using datasets

app.py

@@ -7,7 +7,7 @@ from datetime import datetime
 from pathlib import Path
 import pandas as pd
 import pytz
-import requests
+from datasets import load_dataset
 import streamlit as st
 from huggingface_hub import CommitScheduler, HfApi
 
@@ -20,7 +20,7 @@ PLAYERS_JSON = 'players.json'
 image_path = 'ipl_image.png'
 
 
-PREDICTIONS_FOLDER = Path("ipl_predictions")
+PREDICTIONS_FOLDER = Path("predictions")
 PREDICTIONS_FOLDER.mkdir(parents=True, exist_ok=True)
 
 users_file = Path("leaders") / f"users.json"
@@ -32,7 +32,7 @@ scheduler = CommitScheduler(
     repo_id="DIS_IPL_Dataset",
     repo_type="dataset",
     folder_path=PREDICTIONS_FOLDER,  # Local folder where predictions are saved temporarily
-    path_in_repo="ipl_predictions",  # Path in dataset repo where predictions will be saved
+    path_in_repo="predictions",  # Path in dataset repo where predictions will be saved
     every=5,  # Push every 240 minutes (4 hours)
 )
 
@@ -347,25 +347,13 @@ with st.expander("Leaderboard 🏆"):
 ############################# Admin Panel ##################################
 ADMIN_PASSPHRASE = "admin123"
 
-from huggingface_hub import Repository
-
-# Define the local path to clone the repository
-local_repo_path = "DIS_IPL_Dataset"
-# Define your dataset repository name on Hugging Face Hub
-repo_id = "datasets/Jay-Rajput/DIS_IPL_Dataset"
-# Clone the repository (this only needs to be done once)
-repo = Repository(local_dir=local_repo_path, clone_from=repo_id)
-
+def load_dataset_repo():
+    return load_dataset("Jay-Rajput/DIS_IPL_Dataset")
 
 def fetch_latest_predictions(match_id):
-    # Assuming predictions are stored in the "ipl_predictions" directory in your dataset repo
-    predictions_path = Path("ipl_predictions")
-    latest_predictions = []
-    for prediction_file in predictions_path.glob(f"prediction_{match_id}_*.json"):
-        with open(prediction_file, 'r') as file:
-            prediction_data = json.load(file)
-            latest_predictions.append(prediction_data)
-    return latest_predictions
+    dataset = load_dataset_repo()
+    predictions = dataset['train'].filter(lambda example: example['match_id'] == match_id)
+    return predictions
 
 
 def save_match_outcomes(outcomes):
@@ -378,9 +366,9 @@ def update_leaderboard_and_outcomes(match_id, winning_team, man_of_the_match):
     predictions = fetch_latest_predictions(match_id)
     
     outcomes = load_data(OUTCOMES_JSON)  # Load existing match outcomes
-    users_file_path = repo.git_pull("leaders/users.json")
-    with open(users_file_path, 'r') as file:
-        users = json.load(file)
+    # Load existing match outcomes and user data from the test split
+    dataset = load_dataset_repo()
+    users = {item['user_name']: item for item in dataset['test']}
 
     # Directly update or add the match outcome
     outcome_exists = False
@@ -395,7 +383,9 @@ def update_leaderboard_and_outcomes(match_id, winning_team, man_of_the_match):
     # Update user points based on prediction accuracy
     for prediction in predictions:
         user_name = prediction['user_name']
-        users[user_name] = users.get(user_name, 0)  # Initialize user points if not present
+        # Initialize user points if not present
+        if user_name not in users:
+            users[user_name] = {'user_name': user_name, 'points': 0}
 
         # Update points based on prediction accuracy
         if prediction['predicted_winner'] == winning_team:
@@ -407,14 +397,7 @@ def update_leaderboard_and_outcomes(match_id, winning_team, man_of_the_match):
             users[user_name] -= 200 + prediction['bid_points']  # Penalty for wrong team prediction
 
     save_match_outcomes(outcomes)
-    # Save updated users data back to the dataset repo
-    with open(users_file_path, 'w') as file:
-        json.dump(users, file)
-        
-    # Commit changes to the dataset repo
-    repo.git_add(users_file_path)
-    repo.git_commit("Update leaderboard")
-    repo.git_push()
+    users.save_to_disk(USERS_JSON)
 
 
 with st.sidebar:

requirements.txt

@@ -1,3 +1,4 @@
+datasets
 huggingface_hub
 pandas
 pytz