How to convert columns to categorical data in pandas

Repeated labels in a DataFrame often represent a small vocabulary rather than unrestricted text. The pandas category dtype stores those labels with category metadata and integer codes, which also lets downstream operations recognize the column as categorical.

An inferred conversion is suitable when every label already present in the column is valid and no label order is required. CategoricalDtype is the stronger choice when the allowed labels or their logical order must remain explicit.

An explicit category list turns any unlisted source value into a missing value during conversion. Check the category metadata and missing-value counts before sorting, grouping, plotting, or exporting the converted DataFrame.

1. Create the sample ticket DataFrame in categorical_dtype.py.

   categorical_dtype.py

   import pandas as pd
   from pandas.api.types import CategoricalDtype
    
   df = pd.DataFrame(
       {
           "ticket": [101, 102, 103, 104, 105, 106],
           "team": ["api", "frontend", "api", "ops", "frontend", "api"],
           "priority": ["normal", "urgent", "low", "normal", "low", "urgent"],
       }
   )

   The sample DataFrame supplies a runnable starting point; an existing DataFrame can use the same column-level conversions.

2. Append the inferred categorical conversion for the team column.

   df["team"] = df["team"].astype("category")

   astype(“category”) infers an unordered category set from the labels present in the column.

3. Define the ordered category vocabulary for the priority column.

   priority_dtype = CategoricalDtype(
       categories=["low", "normal", "urgent"],
       ordered=True,
   )

4. Append the explicit categorical conversion for the priority column.

   df["priority"] = df["priority"].astype(priority_dtype)

   A non-missing source value outside low, normal, and urgent becomes missing during this conversion. The allowed category list must cover every valid production label.

5. Append the validation block below to categorical_dtype.py.

   assert str(df["team"].dtype) == "category"
   assert df["priority"].cat.ordered
   assert df["priority"].cat.categories.to_list() == ["low", "normal", "urgent"]
   assert df[["team", "priority"]].isna().sum().eq(0).all()
    
   print("column dtypes")
   print(df.dtypes)
   print()
   print("team categories")
   print(df["team"].cat.categories.to_list())
   print()
   print("priority categories")
   print(df["priority"].cat.categories.to_list())
   print("priority ordered")
   print(df["priority"].cat.ordered)
   print()
   print("missing after conversion")
   print(df[["team", "priority"]].isna().sum())
   print()
   print("sorted by priority")
   print(df.sort_values("priority")[["ticket", "priority"]].to_string(index=False))

   The assertions stop the script when either dtype, the declared order, or the missing-value boundary differs from the intended conversion.

6. Run categorical_dtype.py to confirm the converted metadata and priority sort order.

   $ python3 categorical_dtype.py
   column dtypes
   ticket         int64
   team        category
   priority    category
   dtype: object
   
   team categories
   ['api', 'frontend', 'ops']
   
   priority categories
   ['low', 'normal', 'urgent']
   priority ordered
   True
   
   missing after conversion
   team        0
   priority    0
   dtype: int64
   
   sorted by priority
    ticket priority
       103      low
       105      low
       101   normal
       104   normal
       102   urgent
       106   urgent

   The zero missing-value counts preserve every source label, while the sorted rows follow low, normal, and urgent instead of lexical order.