Seu template de dados de gestão de despesas

import pandas as pd
import numpy as np

# --- Configuration ---
TRANSACTIONS_FILE = 'transactions_export.csv'
REIMBURSEMENTS_FILE = 'reimbursements_export.csv'
OUTPUT_FILE = 'ramp_event_log.csv'

# --- Load Data ---
def load_data():
    try:
        transactions_df = pd.read_csv(TRANSACTIONS_FILE)
        reimbursements_df = pd.read_csv(REIMBURSEMENTS_FILE)
        print("Successfully loaded transactions and reimbursements files.")
        return transactions_df, reimbursements_df
    except FileNotFoundError as e:
        print(f"Error: {e}. Make sure '{TRANSACTIONS_FILE}' and '{REIMBURSEMENTS_FILE}' are in the same directory.")
        return None, None

# --- Main Transformation Logic ---
def transform_to_event_log(transactions_df, reimbursements_df):
    events = []

    # Standardize column names for easier processing
    # These names are based on typical Ramp CSV exports, adjust if your export has different headers
    transactions_df.rename(columns={
        'Transaction ID': 'ExpenseReportId',
        'User Name': 'UserName',
        'Department': 'EmployeeDepartment',
        'Amount': 'ReportTotalAmount',
        'Category Name': 'ExpenseCategory',
        'Transaction Date': 'ExpenseIncurredTimestamp',
        'Receipts Attached At': 'ReceiptAttachedTimestamp',
        'Submitted At': 'ExpenseSubmittedTimestamp',
        'Manager Approved At': 'ManagerApprovedTimestamp',
        'Needs Review At': 'FinanceReviewPendingTimestamp',
        'Admin Approved At': 'FinanceApprovedTimestamp',
        'Accounting Sync At': 'SyncedTimestamp',
        'Memo': 'RevisionReason',
        'State': 'Status'
    }, inplace=True)

    reimbursements_df.rename(columns={
        'Reimbursement ID': 'ExpenseReportId',
        'User Name': 'UserName',
        'Department': 'EmployeeDepartment',
        'Amount': 'ReportTotalAmount',
        'Category Name': 'ExpenseCategory',
        'Created At': 'ReimbursementScheduledTimestamp',
        'Paid Out At': 'ReimbursementExecutedTimestamp',
        'Memo': 'RevisionReason',
        'State': 'Status'
    }, inplace=True)

    # Combine card transactions and reimbursements into a single dataframe for processing common events
    # Note: Reimbursements will be processed separately for reimbursement-specific events
    combined_df = transactions_df

    # Process activities from the combined dataframe
    for _, row in combined_df.iterrows():
        base_event = {
            'ExpenseReportId': row.get('ExpenseReportId'),
            'UserName': row.get('UserName'),
            'EmployeeDepartment': row.get('EmployeeDepartment'),
            'ReportTotalAmount': row.get('ReportTotalAmount'),
            'ExpenseCategory': row.get('ExpenseCategory'),
            'PolicyViolationFlag': 'FLAGGED' in str(row.get('Flags')), # Infer from 'Flags' column
            'RevisionReason': row.get('RevisionReason')
        }

        # Activity: Expense Incurred
        if pd.notna(row.get('ExpenseIncurredTimestamp')):
            events.append({**base_event, 'ActivityName': 'Expense Incurred', 'EventTimestamp': row.get('ExpenseIncurredTimestamp')})

        # Activity: Receipt Attached
        if pd.notna(row.get('ReceiptAttachedTimestamp')):
            events.append({**base_event, 'ActivityName': 'Receipt Attached', 'EventTimestamp': row.get('ReceiptAttachedTimestamp')})

