Bitmaps

Bit-level operations on strings for efficient analytics, permissions, and flags

Redis bitmaps are not a separate data type but bit-level operations on strings. They provide space-efficient storage for binary flags, user activity tracking, and analytics with up to 512MB (4 billion bits) per key.

Use Cases

• User activity tracking: Daily active users, feature usage
• Permissions and flags: User roles, feature flags
• Real-time analytics: Unique visitors, conversion tracking
• A/B testing: Experiment participation tracking
• Bloom filters: Approximate set membership
• Attendance tracking: Event participation, presence

Key Commands

Basic Operations

# Set bit at offset
redis> SETBIT users:active:2024-03-03 123 1
(integer) 0
redis> SETBIT users:active:2024-03-03 456 1
(integer) 0

# Get bit at offset
redis> GETBIT users:active:2024-03-03 123
(integer) 1
redis> GETBIT users:active:2024-03-03 789
(integer) 0

# Count set bits
redis> BITCOUNT users:active:2024-03-03
(integer) 2

# Count in range (byte offsets)
redis> BITCOUNT users:active:2024-03-03 0 10
(integer) 2

Bit Operations

# Set multiple bits
redis> SETBIT day1 0 1
(integer) 0
redis> SETBIT day1 1 0
(integer) 0
redis> SETBIT day1 2 1
(integer) 0

redis> SETBIT day2 0 1
(integer) 0
redis> SETBIT day2 1 1
(integer) 0
redis> SETBIT day2 2 0
(integer) 0

# AND operation (users active both days)
redis> BITOP AND result day1 day2
(integer) 1
redis> BITCOUNT result
(integer) 1

# OR operation (users active either day)
redis> BITOP OR result day1 day2
(integer) 1
redis> BITCOUNT result
(integer) 3

# XOR operation (users active only one day)
redis> BITOP XOR result day1 day2
(integer) 1
redis> BITCOUNT result
(integer) 2

# NOT operation (invert bits)
redis> BITOP NOT result day1
(integer) 1

Position Operations

# Find first set bit
redis> SETBIT bits 10 1
(integer) 0
redis> SETBIT bits 20 1
(integer) 0
redis> BITPOS bits 1
(integer) 10

# Find first clear bit
redis> BITPOS bits 0
(integer) 0

# Find in range
redis> BITPOS bits 1 0 2
(integer) 10

Bitfield Operations

# Set integer value at bit position
redis> BITFIELD stats SET u8 0 255
1) (integer) 0

# Get integer value
redis> BITFIELD stats GET u8 0
1) (integer) 255

# Increment
redis> BITFIELD stats INCRBY u8 0 1
1) (integer) 256

# Multiple operations
redis> BITFIELD stats 
  SET u4 0 15 
  GET u4 0 
  INCRBY u4 0 1
1) (integer) 0
2) (integer) 15
3) (integer) 0

Time Complexity

Command	Time Complexity	Description
SETBIT	O(1)	Set bit value
GETBIT	O(1)	Get bit value
BITCOUNT	O(N)	N=bytes (with AVX2: very fast)
BITPOS	O(N)	N=bytes scanned
BITOP	O(N)	N=longest string
BITFIELD	O(1) per operation	Multiple field operations

BITCOUNT is highly optimized with AVX2/AVX512/NEON SIMD instructions on modern CPUs, making it extremely fast even for large bitmaps.

Patterns and Examples

Daily Active Users

# Mark users active (user ID as bit offset)
redis> SETBIT users:active:2024-03-03 1001 1
(integer) 0
redis> SETBIT users:active:2024-03-03 1002 1
(integer) 0
redis> SETBIT users:active:2024-03-03 5000 1
(integer) 0

# Count daily active users
redis> BITCOUNT users:active:2024-03-03
(integer) 3

# Check if user was active
redis> GETBIT users:active:2024-03-03 1001
(integer) 1

# Find users active both days
redis> BITOP AND active_both users:active:2024-03-03 users:active:2024-03-04
(integer) 625
redis> BITCOUNT active_both
(integer) 2

# Monthly active users (union of all days)
redis> BITOP OR mau:2024-03 users:active:2024-03-01 users:active:2024-03-02 ...
(integer) 625
redis> BITCOUNT mau:2024-03
(integer) 4523

Feature Flags

# Define feature bits: 0=beta, 1=premium, 2=admin, 3=verified
redis> SETBIT user:1001:flags 0 1  # Beta access
(integer) 0
redis> SETBIT user:1001:flags 3 1  # Verified
(integer) 0

