BITMAP Index

A bitmap index speeds up filtering on low-cardinality scalar fields — fields that have a small number of distinct values (for example: is_public, status, category, tier, region).

Instead of scanning every record to evaluate a filter, a bitmap index precomputes compact bitsets and uses fast bitwise operations (AND/OR/NOT) to narrow results.

When to use BITMAP

Use a bitmap index when:

You filter heavily on a field with few distinct values (roughly < 500 unique values is a good rule of thumb).
Your queries often combine multiple filters (example: status == "active" AND is_public == true).

Avoid bitmap indexes when:

The field has high cardinality (like user_id, session_id, timestamps, or random strings). In that case, the bitmap becomes large and the benefit drops.

How it works (simple intuition)

Imagine a collection with category and is_public.

For category = "tech", the engine keeps a bitmap like: [1, 0, 1, 0, 0]
For is_public = true, it keeps another bitmap like: [1, 0, 0, 1, 0]

To answer:

category == "tech" AND is_public == true

it performs a fast bitwise AND on those bitmaps to find matching records—without scanning the full dataset.

Create a BITMAP index

In Dodil, you create scalar indexes through the VBase client.

Note: The exact method names may vary slightly by SDK version. If yours differs, keep the same intent and parameters: field_name, index_type="BITMAP", and index_name.


# Build index params
index_params = vbase.create_index_params()
index_params.add_index(
    field_name="category",
    index_type="BITMAP",
    index_name="category_bitmap_index",
)
 
# Create the index
vbase.create_index(
    collection_name="my_collection",
    index_params=index_params,
)