        # Activity: Expense Submitted
        if pd.notna(row.get('ExpenseSubmittedTimestamp')):
            events.append({**base_event, 'ActivityName': 'Expense Submitted', 'EventTimestamp': row.get('ExpenseSubmittedTimestamp')})
            # Inferred Activity: Policy Check Performed (assumed to happen right after submission)
            policy_check_ts = pd.to_datetime(row.get('ExpenseSubmittedTimestamp')) + pd.Timedelta(seconds=1)
            events.append({**base_event, 'ActivityName': 'Policy Check Performed', 'EventTimestamp': policy_check_ts.isoformat()})
            # Inferred Activity: Manager Review Pending (starts when submitted)
            events.append({**base_event, 'ActivityName': 'Manager Review Pending', 'EventTimestamp': row.get('ExpenseSubmittedTimestamp')})

        # Activity: Manager Approved
        if pd.notna(row.get('ManagerApprovedTimestamp')):
            events.append({**base_event, 'ActivityName': 'Manager Approved', 'EventTimestamp': row.get('ManagerApprovedTimestamp')})

        # Activity: Finance Review Pending
        if pd.notna(row.get('FinanceReviewPendingTimestamp')):
            events.append({**base_event, 'ActivityName': 'Finance Review Pending', 'EventTimestamp': row.get('FinanceReviewPendingTimestamp')})

        # Activity: Finance Approved
        if pd.notna(row.get('FinanceApprovedTimestamp')):
            events.append({**base_event, 'ActivityName': 'Finance Approved', 'EventTimestamp': row.get('FinanceApprovedTimestamp')})

        # Activity: Synced with Accounting System
        if pd.notna(row.get('SyncedTimestamp')):
            events.append({**base_event, 'ActivityName': 'Synced with Accounting System', 'EventTimestamp': row.get('SyncedTimestamp')})

        # Activity: Expense Sent Back for Revision
        if row.get('Status') == 'REJECTED' and pd.notna(row.get('Updated At')):
             events.append({**base_event, 'ActivityName': 'Expense Sent Back for Revision', 'EventTimestamp': row.get('Updated At')}) # Using last updated time for rejection

    # Process reimbursement-specific activities
    for _, row in reimbursements_df.iterrows():
        base_event = {
            'ExpenseReportId': row.get('ExpenseReportId'),
            'UserName': row.get('UserName'),
            'EmployeeDepartment': row.get('EmployeeDepartment'),
            'ReportTotalAmount': row.get('ReportTotalAmount'),
            'ExpenseCategory': row.get('ExpenseCategory'),
            'PolicyViolationFlag': False, # Typically less flagging on reimbursements
            'RevisionReason': row.get('RevisionReason')
        }

        # Activity: Reimbursement Scheduled
        if row.get('Status') == 'APPROVED' and pd.notna(row.get('ReimbursementScheduledTimestamp')):
            events.append({**base_event, 'ActivityName': 'Reimbursement Scheduled', 'EventTimestamp': row.get('ReimbursementScheduledTimestamp')})

        # Activity: Reimbursement Executed
        if row.get('Status') == 'PAID' and pd.notna(row.get('ReimbursementExecutedTimestamp')):
            events.append({**base_event, 'ActivityName': 'Reimbursement Executed', 'EventTimestamp': row.get('ReimbursementExecutedTimestamp')})

    if not events:
        print("Warning: No events were generated. Check the input files and column names.")
        return None

    # Create final DataFrame
    event_log_df = pd.DataFrame(events)
    event_log_df['EventTimestamp'] = pd.to_datetime(event_log_df['EventTimestamp'])
    event_log_df.sort_values(by=['ExpenseReportId', 'EventTimestamp'], inplace=True)
    
    return event_log_df

# --- Main execution block ---
if __name__ == "__main__":
    transactions_df, reimbursements_df = load_data()
    if transactions_df is not None and reimbursements_df is not None:
        event_log = transform_to_event_log(transactions_df, reimbursements_df)
        if event_log is not None:
            event_log.to_csv(OUTPUT_FILE, index=False, date_format='%Y-%m-%dT%H:%M:%S.%fZ')
            print(f"Successfully created event log with {len(event_log)} events. Saved to '{OUTPUT_FILE}'.")

