Data Types

ScramDB supports the most common PostgreSQL data types. The table below shows the mapping between PostgreSQL types and their ScramDB equivalents.

PostgreSQL → ScramDB Type Mapping

PostgreSQL Type	ScramDB Type	Size	Status	Notes
SMALLINT / INT2	Int16	2 bytes	✅	Stored natively as 16-bit
INTEGER / INT4	Int32	4 bytes	✅
BIGINT / INT8	Int64	8 bytes	✅
REAL / FLOAT4	Float32	4 bytes	✅	IEEE 754 single precision
DOUBLE PRECISION / FLOAT8	Float64	8 bytes	✅	IEEE 754 double precision
BOOLEAN	Boolean	1 bit	✅	Validity bitmask
VARCHAR(n) / TEXT	Utf8	Variable	✅	Variable-length UTF-8
CHAR(n)	Utf8	Variable	✅	Fixed-width, space-padded
DATE	Date32	4 bytes	✅	Days since epoch
TIMESTAMP	TimestampMicro	8 bytes	✅	Microsecond precision
DECIMAL(p,s) / NUMERIC	Decimal128	16 bytes	✅	128-bit decimal with precision/scale
INTERVAL	IntervalMDN	16 bytes	✅	Month-Day-Nano representation
VECTOR(n)	Vector	n × 4 bytes	✅	pgvector-compatible, float32 elements
UUID	-	-	⚠️ Planned	v28
JSONB	-	-	⚠️ Planned	v28
ARRAY types	-	-	⚠️ Planned	v28
BYTEA	-	-	⚠️ Planned
SERIAL / BIGSERIAL	-	-	⚠️ Planned	v27 (auto-increment)

Type Behavior

Numeric Types

• Integer arithmetic follows standard SQL semantics (exact)
• Floating-point arithmetic uses IEEE 754
• Integer overflow is checked at runtime
• Implicit casting follows PostgreSQL promotion rules: INT2 → INT4 → INT8 → FLOAT4 → FLOAT8

String Types

• All strings are stored as UTF-8
• VARCHAR(n) enforces max length at insertion
• TEXT and VARCHAR (without length) are equivalent
• String comparison is byte-wise (no collation support yet)
• Short strings (≤8 bytes) are optimized for JIT - packed into 64-bit registers

Date/Time Types

• DATE is stored as days since Unix epoch (1970-01-01)
• TIMESTAMP is stored as microseconds since epoch
• EXTRACT() supports: YEAR, MONTH, DAY, HOUR, MINUTE, SECOND, EPOCH, DOW, DOY, QUARTER, WEEK
• Date arithmetic is implemented via Howard Hinnant's civil date algorithm

NULL Handling

• All types are nullable
• NULL semantics follow SQL standard (three-valued logic)
• IS NULL / IS NOT NULL for NULL checks
• COALESCE() and NULLIF() are supported
• Validity tracked via per-column bitmask (1 bit per row)

Type Casting

Explicit casts use the CAST function or :: syntax:

SELECT CAST(price AS INTEGER) FROM products;
SELECT price::INTEGER FROM products;

Supported Casts

From	To	Method
Any integer	Any integer	Truncation/extension
Integer	Float	Exact conversion
Float	Integer	Truncation
String	Integer/Float	Parse
Integer/Float	String	Format
String	Date/Timestamp	Parse (ISO 8601)
Date	Timestamp	Midnight
Timestamp	Date	Truncate time