How to filter rows in a pandas DataFrame

Filtering rows in a pandas DataFrame keeps records that meet a condition before analysis, reporting, or export. Boolean masks work well for thresholds, membership checks, status fields, and text matches because each row receives a True or False decision.

Use .loc when the filter also selects output columns, because the expression before the comma chooses rows and the expression after the comma chooses columns. A mask built from a column comparison must align with the DataFrame rows, and pandas preserves the original index labels in the filtered result.

Combine conditions with & and | inside parentheses rather than Python and or or, which cannot reduce a Series of booleans to one truth value. Create a .copy() of the filtered result before changing it independently from the source DataFrame.

1. Save a short row-filtering script.

   filter_rows.py

   import pandas as pd
    
   orders = pd.DataFrame(
       {
           "order_id": [1001, 1002, 1003, 1004, 1005, 1006],
           "customer": ["Ada", "Lin", "Maya", "Omar", "Nia", "Kai"],
           "region": ["EMEA", "APAC", "AMER", "EMEA", "APAC", "AMER"],
           "category": ["hardware", "software", "hardware", "service", "software", "hardware"],
           "qty": [3, 12, 7, 2, 15, 9],
           "total_usd": [150.0, 240.0, 875.0, 95.0, 360.0, 420.0],
           "status": ["paid", "paid", "open", "paid", "open", "paid"],
       }
   )
    
   print(f"pandas {pd.__version__}")
   print()
    
   print("BASE")
   print(orders.to_string(index=False))
   print()
    
   high_value = orders.loc[
       orders["total_usd"] >= 300,
       ["order_id", "customer", "region", "total_usd"],
   ]
    
   print("FILTER_TOTAL")
   print(high_value.to_string(index=False))
   print()
    
   regions = ["EMEA", "APAC"]
   priority_mask = (orders["region"].isin(regions)) & (orders["qty"] >= 10)
   priority = orders.loc[
       priority_mask,
       ["order_id", "customer", "region", "qty", "total_usd"],
   ]
    
   print("FILTER_REGION_QTY")
   print(priority.to_string(index=False))
   print()
    
   paid_hardware = orders.loc[
       (orders["status"] == "paid") & (orders["category"] == "hardware"),
       ["order_id", "customer", "category", "status", "total_usd"],
   ]
    
   print("FILTER_STATUS_CATEGORY")
   print(paid_hardware.to_string(index=False))
   print()
    
   customer_match = orders.loc[
       orders["customer"].str.contains("a", case=False, regex=False, na=False),
       ["order_id", "customer", "region"],
   ]
    
   print("FILTER_CUSTOMER_TEXT")
   print(customer_match.to_string(index=False))
   print()
    
   print("VERIFY")
   print(f"priority rows: {len(priority)}")
   print(f"priority index: {priority.index.tolist()}")
   print(f"all priority qty >= 10: {priority['qty'].ge(10).all()}")
   print(f"source rows unchanged: {len(orders)}")

   Replace the sample orders object with the DataFrame already loaded in the working script.

2. Run the script and confirm that each section returns only matching rows.

   $ python3 filter_rows.py
   pandas 3.0.3
   
   BASE
    order_id customer region category  qty  total_usd status
        1001      Ada   EMEA hardware    3      150.0   paid
        1002      Lin   APAC software   12      240.0   paid
        1003     Maya   AMER hardware    7      875.0   open
        1004     Omar   EMEA  service    2       95.0   paid
        1005      Nia   APAC software   15      360.0   open
        1006      Kai   AMER hardware    9      420.0   paid
   
   FILTER_TOTAL
    order_id customer region  total_usd
        1003     Maya   AMER      875.0
        1005      Nia   APAC      360.0
        1006      Kai   AMER      420.0
   
   FILTER_REGION_QTY
    order_id customer region  qty  total_usd
        1002      Lin   APAC   12      240.0
        1005      Nia   APAC   15      360.0
   
   FILTER_STATUS_CATEGORY
    order_id customer category status  total_usd
        1001      Ada hardware   paid      150.0
        1006      Kai hardware   paid      420.0
   
   FILTER_CUSTOMER_TEXT
    order_id customer region
        1001      Ada   EMEA
        1003     Maya   AMER
        1004     Omar   EMEA
        1005      Nia   APAC
        1006      Kai   AMER
   
   VERIFY
   priority rows: 2
   priority index: [1, 4]
   all priority qty >= 10: True
   source rows unchanged: 6

3. Filter rows with a numeric threshold.

   high_value = orders.loc[
       orders["total_usd"] >= 300,
       ["order_id", "customer", "region", "total_usd"],
   ]

   The row condition comes before the comma, and the selected output columns come after it.
   Related: How to select DataFrame rows and columns with loc and iloc in pandas

4. Match rows whose value appears in an allowed list.

   regions = ["EMEA", "APAC"]
   priority_mask = (orders["region"].isin(regions)) & (orders["qty"] >= 10)
   priority = orders.loc[
       priority_mask,
       ["order_id", "customer", "region", "qty", "total_usd"],
   ]

   Use parentheses around each comparison when combining masks. Python and and or raise an ambiguous truth-value error with pandas Series objects.

5. Filter rows by exact status and category values.

   paid_hardware = orders.loc[
       (orders["status"] == "paid") & (orders["category"] == "hardware"),
       ["order_id", "customer", "category", "status", "total_usd"],
   ]

6. Filter text rows with a string condition.

   customer_match = orders.loc[
       orders["customer"].str.contains("a", case=False, regex=False, na=False),
       ["order_id", "customer", "region"],
   ]

   na=False keeps missing text values from matching the filter.

7. Verify the filtered row count, preserved index labels, and source row count.

   print(f"priority rows: {len(priority)}")
   print(f"priority index: {priority.index.tolist()}")
   print(f"all priority qty >= 10: {priority['qty'].ge(10).all()}")
   print(f"source rows unchanged: {len(orders)}")

8. Copy a filtered result before changing it independently.

   priority_review = priority.copy()
   priority_review["review_required"] = True

   .copy() makes the filtered DataFrame a separate working object before later assignment.
   Related: How to migrate pandas code for Copy-on-Write

9. Remove the sample script after confirming the filter behavior.

   $ rm filter_rows.py