import os
import requests
import pandas as pd
from datetime import datetime, timezone

class=class="hl-string">"hl-comment"># --- Configuration ---
class=class="hl-string">"hl-comment"># It is recommended to use environment variables for credentials
CLIENT_ID = os.getenv(class="hl-string">'RAMP_CLIENT_ID', class="hl-string">'[Your_Ramp_Client_ID]')
CLIENT_SECRET = os.getenv(class="hl-string">'RAMP_CLIENT_SECRET', class="hl-string">'[Your_Ramp_Client_Secret]')

API_BASE_URL = class="hl-string">'https://api.ramp.com/v1'
TOKEN_URL = class="hl-string">'https://api.ramp.com/v1/public/customer/token'

class=class="hl-string">"hl-comment"># --- Main Functions ---

def get_access_token():
    class="hl-string">"""Authenticates with the Ramp API to get a bearer token."""
    headers = {class="hl-string">'Content-Type': class="hl-string">'application/x-www-form-urlencoded'}
    data = {
        class="hl-string">'grant_type': class="hl-string">'client_credentials',
        class="hl-string">'client_id': CLIENT_ID,
        class="hl-string">'client_secret': CLIENT_SECRET,
        class="hl-string">'scope': class="hl-string">'transactions:read reimbursements:read users:read'
    }
    try:
        response = requests.post(TOKEN_URL, headers=headers, data=data)
        response.raise_for_status()
        return response.json()[class="hl-string">'access_token']
    except requests.exceptions.RequestException as e:
        print(fclass="hl-string">"Error obtaining access token: {e}")
        if e.response is not None:
            print(fclass="hl-string">"Response body: {e.response.text}")
        return None

def fetch_paginated_data(endpoint, access_token, params=None):
    class="hl-string">"""Fetches all data from a paginated Ramp API endpoint."""
    all_data = []
    headers = {class="hl-string">'Authorization': fclass="hl-string">'Bearer {access_token}'}
    url = fclass="hl-string">"{API_BASE_URL}{endpoint}"
    
    while url:
        try:
            response = requests.get(url, headers=headers, params=params)
            response.raise_for_status()
            data = response.json()
            all_data.extend(data.get(class="hl-string">'data', []))
            url = data.get(class="hl-string">'next') class=class="hl-string">"hl-comment"># Ramp uses a full URL for the next page
            params = None class=class="hl-string">"hl-comment"># Subsequent requests use the class="hl-string">'next' URL which includes params
        except requests.exceptions.RequestException as e:
            print(fclass="hl-string">"Error fetching data from {endpoint}: {e}")
            break
            
    return all_data

def map_state_to_activity(state, reason_code):
    class="hl-string">"""Maps reimbursement state history to a user-friendly activity name."""
    if reason_code == class="hl-string">'REJECTED':
        return class="hl-string">'Expense Sent Back for Revision'
    
    state_map = {
        class="hl-string">'SUBMITTED': class="hl-string">'Expense Submitted',
        class="hl-string">'MANAGER_APPROVAL': class="hl-string">'Manager Review Pending',
        class="hl-string">'ACCOUNTING_APPROVAL': class="hl-string">'Finance Review Pending',
        class="hl-string">'MANAGER_APPROVED': class="hl-string">'Manager Approved',
        class="hl-string">'ACCOUNTING_APPROVED': class="hl-string">'Finance Approved',
        class="hl-string">'PAYMENT_SCHEDULED': class="hl-string">'Reimbursement Scheduled',
        class="hl-string">'PAYMENT_PROCESSING': class="hl-string">'Reimbursement Executed', class=class="hl-string">"hl-comment"># Approximation
        class="hl-string">'FINISHED': class="hl-string">'Reimbursement Executed',
        class="hl-string">'SYNCED': class="hl-string">'Synced with Accounting System'
    }
    return state_map.get(state)

