data_processor.py

import re
import pandas as pd
import os
# from huggingface_hub import InferenceClient
from mistralai import Mistral     

class DataProcessor:
    INTERVENTION_COLUMN_OPTIONS = [
        'Did the intervention happen today?',
        'Did the intervention take place today?'
    ]
    YES_RESPONSES = ['yes', 'assessment day']  # Added this line
    ENGAGED_STR = 'Engaged'
    PARTIALLY_ENGAGED_STR = 'Partially Engaged'
    NOT_ENGAGED_STR = 'Not Engaged'

    # HF INFERENCE
    # def __init__(self, student_metrics_df=None):
    #     self.hf_api_key = os.getenv('HF_API_KEY')
    #     if not self.hf_api_key:
    #         raise ValueError("HF_API_KEY not set in environment variables")
    #     self.client = InferenceClient(api_key=self.hf_api_key)
    #     self.student_metrics_df = student_metrics_df
    #     self.intervention_column = None  # Will be set when processing data

    # MISTRAL INFERENCE
    def __init__(self, student_metrics_df=None):
        self.mistral_api_key = os.getenv('MISTRAL_API_KEY')
        if not self.mistral_api_key:
            raise ValueError("MISTRAL_API_KEY not set in environment variables")
        self.client = Mistral(api_key=self.mistral_api_key)
        self.student_metrics_df = student_metrics_df
        self.intervention_column = None  # Will be set when processing data

    def read_excel(self, uploaded_file):
        return pd.read_excel(uploaded_file)

    def format_session_data(self, df):
        date_column = next((col for col in df.columns if col in ["Date of Session", "Date"]), None)
        if date_column:
            df[date_column] = pd.to_datetime(df[date_column], errors='coerce').dt.date
        else:
            print("Warning: Neither 'Date of Session' nor 'Date' column found in the dataframe.")
        
        df['Timestamp'] = self.safe_convert_to_datetime(df['Timestamp'], '%I:%M %p')
        df['Session Start Time'] = self.safe_convert_to_time(df['Session Start Time'], '%I:%M %p')
        df['Session End Time'] = self.safe_convert_to_time(df['Session End Time'], '%I:%M %p')
        return df

    def safe_convert_to_time(self, series, format_str='%I:%M %p'):
        try:
            converted = pd.to_datetime(series, format='%H:%M:%S', errors='coerce')
            if format_str:
                return converted.dt.strftime(format_str)
            return converted
        except Exception as e:
            print(f"Error converting series to time: {e}")
            return series

    def safe_convert_to_datetime(self, series, format_str=None):
        try:
            converted = pd.to_datetime(series, errors='coerce')
            if format_str:
                return converted.dt.strftime(format_str)
            return converted
        except Exception as e:
            print(f"Error converting series to datetime: {e}")
            return series

    def replace_student_names_with_initials(self, df):
        updated_columns = []
        for col in df.columns:
            if 'Student Attendance' in col:
                # Search for the last occurrence of text within square brackets at the end of the string
                match = re.search(r'\[(.+?)\]$', col)
                if not match:
                    # Handle cases where the closing bracket might be missing
                    match = re.search(r'\[(.+)$', col)
                if match:
                    name = match.group(1).strip()
                    # Remove any trailing closing bracket if it wasn't matched earlier
                    name = name.rstrip(']')
                    # Get initials
                    initials = ''.join([part[0] for part in name.strip().split()])
                    updated_col = f'Student Attendance [{initials}]'
                    updated_columns.append(updated_col)
                else:
                    # If no match is found, keep the column name as is
                    updated_columns.append(col)
            else:
                updated_columns.append(col)
        df.columns = updated_columns
        return df

    def find_intervention_column(self, df):
        for column in self.INTERVENTION_COLUMN_OPTIONS:
            if column in df.columns:
                self.intervention_column = column
                return column
        raise ValueError("No intervention column found in the dataframe.")

    def get_intervention_column(self, df):
        if self.intervention_column is None:
            self.intervention_column = self.find_intervention_column(df)
        return self.intervention_column

    def compute_intervention_statistics(self, df):
        intervention_column = self.get_intervention_column(df)
        total_days = len(df)
        sessions_held = df[intervention_column].str.strip().str.lower().isin(self.YES_RESPONSES).sum()  # Modified line
        intervention_frequency = (sessions_held / total_days) * 100 if total_days > 0 else 0
        return pd.DataFrame({
            'Intervention Dosage (%)': [round(intervention_frequency, 0)],
            'Intervention Sessions Held': [sessions_held],
            'Intervention Sessions Not Held': [total_days - sessions_held],
            'Total Number of Days Available': [total_days]
        })

    def classify_engagement(self, engagement_str):
        engagement_str = str(engagement_str).lower()
        if engagement_str.startswith(self.ENGAGED_STR.lower()):
            return self.ENGAGED_STR
        elif engagement_str.startswith(self.PARTIALLY_ENGAGED_STR.lower()):
            return self.PARTIALLY_ENGAGED_STR
        elif engagement_str.startswith(self.NOT_ENGAGED_STR.lower()):
            return self.NOT_ENGAGED_STR
        else:
            return 'Unknown'
    
    def compute_student_metrics(self, df):
        intervention_column = self.get_intervention_column(df)
        intervention_df = df[df[intervention_column].str.strip().str.lower().isin(self.YES_RESPONSES)]
        intervention_sessions_held = len(intervention_df)
        student_columns = [col for col in df.columns if col.startswith('Student Attendance')]
    
        student_metrics = {}
        for col in student_columns:
            student_name = col.replace('Student Attendance [', '').replace(']', '').strip()
            student_data = intervention_df[[col]].copy()

