agricultural production",
        "Graph neural networks for social network analysis",
        "Biodiversity conservation strategies in the face of climate change",
        "Exploring the potential of quantum computing in drug discovery",
        "A survey of reinforcement learning algorithms and applications",
        "The role of artificial intelligence in combating climate change",
    ]*10,
    "Authors": [
        "Smith, J.; Doe, J.; Brown, M.",
        "Garcia, L.; Johnson, N.; Patel, K.",
        "Kim, D.; Taylor, R.; Yamamoto, Y.",
        "Roberts, A.; Jackson, T.; Davis, M.",
        "Turner, B.; Adams, C.; Evans, D.",
        "Baker, E.; Stewart, F.; Roberts, G.",
        "Nelson, H.; Mitchell, I.; Cooper, J.",
        "Parker, K.; Lewis, L.; Jenkins, M.",
        "Edwards, N.; Harrison, O.; Simmons, P.",
        "Fisher, Q.; Grant, R.; Turner, S.",
    ]*10,
    "Year": [2020, 2019, 2018, 2021, 2019, 2020, 2018, 2021, 2019, 2020]*10,
    "Keywords": [
        "climate change, biodiversity, ecosystems",
        "deep learning, image classification, convolutional neural networks",
        "quantum computing, cryptography, Shor's algorithm",
        "machine learning, natural language processing, text analysis",
        "climate change, agriculture, crop modeling",
        "graph neural networks, social network analysis, machine learning",
        "biodiversity conservation, climate change, environmental management",
        "quantum computing, drug discovery, computational chemistry",
        "reinforcement learning, algorithms, applications",
        "artificial intelligence, climate change, mitigation strategies",
    ]*10,
    "Subject_Area": [
        "Environmental Science",
        "Computer Science",
        "Physics",
        "Computer Science",
        "Environmental Science",
        "Computer Science",
        "Environmental Science",
        "Physics",
        "Computer Science",
        "Environmental Science",
    ]*10,
    "Journal": [
        "Nature",
        "IEEE Transactions on Pattern Analysis and Machine Intelligence",
        "Physical Review Letters",
        "Journal of Machine Learning Research",
        "Agricultural Systems",
        "IEEE Transactions on Neural Networks and Learning Systems",
        "Conservation Biology",
        "Journal of Chemical Information and Modeling",
        "Neural Computing and Applications",
        "Science",
    ]*10,
    "Is_Open_Access": [True, False, True, False, True, False, True, False, True, False]*10,
    "Abstract": [
        "This study analyzes the impact of climate change on biodiversity and ecosystem health...",
        "We present novel deep learning algorithms for image classification using convolutional neural networks...",
        "Quantum computing has the potential to revolutionize cryptography, and in this paper, we discuss...",
        "Natural language processing is a growing field in machine learning, and in this review, we explore...",
        "Climate change poses significant challenges to agriculture, and this paper investigates...",
        "Graph neural networks have gained popularity in recent years for their ability to model complex...",
        "Biodiversity conservation is crucial in the face of climate change, and this study outlines...",
        "Quantum computing offers new opportunities for drug discovery, and in this paper, we analyze...",
        "Reinforcement learning is a powerful machine learning paradigm, and in this survey, we...",
        "Artificial intelligence has the potential to help combat climate change by providing new...",
    ]*10,
}


def rank_results(query, filtered_papers):
    # Generate embeddings for user query and filtered paper abstracts
    abstracts = [abstract for abstract in filtered_papers['Abstract']]
    features = tokenizer([query for _ in range(len(abstracts))], abstracts,  padding=True, truncation=True, return_tensors="pt")
    with torch.no_grad():
        scores = model(**features).logits
    
    # Rank papers based on similarity scores
    filtered_papers['Similarity Score'] = scores.numpy()
    ranked_papers = filtered_papers.sort_values(by='Similarity Score', ascending=False)
    
    return ranked_papers

# Function to generate a download link for a PDF file
def generate_pdf_link(pdf_file_path, link_text):
    with open(pdf_file_path, "rb") as f:
        pdf_data = f.read()
        
    b64_pdf_data = base64.b64encode(pdf_data).decode()
    href = f'<a href="data:application/octet-stream;base64,{b64_pdf_data}" download="{os.path.basename(pdf_file_path)}">{link_text}</a>'
    return href