class=class="hl-string">"hl-comment"># --- Main Execution ---
def main():
    class="hl-string">"""Main script to extract, transform, and save the event log."""
    print(class="hl-string">"Starting Ramp data extraction...")
    access_token = get_access_token()
    if not access_token:
        print(class="hl-string">"Failed to get access token. Exiting.")
        return

    class=class="hl-string">"hl-comment"># 1. Fetch Users for enrichment
    print(class="hl-string">"Fetching users...")
    users = fetch_paginated_data(class="hl-string">'/users', access_token)
    user_map = {u[class="hl-string">'id']: {class="hl-string">'name': fclass="hl-string">"{u['first_name']} {u['last_name']}", class="hl-string">'department': u.get(class="hl-string">'department')} for u in users}
    print(fclass="hl-string">"Found {len(user_map)} users.")

    class=class="hl-string">"hl-comment"># 2. Fetch Reimbursements within a date range
    START_DATE = class="hl-string">'2023-01-01T00:00:00Z'
    END_DATE = class="hl-string">'2023-12-31T23:59:59Z'
    print(fclass="hl-string">"Fetching reimbursements from {START_DATE} to {END_DATE}...")
    reimbursement_params = {class="hl-string">'created_at_gte': START_DATE, class="hl-string">'created_at_lte': END_DATE}
    reimbursements = fetch_paginated_data(class="hl-string">'/reimbursements', access_token, params=reimbursement_params)
    print(fclass="hl-string">"Found {len(reimbursements)} reimbursements.")

    event_log = []

    class=class="hl-string">"hl-comment"># 3. Process each reimbursement to create the event log
    for r in reimbursements:
        case_id = r[class="hl-string">'id']
        report_total = r[class="hl-string">'amount']
        submitter_id = r[class="hl-string">'user_id']
        submitter_details = user_map.get(submitter_id, {class="hl-string">'name': class="hl-string">'Unknown User', class="hl-string">'department': None})
        expense_category = r.get(class="hl-string">'merchant_name') class=class="hl-string">"hl-comment"># Best available proxy for category
        
        class=class="hl-string">"hl-comment"># Process reimbursement state history
        for history in r.get(class="hl-string">'state_history', []):
            activity_name = map_state_to_activity(history[class="hl-string">'state'], history.get(class="hl-string">'reason_code'))
            if activity_name:
                event_log.append({
                    class="hl-string">'ExpenseReportId': case_id,
                    class="hl-string">'ActivityName': activity_name,
                    class="hl-string">'EventTimestamp': history[class="hl-string">'created_at'],
                    class="hl-string">'UserName': user_map.get(history.get(class="hl-string">'user_id'), submitter_details)[class="hl-string">'name'],
                    class="hl-string">'EmployeeDepartment': submitter_details[class="hl-string">'department'],
                    class="hl-string">'ReportTotalAmount': report_total,
                    class="hl-string">'ExpenseCategory': expense_category,
                    class="hl-string">'PolicyViolationFlag': None, class=class="hl-string">"hl-comment"># Not directly available in reimbursement history
                    class="hl-string">'RevisionReason': history.get(class="hl-string">'reason_notes')
                })

        class=class="hl-string">"hl-comment"># Fetch and process associated transactions
        for tx_id in r.get(class="hl-string">'transaction_ids', []):
            try:
                tx_response = requests.get(fclass="hl-string">"{API_BASE_URL}/transactions/{tx_id}", headers={class="hl-string">'Authorization': fclass="hl-string">'Bearer {access_token}'})
                tx_response.raise_for_status()
                tx = tx_response.json()

                class=class="hl-string">"hl-comment"># Activity: Expense Incurred
                event_log.append({
                    class="hl-string">'ExpenseReportId': case_id,
                    class="hl-string">'ActivityName': class="hl-string">'Expense Incurred',
                    class="hl-string">'EventTimestamp': tx[class="hl-string">'user_transaction_time'],
                    class="hl-string">'UserName': submitter_details[class="hl-string">'name'],
                    class="hl-string">'EmployeeDepartment': submitter_details[class="hl-string">'department'],
                    class="hl-string">'ReportTotalAmount': report_total,
                    class="hl-string">'ExpenseCategory': tx.get(class="hl-string">'merchant_name'),
                    class="hl-string">'PolicyViolationFlag': tx.get(class="hl-string">'flagged_for_review'),
                    class="hl-string">'RevisionReason': None
                })

