import plotly.express as px
import pandas as pd

df = pd.read_csv('data/results.csv', sep="\t")
custom_order = ["MICCAI", "MIDL", "Nature", "arXiv"]

# Calculate total number of URLs per year and venue
total_urls_per_year_venue = df.groupby(['year', 'venue']).size().reset_index(name='total_urls')

# Calculate the number of URLs with errors per year and venue
errors_per_year_venue = df[df["pred_valid"] != False].groupby(['year', 'venue']).size().reset_index(name='errors')

# Merge the DataFrames to calculate the error rate
error_rate_df = pd.merge(total_urls_per_year_venue, errors_per_year_venue, on=['year', 'venue'], how='left')
error_rate_df['errors'] = error_rate_df['errors'].fillna(0)  # Replace NaN with 0 for venues with no errors
error_rate_df['error_rate'] = error_rate_df['errors'] / error_rate_df['total_urls']

# Plot the error rates using Plotly, with year on x-axis and color by venue
fig = px.bar(
    error_rate_df,
    x='year',
    y='error_rate',
    color='venue',
    barmode='group',
    title=f'Success Rate per Venue and Year for "valid_url"',
    labels={'error_rate': 'Success Rate', 'year': 'Year'},
    category_orders={'venue': custom_order}
)

fig.update_yaxes(range=[0, 1])
fig.update_xaxes(range=[2017.5, 2024.5])
fig.show()


for topic in ["pred_live", "pred_dependencies", "pred_training", "pred_evaluation", "pred_weights", "pred_readme", "pred_license"]:
    # Calculate total number of URLs per year and venue
    total_valid_urls_per_year_venue = df[df["pred_valid"] == True].groupby(['year', 'venue']).size().reset_index(name='total_urls')

    # Calculate the number of URLs with errors per year and venue
    passes_per_year_venue = df[df[topic] != "No"].groupby(['year', 'venue']).size().reset_index(name='successes')

    # Merge the DataFrames to calculate the error rate
    success_rate_df = pd.merge(total_urls_per_year_venue, passes_per_year_venue, on=['year', 'venue'], how='left')
    success_rate_df['successes'] = success_rate_df['successes'].fillna(0)  # Replace NaN with 0 for venues with no errors
    success_rate_df['success_rate'] = success_rate_df['successes'] / success_rate_df['total_urls']

    # Plot the error rates using Plotly, with year on x-axis and color by venue
    fig = px.bar(
        success_rate_df,
        x='year',
        y='success_rate',
        color='venue',
        barmode='group',
        title=f'Success Rate per Venue and Year for "{topic}"',
        labels={'error_rate': 'Success Rate', 'year': 'Year'},
        category_orders={'venue': custom_order}
    )

    fig.update_yaxes(range=[0, 1])
    fig.update_xaxes(range=[2017.5, 2024.5])
    fig.show()


# List of columns to check for "No"
columns_to_check = ["pred_dependencies", "pred_training", "pred_evaluation", "pred_weights", "pred_readme", "pred_license"]

# Step 1: Calculate the number of "No" answers per row for the specified columns
df['no_count'] = df[columns_to_check].apply(lambda row: (row != 'No').sum(), axis=1)

# Step 2: Create scatter plot with pred_stars on x-axis and no_count on y-axis, color-coded by venue
fig = px.scatter(
    df,
    x='pred_citations',
    y='no_count',
    color='venue',
    title='Number of "No" Answers vs Predicted Stars, Color Coded by Venue',
    labels={'pred_stars': 'Predicted Stars', 'no_count': 'Automated Reproducibility score (0-6)'},
    category_orders={'venue': custom_order},  # Ensure custom order for venue if necessary
    log_x=True
)

# Step 3: Display the scatter plot
fig.show()

# List of columns to check for "No"
columns_to_check = ["pred_dependencies", "pred_training", "pred_evaluation", "pred_weights", "pred_readme", "pred_license"]

# Step 1: Calculate the number of "No" answers per row for the specified columns
df['no_count'] = df[columns_to_check].apply(lambda row: (row != 'No').sum(), axis=1)

# Step 2: Create a strip plot (scatter-like) with jitter to show individual "No" counts
fig = px.strip(
    df,
    x='venue',
    y='no_count',
    color='venue',
    title='Individual "No" Scores with Jitter per Venue',
    labels={'no_count': 'Automated Reproducibility Score (0-6)', 'venue': 'Venue'},
    category_orders={'venue': custom_order},  # Ensure custom order for venues
    stripmode='overlay'  # Allows all individual points to overlay each other
)

# Step 3: Add some jitter to the x-axis so points don't overlap
fig.update_traces(jitter=0.3, marker={'size': 8}, selector=dict(mode='markers'))

# Step 4: Optionally overlay a bar plot or box plot to show mean/median and spread
fig.add_trace(px.box(
    df,
    x='venue',
    y='no_count',
    category_orders={'venue': custom_order}
).data[0])  # We add the first trace of the box plot to overlay

# Step 5: Show the plot
fig.show()

for topic in ["pred_live", "pred_dependencies", "pred_training", "pred_evaluation", "pred_weights", "pred_readme", "pred_license"]:
    # Calculate total number of URLs per venue
    total_urls_per_venue = df.groupby('venue').size().reset_index(name='total_urls')

    # Calculate the number of URLs with errors per venue
    errors_per_venue = df[df[topic] != "No"].groupby('venue').size().reset_index(name='errors')

    # Merge the DataFrames to calculate the error rate
    error_rate_df = pd.merge(total_urls_per_venue, errors_per_venue, on='venue', how='left')
    error_rate_df['errors'] = error_rate_df['errors'].fillna(0)  # Replace NaN with 0 for venues with no errors
    error_rate_df['error_rate'] = error_rate_df['errors'] / error_rate_df['total_urls']

    # Plot the error rates using Plotly, with venue on x-axis
    fig = px.bar(
        error_rate_df,
        x='venue',
        y='error_rate',
        color='venue',
        title=f'Success Rate per Venue for "{topic}"',
        labels={'error_rate': 'Success Rate', 'venue': 'Venue'},
        category_orders={'venue': custom_order}
    )

    fig.update_yaxes(range=[0, 1])
    fig.show()