app.py

import streamlit as st
import os
import pandas as pd
import json
from html import escape
import difflib

def generate_diff_html_word_level(text1, text2):
    """
    Generates word-level difference between text1 and text2 as HTML, correctly handling spaces.
    """
    words1 = text1.split()
    words2 = text2.split()
    
    diff = []
    matcher = difflib.SequenceMatcher(None, words1, words2)
    
    for opcode in matcher.get_opcodes():
        tag, i1, i2, j1, j2 = opcode
        if tag == 'replace':
            diff.append('<del style="background-color: #fbb6ce;">' + escape(' '.join(words1[i1:i2])) + '</del>')
            diff.append('<ins style="background-color: #b7e4c7;">' + escape(' '.join(words2[j1:j2])) + '</ins>')
        elif tag == 'delete':
            diff.append('<del style="background-color: #fbb6ce;">' + escape(' '.join(words1[i1:i2])) + '</del>')
        elif tag == 'insert':
            diff.append('<ins style="background-color: #b7e4c7;">' + escape(' '.join(words2[j1:j2])) + '</ins>')
        elif tag == 'equal':
            diff.append(escape(' '.join(words1[i1:i2])))

    final_html = ' '.join(diff).replace('</del> <ins', '</del>&nbsp;<ins')
    return f'<pre style="white-space: pre-wrap;">{final_html}</pre>'

os.environ["PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION"] = "python"
st.set_page_config(layout="wide")

@st.cache_data
def convert_df(df):
    return df.to_csv(index=False, quotechar='"').encode('utf-8')

@st.cache_data
def load_narratives_data():
    data = []
    with open("narratives.jsonl", "r") as f:
        for line in f:
            data.append(json.loads(line))
    return pd.DataFrame(data)

narratives_df = load_narratives_data()

# Extract language from id
narratives_df['language'] = narratives_df['id'].str.extract('-(rus|zho|fas)-')

col1, col2 = st.columns([1, 3], gap="large")

with st.sidebar:
    st.title("Options")

with col1:
    st.title("Narratives")
    
    # Add language filter
    selected_language = st.selectbox(
        "Select language",
        ["All", "rus", "zho", "fas"]
    )
    
    if selected_language != "All":
        filtered_df = narratives_df[narratives_df['language'] == selected_language]
    else:
        filtered_df = narratives_df
    
    narrative_ids = filtered_df["id"].tolist()
    container_for_nav = st.container()
    
    def sync_from_drop():
        if st.session_state.selectbox_narrative == "Overview":
            st.session_state.narrative_index = -1
        else:
            st.session_state.narrative_index = narrative_ids.index(st.session_state.selectbox_narrative)

    def sync_from_number():
        st.session_state.narrative_index = st.session_state.narrative_number
        if st.session_state.narrative_number == -1:
            st.session_state.selectbox_narrative = "Overview"
        else:
            st.session_state.selectbox_narrative = narrative_ids[st.session_state.narrative_number]

    narrative_number = container_for_nav.number_input(
        min_value=-1, step=1, max_value=len(narrative_ids) - 1,
        on_change=sync_from_number,
        label=f"Select narrative by index (up to **{len(narrative_ids) - 1}**)",
        key="narrative_number"
    )
    selectbox_narrative = container_for_nav.selectbox(
        "Select narrative by ID",
        ["Overview"] + narrative_ids,
        on_change=sync_from_drop,
        key="selectbox_narrative"
    )
    st.divider()

with col2:
    narrative_index = narrative_number

    if narrative_index >= 0:
        narrative = filtered_df.iloc[narrative_index]
        
        st.markdown("<h1 style='text-align: center; color: black;text-decoration: underline;'>Editor</h1>", unsafe_allow_html=True)

        container = st.container()

        container.subheader(f"Narrative ID: {narrative['id']}")
        container.divider()

        container.subheader("Diff: Original English vs Altered English")
        processed_diff = generate_diff_html_word_level(narrative['original_english'].strip(), narrative['altered_english'].strip())
        with container.container(border=True):
            st.markdown(processed_diff, unsafe_allow_html=True)
        container.divider()

        container.subheader("Original Text")
        original_input = container.text_area("Edit the original text", value=narrative['original'].strip(), height=300)

    elif narrative_index < 0:
        st.title("Overview")
        st.write(f"Total number of narratives: {len(filtered_df)}")
        if selected_language != "All":
            st.write(f"Selected language: {selected_language}")
        st.write("Select a narrative from the sidebar to view and edit its details.")