                class=class="hl-string">"hl-comment"># Activity: Receipt Attached
                if tx.get(class="hl-string">'receipts') and len(tx[class="hl-string">'receipts']) > 0:
                    receipt_time = tx[class="hl-string">'receipts'][0].get(class="hl-string">'upload_time', tx[class="hl-string">'user_transaction_time']) class=class="hl-string">"hl-comment"># Use transaction time as fallback
                    event_log.append({
                        class="hl-string">'ExpenseReportId': case_id,
                        class="hl-string">'ActivityName': class="hl-string">'Receipt Attached',
                        class="hl-string">'EventTimestamp': receipt_time,
                        class="hl-string">'UserName': submitter_details[class="hl-string">'name'],
                        class="hl-string">'EmployeeDepartment': submitter_details[class="hl-string">'department'],
                        class="hl-string">'ReportTotalAmount': report_total,
                        class="hl-string">'ExpenseCategory': tx.get(class="hl-string">'merchant_name'),
                        class="hl-string">'PolicyViolationFlag': tx.get(class="hl-string">'flagged_for_review'),
                        class="hl-string">'RevisionReason': None
                    })

                class=class="hl-string">"hl-comment"># Activity: Policy Check Performed (Approximation)
                if tx.get(class="hl-string">'flagged_for_review'):
                    event_log.append({
                        class="hl-string">'ExpenseReportId': case_id,
                        class="hl-string">'ActivityName': class="hl-string">'Policy Check Performed',
                        class="hl-string">'EventTimestamp': r.get(class="hl-string">'created_at'), class=class="hl-string">"hl-comment"># Use submission time as proxy
                        class="hl-string">'UserName': class="hl-string">'System',
                        class="hl-string">'EmployeeDepartment': submitter_details[class="hl-string">'department'],
                        class="hl-string">'ReportTotalAmount': report_total,
                        class="hl-string">'ExpenseCategory': tx.get(class="hl-string">'merchant_name'),
                        class="hl-string">'PolicyViolationFlag': True,
                        class="hl-string">'RevisionReason': None
                    })

            except requests.exceptions.RequestException as e:
                print(fclass="hl-string">"Could not fetch transaction {tx_id}: {e}")

    if not event_log:
        print(class="hl-string">"No events were generated. Check the date range and API responses.")
        return

    class=class="hl-string">"hl-comment"># 4. Create and save DataFrame
    print(class="hl-string">"Finalizing event log...")
    df = pd.DataFrame(event_log)
    df[class="hl-string">'EventTimestamp'] = pd.to_datetime(df[class="hl-string">'EventTimestamp'])
    df = df.sort_values(by=[class="hl-string">'ExpenseReportId', class="hl-string">'EventTimestamp'])

    class=class="hl-string">"hl-comment"># Ensure required columns are present
    final_columns = [
        class="hl-string">'ExpenseReportId', class="hl-string">'ActivityName', class="hl-string">'EventTimestamp', class="hl-string">'UserName',
        class="hl-string">'EmployeeDepartment', class="hl-string">'ReportTotalAmount', class="hl-string">'ExpenseCategory',
        class="hl-string">'PolicyViolationFlag', class="hl-string">'RevisionReason'
    ]
    for col in final_columns:
        if col not in df.columns:
            df[col] = None
    
    df = df[final_columns]

    output_filename = class="hl-string">'ramp_event_log.csv'
    df.to_csv(output_filename, index=False, date_format=class="hl-string">'%Y-%m-%dT%H:%M:%S.%f%z')
    print(fclass="hl-string">"Extraction complete. Event log saved to {output_filename}")

if __name__ == class="hl-string">'__main__':
    main()

import pandas as pd
import numpy as np

# --- Configuration ---
TRANSACTIONS_FILE = 'transactions_export.csv'
REIMBURSEMENTS_FILE = 'reimbursements_export.csv'
OUTPUT_FILE = 'ramp_event_log.csv'

