diff --git a/write_to_db.py b/write_to_db.py
index b22820e..5a4ab40 100644
--- a/write_to_db.py
+++ b/write_to_db.py
@@ -1,7 +1,9 @@
 import os
 import pandas as pd
-from sqlalchemy import create_engine
+from sqlalchemy import create_engine, MetaData, Table
+from sqlalchemy.dialects.mysql import insert
 from dotenv import load_dotenv
+import numpy as np
 
 # Load environment variables from .env file
 load_dotenv()
@@ -27,6 +29,9 @@ file_paths = [
     ('sec_data/2015q1/pre.txt', 'pre', ['adsh', 'report', 'line'])
 ]
 
+# Initialize metadata
+metadata = MetaData()
+
 # Loop through each file and write the data to the database
 for i, (file_path, table_name, primary_keys) in enumerate(file_paths):
     print(f"\nAnalyzing {file_path} (File {i+1}/4)...")
@@ -44,18 +49,41 @@ for i, (file_path, table_name, primary_keys) in enumerate(file_paths):
         print("\nUpdated 'coreg' column (NaN values replaced with 'nocoreg'):")
         print(df[['coreg']].head(10))  # Display first 10 rows of the 'coreg' column for verification
     
-    # Dropping rows with any missing values in the primary keys and NOT NULL columns
+    # Dropping rows with any missing values in the primary keys
     df.dropna(subset=primary_keys, inplace=True)
 
     # Dropping duplicate rows based on primary keys
-    df.drop_duplicates(subset=primary_keys, keep='first', inplace=True)
+    # df.drop_duplicates(subset=primary_keys, keep='first', inplace=True)
+    
+    # Replace NaN values with None to ensure compatibility with SQL NULL
+    df = df.replace([np.nan, np.inf, -np.inf], None)
     
     # Get Updated Information
     print("\nUpdated Information:")
     print(df.info())
 
-    # Write the cleaned DataFrame to the corresponding table in the MariaDB database
-    df.to_sql(table_name, con=engine, if_exists='append', index=False)
-    print(f"\nCleaned data from {file_path} has been written to the '{table_name}' table in the database.\n")
+    # Reflect the already existing table from the database schema
+    table = Table(table_name, metadata, autoload_with=engine)
 
-print("\nAll files have been processed and cleaned data has been written to the database.")
\ No newline at end of file
+    # Perform Upsert operation for each row in the DataFrame
+    with engine.connect() as conn:
+        for row in df.itertuples(index=False):
+            # Create a dictionary of the row data
+            data = {key: getattr(row, key) for key in df.columns}
+            
+            # Prepare insert statement using SQLAlchemy with MySQL-specific ON DUPLICATE KEY UPDATE
+            insert_stmt = insert(table).values(**data)
+            
+            # Construct the `ON DUPLICATE KEY UPDATE` part
+            update_stmt = insert_stmt.on_duplicate_key_update(
+                {col.name: insert_stmt.inserted[col.name] for col in table.columns}
+            )
+            
+            # Execute the upsert statement
+            conn.execute(update_stmt)
+    
+    print(f"\nCleaned data from {file_path} has been written to the '{table_name}' table in the database with upsert functionality.\n")
+
+print("\nAll files have been processed and cleaned data has been written to the database.")
+
+#FIXME: Foreign key missing because usgapp is in the past constantly, Q1 gaap is based on the year before gaap.
\ No newline at end of file