# Check if user has feature
redis> GETBIT user:1001:flags 0
(integer) 1  # Has beta access

redis> GETBIT user:1001:flags 2
(integer) 0  # Not admin

# Grant premium feature
redis> SETBIT user:1001:flags 1 1
(integer) 0

# Revoke beta access
redis> SETBIT user:1001:flags 0 0
(integer) 1

Real-Time Analytics

# Track unique visitors per hour (IP hash as offset)
redis> SETBIT visitors:2024-03-03:10 12345 1
(integer) 0
redis> SETBIT visitors:2024-03-03:10 67890 1
(integer) 0

# Get unique visitors for hour
redis> BITCOUNT visitors:2024-03-03:10
(integer) 2

# Unique visitors for entire day
redis> BITOP OR visitors:2024-03-03 
  visitors:2024-03-03:00 
  visitors:2024-03-03:01 
  ...
  visitors:2024-03-03:23
(integer) 1250
redis> BITCOUNT visitors:2024-03-03
(integer) 8945

User Permissions

# Permission bits: 0=read, 1=write, 2=delete, 3=admin
redis> SETBIT perms:user:123:doc:456 0 1  # Read
(integer) 0
redis> SETBIT perms:user:123:doc:456 1 1  # Write
(integer) 0

# Check permission
redis> GETBIT perms:user:123:doc:456 2
(integer) 0  # No delete permission

# Grant all permissions
redis> SETBIT perms:user:123:doc:456 0 1
redis> SETBIT perms:user:123:doc:456 1 1
redis> SETBIT perms:user:123:doc:456 2 1
redis> SETBIT perms:user:123:doc:456 3 1

Attendance Tracking

# Track attendance (day of year as bit)
redis> SETBIT attendance:student:123:2024 62 1  # March 3rd (day 62)
(integer) 0
redis> SETBIT attendance:student:123:2024 63 1  # March 4th
(integer) 0

# Count days present
redis> BITCOUNT attendance:student:123:2024
(integer) 2

# Find first absence
redis> BITPOS attendance:student:123:2024 0
(integer) 0  # Day 0 (Jan 1st)

# Check specific day
redis> GETBIT attendance:student:123:2024 62
(integer) 1  # Present

Memory Efficiency

Bitmaps are extremely space-efficient:

# Track 1 million users: only 125KB
# 1,000,000 bits / 8 = 125,000 bytes = 125 KB

redis> SETBIT users 1000000 1
(integer) 0
redis> MEMORY USAGE users
(integer) 125056  # ~125KB

# Compare with set of 1M user IDs: ~15-20MB
redis> SADD users_set user:1 user:2 ... user:1000000
# Much larger memory usage

Sparse vs Dense Bitmaps

# Sparse bitmap (few bits set)
redis> SETBIT sparse 1000000 1
(integer) 0
redis> MEMORY USAGE sparse
(integer) 125032  # Still allocates full range

# For sparse data, consider sets
redis> SADD sparse_set 1000000
(integer) 1
redis> MEMORY USAGE sparse_set
(integer) 296  # Much smaller for sparse data

Bitmaps allocate memory based on the highest bit offset. SETBIT at offset 1 billion requires 125MB even if only one bit is set. For sparse data, use sets or hashes instead.

Best Practices

1. Use for dense data - when most IDs in a range are used
2. Combine with BITOP for efficient set operations
3. Set expiration on daily/hourly bitmaps
4. Use consistent key naming for easy BITOP operations
5. Consider sets for sparse data (few bits set)

For optimal performance with BITOP on large bitmaps, ensure all operand keys exist. Missing keys are treated as all zeros but still affect performance.

Bitmaps vs Other Types

Use Case	Bitmap	Set	HyperLogLog
Dense IDs (0-1M)	Best	High memory	Approximate
Sparse IDs	High memory	Best	Approximate
Exact count	Yes	Yes	No (±0.81%)
Set operations	BITOP (fast)	SINTER/SUNION	No
Memory (1M IDs)	125KB	15-20MB	12KB
Best for	Sequential IDs	Sparse IDs	Counting only

Advanced: Bitfield

Store multiple integers in a single string:

# Store RGB color (3 bytes)
redis> BITFIELD color SET u8 0 255 SET u8 8 128 SET u8 16 64
1) (integer) 0
2) (integer) 0
3) (integer) 0

# Get RGB values
redis> BITFIELD color GET u8 0 GET u8 8 GET u8 16
1) (integer) 255
2) (integer) 128
3) (integer) 64

# Increment red channel
redis> BITFIELD color INCRBY u8 0 10
1) (integer) 265

Next Steps

String Commands

Bitmap command reference

HyperLogLog

For approximate counting