            # Check if the student is still active (has non-empty cells in recent sessions)
            is_active = student_data[col].notna().any()
            if not is_active:
                continue  # Skip inactive students
            
            student_data[col] = student_data[col].fillna('Absent')
    
            # Classify each entry
            student_data['Engagement'] = student_data[col].apply(self.classify_engagement)
    
            # Calculate attendance
            attendance_values = student_data['Engagement'].apply(
                lambda x: 1 if x in [self.ENGAGED_STR, self.PARTIALLY_ENGAGED_STR, self.NOT_ENGAGED_STR] else 0
            )
    
            sessions_attended = attendance_values.sum()
            attendance_pct = (sessions_attended / intervention_sessions_held * 100) if intervention_sessions_held > 0 else 0
            attendance_pct = round(attendance_pct)
    
            # Engagement counts (excluding 'Absent')
            engagement_counts = {
                self.ENGAGED_STR: 0,
                self.PARTIALLY_ENGAGED_STR: 0,
                self.NOT_ENGAGED_STR: 0
            }
    
            # Count the engagement types, excluding 'Absent'
            for x in student_data['Engagement']:
                if x in engagement_counts:
                    engagement_counts[x] += 1
                # 'Absent' is not counted in engagement_counts
    
            total_present_sessions = sum(engagement_counts.values())
    
            engaged_pct = (
                (engagement_counts[self.ENGAGED_STR] / total_present_sessions * 100)
                if total_present_sessions > 0 else 0
            )
            engaged_pct = round(engaged_pct)
    
            partially_engaged_pct = (
                (engagement_counts[self.PARTIALLY_ENGAGED_STR] / total_present_sessions * 100)
                if total_present_sessions > 0 else 0
            )
            partially_engaged_pct = round(partially_engaged_pct)
    
            not_engaged_pct = (
                (engagement_counts[self.NOT_ENGAGED_STR] / total_present_sessions * 100)
                if total_present_sessions > 0 else 0
            )
            not_engaged_pct = round(not_engaged_pct)
    
            # Engagement percentage is based on Engaged and Partially Engaged sessions
            engagement_pct = (
                ((engagement_counts[self.ENGAGED_STR] + engagement_counts[self.PARTIALLY_ENGAGED_STR]) / total_present_sessions * 100)
                if total_present_sessions > 0 else 0
            )
            engagement_pct = round(engagement_pct)
    
            # Absent percentage (for reference, not used in engagement calculation)
            absent_sessions = student_data['Engagement'].value_counts().get('Absent', 0)
            absent_pct = (absent_sessions / intervention_sessions_held * 100) if intervention_sessions_held > 0 else 0
            absent_pct = round(absent_pct)
    
            # Determine if the student attended ≥ 90% of sessions
            attended_90 = "Yes" if attendance_pct >= 90 else "No"
    
            # Determine if the student was engaged ≥ 80% of the time
            engaged_80 = "Yes" if engagement_pct >= 80 else "No"
    
            # Store metrics
            student_metrics[student_name] = {
                'Attended ≥ 90%': attended_90,
                'Engagement ≥ 80%': engaged_80,
                'Attendance (%)': attendance_pct,
                # 'Engagement (%)': engagement_pct, REMOVED REMOVED
                f'{self.ENGAGED_STR} (%)': engaged_pct,
                f'{self.PARTIALLY_ENGAGED_STR} (%)': partially_engaged_pct,
                f'{self.NOT_ENGAGED_STR} (%)': not_engaged_pct,
                'Absent (%)': absent_pct
            }
    
        # Create a DataFrame from student_metrics
        student_metrics_df = pd.DataFrame.from_dict(student_metrics, orient='index').reset_index()
        student_metrics_df.rename(columns={'index': 'Student'}, inplace=True)
        return student_metrics_df
    
    # def compute_average_metrics(self, student_metrics_df):
    #     # Filter out rows with NaN values (inactive students)
    #     active_students_df = student_metrics_df.dropna()
        
    #     # Calculate the attendance and engagement average percentages across active students
    #     attendance_avg_stats = active_students_df['Attendance (%)'].mean()
    #     engagement_avg_stats = active_students_df[f'{self.ENGAGED_STR} (%)'].mean()
        
    #     # Round the averages to whole numbers
    #     attendance_avg_stats = round(attendance_avg_stats)
    #     engagement_avg_stats = round(engagement_avg_stats)
        
    #     return attendance_avg_stats, engagement_avg_stats

    def compute_average_metrics(self, student_metrics_df):
        # Filter out rows with NaN values (inactive students)
        active_students_df = student_metrics_df.dropna()
        
        # Calculate the attendance average percentage across active students
        attendance_avg_stats = active_students_df['Attendance (%)'].mean()
        
        # Calculate the engagement average percentage across active students
        # Only consider 'Engaged' and 'Partially Engaged' percentages, exclude 'Absent'
        total_engagement = active_students_df[f'{self.ENGAGED_STR} (%)'] + active_students_df[f'{self.PARTIALLY_ENGAGED_STR} (%)']
        engagement_avg_stats = total_engagement.mean()
    
        # Round the averages to whole numbers
        attendance_avg_stats = round(attendance_avg_stats)
        engagement_avg_stats = round(engagement_avg_stats)
        
        return attendance_avg_stats, engagement_avg_stats

    def evaluate_student(self, row, attendance_threshold=90, engagement_threshold=80):
        if row["Attended ≥ 90%"] == "No":
            return "Address Attendance"
        elif row["Engagement ≥ 80%"] == "No":
            return "Address Engagement"
        else:
            return "Consider barriers, fidelity, and progress monitoring"