Skip to main content

Trunc

function Trunc(E: Extended): LongInt;

Example​

procedure ScriptEvent(var Value: variant);
begin
  Value := Trunc(-12.75); // -12
end;

Usage​

Trunc discards an Extended value's fraction toward zero and narrows the result to a 32-bit LongInt.

Parameters

NameTypeDescription
EExtendedFloating value whose fractional component is discarded toward zero.

Returns

For a truncated value within Low(LongInt)..High(LongInt), the expected 32-bit integer. Otherwise, the low 32 bits of the Velox Int64 result are reinterpreted as signed LongInt.

Velox Int64 resultVelox LongInt result
2147483648-2147483648
4294967295-1
42949672960
-21474836492147483647

Additional Technical Info​

Trunc discards the fractional part of an Extended value by rounding toward zero, obtains Delphi's Int64 result, then stores only a PascalScript 32-bit LongInt.

The unchecked 32-bit narrowing is a critical difference from Delphi's public Int64 contract. Values outside the LongInt range silently wrap modulo 2^32 if they are still convertible to Int64. The example is fictional and source-reviewed only.

Implementation

  1. The modified PascalScript compiler declares Trunc in the Numeric group.
  2. Runtime DefProc selector 19 reads the argument as real.
  3. Delphi System.Trunc produces an Int64 by rounding toward zero.
  4. Stack.SetInt accepts a LongInt; PascalScript.inc has disabled range/overflow checks, so narrowing wraps.

Edge cases and errors

  • Trunc(-12.75) is -12, not floor -13.
  • NaN, infinity or a value outside Delphi's Int64 conversion range can raise EInvalidOp/a PascalScript math error before narrowing.
  • Extended is Double on default Win64 and wider on Win32; near-boundary binary representation can differ.
  • This is not a checked numeric conversion. Validate against the 32-bit LongInt range before calling if wraparound is not acceptable.

Related entries

  • Int also removes the fraction toward zero but returns floating Extended and does not narrow to LongInt.
  • Frac returns the complementary signed fractional part.
  • Round rounds to nearest before applying the same unchecked 32-bit narrowing.

External references

Created 2026-07-15