# Function to filter papers based on user input
def filter_papers(papers,year_range, is_open_access, abstract_query):
    if year_range:
        papers = papers[(papers['Year'] >= year_range[0]) & (papers['Year'] <= year_range[1])]
    if is_open_access is not None:
        papers = papers[papers['Is_Open_Access'] == is_open_access]

    return papers

# Function to perform complex boolean search
def complex_boolean_search(text, query):
    query = re.sub(r'(?<=[A-Za-z0-9])\s+(?=[A-Za-z0-9])', 'AND', query)
    query = re.sub(r'\b(AND|OR)\b', r'\\\1', query)
    query = re.sub(r'(?<=\s)\bNOT\b(?=\s)', ' -', query)
    query = re.sub(r'(?<=\b)\bNOT\b(?=\s)', '-', query)
    try:
        return bool(re.search(query, text, flags=re.IGNORECASE))
    except re.error:
        return False

papers_df = pd.DataFrame(data)
if "model" not in locals():
  model = AutoModelForSequenceClassification.from_pretrained('cross-encoder/ms-marco-MiniLM-L-6-v2')
  tokenizer = AutoTokenizer.from_pretrained('cross-encoder/ms-marco-MiniLM-L-6-v2')
  model.eval()

# Streamlit interface
st.set_page_config(page_title="Scientific Article Search", layout="wide")

hide_menu_style = """
        <style>
        #MainMenu {visibility: hidden;}
        </style>
        """
st.markdown(hide_menu_style, unsafe_allow_html=True)

# Add custom CSS to scale the sidebar
scale = 0.4
custom_css = """
<style>
    .filterbar .sidebar-content {{
        transform: scale({scale});
        transform-origin: top left;
    }}
</style>"""

st.markdown(custom_css, unsafe_allow_html=True)
page=1
per_page=10
title = ""
filtered_papers = papers_df

# Sidebar for filters
with st.sidebar:
    st.header("Filters")
    search_query= st.text_input("Query")
    so = st.multiselect(
        label='Search Over', 
        options=['Abstract','Everything','Authors'],
        default=['Everything'], 
        help='Search and select multiple options from the dropdown menu')

    sites = st.multiselect(
        label='Search Over', 
        options=['OpenAlex','Google Scholar','Base Search','All Sites'], 
        default=['All Sites'], 
        help='Search and select multiple options from the dropdown menu')

    year_range = st.slider("Year Range", min_value=1900, max_value=2022, value=(1990, 2022), step=1)

    is_open_access = st.multiselect(
        label='Open Access', 
        options=["All","Yes","No"], 
        default="All", 
        help='Search and select multiple options from the dropdown menu')

    # Convert is_open_access to boolean or None
    if is_open_access == "Yes":
        is_open_access = True
    elif is_open_access == "No":
        is_open_access = False
    else:
        is_open_access = None

    # Filter button
    if st.button("Search"):
      filtered_papers = filter_papers(papers_df, year_range, is_open_access,search_query)
    else:
      filtered_papers = papers_df  # Empty dataframe

    filtered_papers = rank_results(search_query, filtered_papers)

if not filtered_papers.empty:
    # Pagination
    no_pages = math.ceil(len(filtered_papers)/per_page)

    # Generate pagination buttons
    if no_pages == 1:
        pagination_buttons = []
    elif no_pages == 2:
        pagination_buttons = [st.button('1'), st.write('2'), ]
    else:
        pagination_buttons = [st.button(str(page-1) if page > 1 else '1'),
                              st.write(str(page)),
                              st.button(str(page+1) if page < no_pages else str(no_pages))]

    # Display results with a more advanced look
    col1, col2 = st.columns([3, 1])
    title, authors, year, journal = st.columns([5, 5, 2, 3])
    with title:
        st.subheader("Title")
    with year:
        st.subheader("Year")
    with journal:
        st.subheader("Journal")

    # Display paginated results
    start_idx = (page - 1) * per_page
    end_idx = start_idx + per_page
    paginated_papers = filtered_papers.iloc[start_idx:end_idx]

    for idx, paper in paginated_papers.iterrows():
        st.write("---")
        title, authors, year, journal = st.columns([5, 5, 2, 3])

        with col1:
            with title:
                st.write(f"{paper['Title']}")
            with authors:
                st.write(f"{paper['Authors']}")
            with year:
                st.write(f"{paper['Year']}")
            with journal:
                st.write(f"{paper['Journal']}")
        abstract = st.expander("Abstract")
        abstract.write(f"{paper['Abstract']}")

        with col2:
            pdf_file_path = "/content/ADVS-6-1801195.pdf"  # Replace with the actual path to the PDF file associated with the paper
            # st.markdown(generate_pdf_link(pdf_file_path, "Show PDF"), unsafe_allow_html=True)

    st.write("---")

    # Display pagination buttons
    per_page = st.selectbox("Results per page", [10, 20, 30], index=0)
    pagination_bar = st.columns(3)
    if no_pages > 1:
        with pagination_bar[1]:
            for button in pagination_buttons:
                button
else:
    st.header("No papers found.")