How to write a CSV file with pandas

Delimited text remains a common handoff between data analysis, spreadsheets, databases, and automation systems. A pandas DataFrame can become a CSV file with an explicit column order and text format that the receiving system can inspect without a Python runtime.

Pandas writes a DataFrame to a path, path-like object, file-like object, or an in-memory string through DataFrame.to_csv(). A path-based export leaves a file that can be inspected and read back, while index=False prevents the row labels from becoming an extra CSV column.

CSV preserves cell text rather than enforcing spreadsheet safety rules. Review values from untrusted sources for formula-leading equals, plus, minus, and at signs before sharing an export that someone will open in spreadsheet software.

1. Create write_orders_csv.py with the initial output-path setup.

   write_orders_csv.py

   from pathlib import Path
    
   import pandas as pd
    
   output = Path("exports/orders.csv")
   output.parent.mkdir(parents=True, exist_ok=True)

   to_csv() does not create missing parent directories, so the output directory must exist before the export runs.

2. Append the order data block below the output-directory setup in write_orders_csv.py.

   orders = pd.DataFrame(
       {
           "order_id": pd.Series(["A100", "A101", "A102"], dtype="string"),
           "customer": ["Ada Lovelace", "Lin Chen", "Maya Patel"],
           "region": ["EMEA", "APAC", "AMER"],
           "total_usd": [149.5, None, 212.0],
           "ordered_at": pd.to_datetime(
               ["2026-06-01", "2026-06-02", "2026-06-03"]
           ),
       }
   )

3. Append the CSV export block below the DataFrame definition in write_orders_csv.py.

   orders.to_csv(
       output,
       sep=",",
       index=False,
       columns=["order_id", "customer", "region", "total_usd", "ordered_at"],
       na_rep="",
       float_format="%.2f",
       date_format="%Y-%m-%d",
       encoding="utf-8",
       lineterminator="\n",
   )
    
   print(f"Wrote {output}")

   columns fixes the export order, na_rep=“” leaves the missing total blank, and the format options make dates, decimal places, encoding, and line endings explicit.

4. Run write_orders_csv.py to create the CSV file.

   $ python3 write_orders_csv.py
   Wrote exports/orders.csv

5. Inspect the saved CSV text at exports/orders.csv.

   $ cat exports/orders.csv
   order_id,customer,region,total_usd,ordered_at
   A100,Ada Lovelace,EMEA,149.50,2026-06-01
   A101,Lin Chen,APAC,,2026-06-02
   A102,Maya Patel,AMER,212.00,2026-06-03

   The file should contain the selected header order, three data rows, one blank total, ISO-style dates, and two-decimal totals.
   Tool: Comma-Separated Values (CSV) Converter

6. Read exports/orders.csv back with pandas to verify the exported table.

   $ python3 - <<'PY'
   from pathlib import Path
   
   import pandas as pd
   
   csv_path = Path("exports/orders.csv")
   round_trip = pd.read_csv(
       csv_path,
       dtype={"order_id": "string"},
       parse_dates=["ordered_at"],
   )
   
   expected_columns = [
       "order_id",
       "customer",
       "region",
       "total_usd",
       "ordered_at",
   ]
   
   assert round_trip.columns.tolist() == expected_columns
   assert round_trip["order_id"].tolist() == ["A100", "A101", "A102"]
   assert round_trip["total_usd"].isna().sum() == 1
   assert round_trip["ordered_at"].dt.strftime("%Y-%m-%d").tolist() == [
       "2026-06-01",
       "2026-06-02",
       "2026-06-03",
   ]
   
   print(round_trip.to_string(index=False))
   print(f"\nRows: {len(round_trip)}")
   print(f"Columns: {', '.join(round_trip.columns)}")
   print(f"Missing totals: {round_trip['total_usd'].isna().sum()}")
   PY
   order_id     customer region  total_usd ordered_at
       A100 Ada Lovelace   EMEA      149.5 2026-06-01
       A101     Lin Chen   APAC        NaN 2026-06-02
       A102   Maya Patel   AMER      212.0 2026-06-03
   
   Rows: 3
   Columns: order_id, customer, region, total_usd, ordered_at
   Missing totals: 1

   The read-back command fails if the file is missing, the schema changes, an order ID is altered, the missing total is lost, or a date no longer matches the expected value.