gnd

The float32 operation converts one input value to a 32-bit IEEE-754 floating-point number.

Floating-point literal - decimal (3.14, 1e-5).
Integer literal - decimal (42) or hexadecimal (0x2A).
Numeric value - already stored in a variable (integer or float).
String literal - text that parses as a decimal or hexadecimal floating-point or integer.

The representable range is roughly +/-1.18 x 10^- ^3^8 … +/-3.40 x 10^3^8.
Values outside that range raise an overflow error.
Values whose magnitude is non-zero but smaller than the minimum normalised float32 are flushed to +/-0.
Conversion rounds to the nearest representable float32 using IEEE-754 “round to nearest, ties to even”.

[ $destination ] float32 value

$destination is optional; if omitted, the converted value is assigned to _.
Exactly one operand must follow float32. Writing float32 with no operand is shorthand for float32 _.

Convert an integer literal:

$ratio float32  10          # 10.0 stored as float32

Convert a decimal string:

$txt    let "2.71828"
$e32    float32 $txt        # rounded to nearest float32

Convert the current _ in place:

let  1e40
float32                      # error: overflow (value too large)

Handle small values:

float32 1e-50                # silently becomes 0 (underflow to sub-normal 0)

Reject non-numeric input:

float32 "hello"              # error: cannot parse as number

float32 always produces one float32 value, never rebinds existing variables, and observes Gendo’s single-assignment rule.