Expanded Data Types Introduced in V15

PBXB64 V18 retains the complete managed UN family introduced in V15 alongside standard PowerBASIC types and specialized numeric, COM, string, and wide-integer extensions: UNL, UNF, UNB, UNC, UNI, and UNQ.

Complete Type Reference

The V15 UN Family

These managed types store numbers beyond the fixed limits of native machine types. They support declarations, assignments, conversions, arithmetic, comparisons, globals, parameters, return values, fixed and dynamic arrays, and automatic cleanup.

Scalar assignment to UNC, UNI, or UNQ fills the first component and initializes the remaining components to zero. Invalid UN literals and conversions produce explicit diagnostics.

UNL and UNx in Real Code

The UN family is superior when a fixed native type would overflow, lose required precision, or force several related numeric components into a hand-built structure. The values remain strongly typed, work with PBXB64 expressions and calls, and clean up their managed storage automatically. Native integers and floating-point types remain the faster, smaller choice when their range and precision are sufficient.

Native H-Lib Data Structure Types

PBXB64 also treats the native H-Lib containers as first-class managed data types. All 22 core families plus the Safe synchronized wrapper are built into the compiler, require no COM object or external library, use dot-method syntax, and release their internal storage automatically when the declaring scope ends.

Type prefixes select the stored PB value: for example LnArr stores LONG, DbArr stores DOUBLE, SsArr stores ANSI strings, and WsArr stores Unicode strings. Up to 13 element variants across the families provide more than 125 recognized container types; supported families can also use UDT element forms such as AS LnArr OF MyType.

DEC128 - 18-Digit Decimal Precision

PBXB64 Addition A 16-byte decimal fixed-point type with 18-digit precision. Ideal for financial calculations where binary floating-point rounding errors are unacceptable.

DEC128 stores values as decimal digits, not binary fractions. 19.99 + 0.01 equals exactly 20.00, not 19.999999999999996.

RATIONAL - Exact Arithmetic

PBXB64 Addition Stores numbers as int64 numerator/denominator pairs with automatic GCD normalization. 1/3 + 1/3 = 2/3 exactly. No rounding, ever.

RATIONAL automatically reduces fractions (GCD normalization) and supports all arithmetic operations: +, -, *, /, comparison, conversion to/from DECIMAL/DOUBLE.

VARIANT - COM-Compatible Values

PBXB64 Addition Full VARIANT implementation with VT_EMPTY, VT_I8, VT_R8, VT_CY, VT_BOOL, VT_BSTR. Supports scalar, fixed arrays, and dynamic arrays with REDIM PRESERVE.

VARIANT arrays support element-wise lifecycle management. REDIM PRESERVE performs deep element copy.

Wide Integers - 128 / 256 / 512 Bit

PBXB64 Addition Inline arithmetic on 128, 256, and 512-bit integers. No library calls, no heap allocation - pure register-based operations.

Wide integers support addition, subtraction, multiplication, division, comparison, and bitwise operations. All done inline in the x64 backend without external libraries.

Datatype Competition: Where They Lead, Where PBXB64 Wins

No compiler is universally superior in every meaning of “datatype.” GHC is more expressive at the type level, Rust is stronger at compile-time memory safety, Ada/SPARK at constrained types and formal assurance, and Julia in scientific dispatch and ecosystem depth. PBXB64's defensible lead is out-of-the-box concrete datatype breadth in one native Windows toolchain: UNL, UNF, UNB, UNC, UNI, UNQ, DEC128, RATIONAL, 128/256/512-bit integers, Windows/COM types, and 22 H-Lib families with 125+ typed variants.

Scope: major relevant compiler families in their standard or commonly shipped form, not every research compiler or third-party package. “PBXB64 wins” means the capability is integrated in the V18 compiler/package and fits the stated use case; it is not a claim that PBXB64 has the world's most expressive type system. Comparison anchors: GHC GADTs, Rust generics, traits and lifetimes, Julia types, and Ada 2022.

Datatype Feature Availability Matrix

This matrix compares the concrete PBXB64 datatype families above with the same 15 toolchains. An orange tools symbol is only awarded when a named, established package or platform route supplies a practical equivalent. A type that could merely be hand-written from scratch remains unavailable here.

Swipe horizontally to compare every compiler.

“Built in” includes the language's normal standard or official platform library. “Library / emulation” means the green result depends on the named package, binding, or composition and may not match PBXB64's syntax, precision model, lifecycle, or deployment profile exactly. Evidence anchors include Julia numbers, .NET numerics, D BigInt, Go math/big, and Zig integer types.

Try It Yourself

All test files are in examples/basic/ - compile and run them to verify.

By The Numbers

Related Features