# --- Load Data ---
def load_data():
    try:
        transactions_df = pd.read_csv(TRANSACTIONS_FILE)
        reimbursements_df = pd.read_csv(REIMBURSEMENTS_FILE)
        print("Successfully loaded transactions and reimbursements files.")
        return transactions_df, reimbursements_df
    except FileNotFoundError as e:
        print(f"Error: {e}. Make sure '{TRANSACTIONS_FILE}' and '{REIMBURSEMENTS_FILE}' are in the same directory.")
        return None, None

# --- Main Transformation Logic ---
def transform_to_event_log(transactions_df, reimbursements_df):
    events = []

    # Standardize column names for easier processing
    # These names are based on typical Ramp CSV exports, adjust if your export has different headers
    transactions_df.rename(columns={
        'Transaction ID': 'ExpenseReportId',
        'User Name': 'UserName',
        'Department': 'EmployeeDepartment',
        'Amount': 'ReportTotalAmount',
        'Category Name': 'ExpenseCategory',
        'Transaction Date': 'ExpenseIncurredTimestamp',
        'Receipts Attached At': 'ReceiptAttachedTimestamp',
        'Submitted At': 'ExpenseSubmittedTimestamp',
        'Manager Approved At': 'ManagerApprovedTimestamp',
        'Needs Review At': 'FinanceReviewPendingTimestamp',
        'Admin Approved At': 'FinanceApprovedTimestamp',
        'Accounting Sync At': 'SyncedTimestamp',
        'Memo': 'RevisionReason',
        'State': 'Status'
    }, inplace=True)

    reimbursements_df.rename(columns={
        'Reimbursement ID': 'ExpenseReportId',
        'User Name': 'UserName',
        'Department': 'EmployeeDepartment',
        'Amount': 'ReportTotalAmount',
        'Category Name': 'ExpenseCategory',
        'Created At': 'ReimbursementScheduledTimestamp',
        'Paid Out At': 'ReimbursementExecutedTimestamp',
        'Memo': 'RevisionReason',
        'State': 'Status'
    }, inplace=True)

    # Combine card transactions and reimbursements into a single dataframe for processing common events
    # Note: Reimbursements will be processed separately for reimbursement-specific events
    combined_df = transactions_df

    # Process activities from the combined dataframe
    for _, row in combined_df.iterrows():
        base_event = {
            'ExpenseReportId': row.get('ExpenseReportId'),
            'UserName': row.get('UserName'),
            'EmployeeDepartment': row.get('EmployeeDepartment'),
            'ReportTotalAmount': row.get('ReportTotalAmount'),
            'ExpenseCategory': row.get('ExpenseCategory'),
            'PolicyViolationFlag': 'FLAGGED' in str(row.get('Flags')), # Infer from 'Flags' column
            'RevisionReason': row.get('RevisionReason')
        }

        # Activity: Expense Incurred
        if pd.notna(row.get('ExpenseIncurredTimestamp')):
            events.append({**base_event, 'ActivityName': 'Expense Incurred', 'EventTimestamp': row.get('ExpenseIncurredTimestamp')})

        # Activity: Receipt Attached
        if pd.notna(row.get('ReceiptAttachedTimestamp')):
            events.append({**base_event, 'ActivityName': 'Receipt Attached', 'EventTimestamp': row.get('ReceiptAttachedTimestamp')})

