demo.py

import gradio as gr
# import pandas as pd
import polars as pl

from math import ceil
import os

from data import df, pitch_stats, league_pitch_stats, player_df
from gradio_function import *
from translate import jp_pitch_to_en_pitch, max_pitch_types

os.makedirs('files', exist_ok=True)

css = '''
.pitch-usage {height: 256px}
.pitch-usage .js-plotly-plot {height: 100%}

.pitch-velo {height: 100px}
.pitch-velo .js-plotly-plot {height: 100%}

.pitch-loc {height: 320px}
.pitch-loc .js-plotly-plot {height: 100%}

.pitch-velo-summary div.plotly-notifier {visibility: hidden}
'''

with gr.Blocks(
    css=css
) as demo:
  gr.Markdown('''
  # NPB data visualization demo
  [Data from SportsNavi](https://sports.yahoo.co.jp/)
  ''')

  source_df = gr.State(df)
  app_df = gr.State(df)
  app_league_df = gr.State(df)
  app_pitch_stats = gr.State(pitch_stats)
  app_league_pitch_stats = gr.State(league_pitch_stats)

  with gr.Tab('Pitcher Dashboard'):
    with gr.Row():
      player = gr.Dropdown(value=None, choices=sorted(player_df.filter(pl.col('name').is_not_null())['name'].to_list()), label='Player')
      handedness = gr.Radio(value='Both', choices=['Both', 'Left', 'Right'], type='value', interactive=False, label='Batter Handedness')

    # preview = gr.DataFrame()
    download_file = gr.DownloadButton(label='Download player data')

    with gr.Group():
      with gr.Row():
        usage = gr.Plot(label='Pitch usage')
        velo_summary = gr.Plot(label='Velocity summary', elem_classes='pitch-velo-summary')
        loc_summary = gr.Plot(label='Overall location')

    max_locs = len(jp_pitch_to_en_pitch)
    locs_per_row = 4
    max_rows = ceil(max_locs/locs_per_row)

    gr.Markdown('''
    ## Pitch Locations
    Pitcher's persective
    <br>
    `NPB` refers to the top 10% of pitches thrown across the league with the current search constraints e.g. handedness
    <br>
    Note: To speed up the KDE, we restrict the league-wide pitches to 5,000 pitches
    ''')
    pitch_rows = []
    pitch_groups = []
    pitch_names = []
    pitch_infos = []
    pitch_velos = []
    pitch_locs = []
    for row in range(max_rows):
      visible = row==0
      pitch_row = gr.Row(visible=visible)
      pitch_rows.append(pitch_row)
      with pitch_row:
        _locs_per_row = locs_per_row if row < max_rows-1 else max_locs - locs_per_row * (max_rows - 1)
        for col in range(_locs_per_row):
          with gr.Column(min_width=256):
            pitch_group = gr.Group(visible=visible)
            pitch_groups.append(pitch_group)
            with pitch_group:
              pitch_names.append(gr.Markdown(f'### Pitch {col+1}', visible=visible))
              pitch_infos.append(gr.DataFrame(pl.DataFrame([{'Whiff%': None, 'CSW%': None}]), interactive=False, visible=visible))
              pitch_velos.append(gr.Plot(show_label=False, elem_classes='pitch-velo', visible=visible))
              pitch_locs.append(gr.Plot(label='Pitch Location', elem_classes='pitch-loc', visible=visible))

    gr.Markdown('## Pitch Velocity')
    velo_stats = gr.DataFrame(pl.DataFrame([{'Avg. Velo': None, 'League Avg. Velo': None}]), interactive=False, label='Pitch Velocity')

    (
        player
        .input(update_dfs, inputs=[player, handedness, source_df], outputs=[app_df, app_league_df, app_pitch_stats, app_league_pitch_stats])
        .then(lambda : gr.update(value='Both', interactive=True), outputs=handedness)
    )
    handedness.input(update_dfs, inputs=[player, handedness, source_df], outputs=[app_df, app_league_df, app_pitch_stats, app_league_pitch_stats])

    # app_df.change(preview_df, inputs=app_df, outputs=preview)
    # app_df.change(set_download_file, inputs=app_df, outputs=download_file)
    # app_df.change(plot_usage, inputs=[app_df, player], outputs=usage)
    # app_df.change(plot_velo_summary, inputs=[app_df, app_league_df, player], outputs=velo_summary)
    # app_df.change(lambda df: plot_loc(df), inputs=app_df, outputs=loc_summary)
    # app_df.change(plot_pitch_cards, inputs=[app_df, app_pitch_stats], outputs=pitch_rows+pitch_groups+pitch_names+pitch_infos+pitch_velos+pitch_locs)
    app_pitch_stats.change(update_velo_stats, inputs=[app_pitch_stats, app_league_pitch_stats], outputs=velo_stats)

    (
        app_df
        .change(create_set_download_file_fn('files/player.csv'), inputs=app_df, outputs=download_file)
        .then(plot_usage, inputs=[app_df, player], outputs=usage)
        .then(plot_velo_summary, inputs=[app_df, app_league_df, player], outputs=velo_summary)
        .then(lambda df: plot_loc(df), inputs=app_df, outputs=loc_summary)
        .then(plot_pitch_cards, inputs=[app_df, app_league_df, app_pitch_stats], outputs=pitch_rows+pitch_groups+pitch_names+pitch_infos+pitch_velos+pitch_locs)
    )

    gr.Markdown('## Bugs and other notes')
    with gr.Accordion('Click to open', open=False):
      gr.Markdown('''
      - Y axis ticks messy when no velocity distribution is plotted
      - DataFrame precision inconsistent
      '''
      )
  
  with gr.Tab('Pitch Leaderboard'):
    def filter_pitch_leaderboard(min_pitches):
      return pitch_stats.filter(pl.col('Count') >= min_pitches).sort('CSW%', descending=True)

    init_min_pitches = 100
    pitch_stats.write_csv('pitch_leaderboard.csv')
    pitch_leaderboard_df = gr.State(filter_pitch_leaderboard(init_min_pitches))

    min_pitches = gr.Number(init_min_pitches, precision=0, label='Min. Pitches')
    pitch_leaderboard_download_file = gr.DownloadButton(value='pitch_leaderboard.csv', label='Download leaderboard')
    pitch_leaderboard = gr.Dataframe(value=pitch_leaderboard_df.value)
    
    min_pitches.change(filter_pitch_leaderboard, inputs=min_pitches, outputs=pitch_leaderboard_df)
    (
        pitch_leaderboard_df
        .change(create_set_download_file_fn('files/pitch_leaderboard.csv'), inputs=pitch_leaderboard_df, outputs=pitch_leaderboard_download_file)
        .then(lambda df: df, inputs=pitch_leaderboard_df, outputs=pitch_leaderboard)
    )

demo.launch(
    share=True,
    debug=True
)