        # Activity: Expense Submitted
        if pd.notna(row.get('ExpenseSubmittedTimestamp')):
            events.append({**base_event, 'ActivityName': 'Expense Submitted', 'EventTimestamp': row.get('ExpenseSubmittedTimestamp')})
            # Inferred Activity: Policy Check Performed (assumed to happen right after submission)
            policy_check_ts = pd.to_datetime(row.get('ExpenseSubmittedTimestamp')) + pd.Timedelta(seconds=1)
            events.append({**base_event, 'ActivityName': 'Policy Check Performed', 'EventTimestamp': policy_check_ts.isoformat()})
            # Inferred Activity: Manager Review Pending (starts when submitted)
            events.append({**base_event, 'ActivityName': 'Manager Review Pending', 'EventTimestamp': row.get('ExpenseSubmittedTimestamp')})

        # Activity: Manager Approved
        if pd.notna(row.get('ManagerApprovedTimestamp')):
            events.append({**base_event, 'ActivityName': 'Manager Approved', 'EventTimestamp': row.get('ManagerApprovedTimestamp')})

        # Activity: Finance Review Pending
        if pd.notna(row.get('FinanceReviewPendingTimestamp')):
            events.append({**base_event, 'ActivityName': 'Finance Review Pending', 'EventTimestamp': row.get('FinanceReviewPendingTimestamp')})

        # Activity: Finance Approved
        if pd.notna(row.get('FinanceApprovedTimestamp')):
            events.append({**base_event, 'ActivityName': 'Finance Approved', 'EventTimestamp': row.get('FinanceApprovedTimestamp')})

        # Activity: Synced with Accounting System
        if pd.notna(row.get('SyncedTimestamp')):
            events.append({**base_event, 'ActivityName': 'Synced with Accounting System', 'EventTimestamp': row.get('SyncedTimestamp')})

        # Activity: Expense Sent Back for Revision
        if row.get('Status') == 'REJECTED' and pd.notna(row.get('Updated At')):
             events.append({**base_event, 'ActivityName': 'Expense Sent Back for Revision', 'EventTimestamp': row.get('Updated At')}) # Using last updated time for rejection

    # Process reimbursement-specific activities
    for _, row in reimbursements_df.iterrows():
        base_event = {
            'ExpenseReportId': row.get('ExpenseReportId'),
            'UserName': row.get('UserName'),
            'EmployeeDepartment': row.get('EmployeeDepartment'),
            'ReportTotalAmount': row.get('ReportTotalAmount'),
            'ExpenseCategory': row.get('ExpenseCategory'),
            'PolicyViolationFlag': False, # Typically less flagging on reimbursements
            'RevisionReason': row.get('RevisionReason')
        }

        # Activity: Reimbursement Scheduled
        if row.get('Status') == 'APPROVED' and pd.notna(row.get('ReimbursementScheduledTimestamp')):
            events.append({**base_event, 'ActivityName': 'Reimbursement Scheduled', 'EventTimestamp': row.get('ReimbursementScheduledTimestamp')})

        # Activity: Reimbursement Executed
        if row.get('Status') == 'PAID' and pd.notna(row.get('ReimbursementExecutedTimestamp')):
            events.append({**base_event, 'ActivityName': 'Reimbursement Executed', 'EventTimestamp': row.get('ReimbursementExecutedTimestamp')})

    if not events:
        print("Warning: No events were generated. Check the input files and column names.")
        return None

    # Create final DataFrame
    event_log_df = pd.DataFrame(events)
    event_log_df['EventTimestamp'] = pd.to_datetime(event_log_df['EventTimestamp'])
    event_log_df.sort_values(by=['ExpenseReportId', 'EventTimestamp'], inplace=True)
    
    return event_log_df

# --- Main execution block ---
if __name__ == "__main__":
    transactions_df, reimbursements_df = load_data()
    if transactions_df is not None and reimbursements_df is not None:
        event_log = transform_to_event_log(transactions_df, reimbursements_df)
        if event_log is not None:
            event_log.to_csv(OUTPUT_FILE, index=False, date_format='%Y-%m-%dT%H:%M:%S.%fZ')
            print(f"Successfully created event log with {len(event_log)} events. Saved to '{OUTPUT_FILE}'.")