Resources Numerical accuracy Calculation policy
The script functions that convert values to single precision or to whole-number types.
The language keeps every number as a 64-bit value. single() rounds each value to the nearest 32-bit number with Math.fround and stores it back in a Float64Array, so later arithmetic is 64-bit on values that carry about 7 digits. The whole-number functions round but keep 64-bit storage and do not limit the range.
Open item. State in the help for single() that results carry about 7 digits, and make the whole-number functions round halves the same way round() does and limit values to each type's range.
2 calculations in 1 source file. The code is read from the released app when this page loads. It is the calculation people run, not a copy kept for this page.
Tests tools/partcraft-test/test-craftscript-b.mjs
single(A): each value rounded to the nearest 32-bit floating-point number.
x ← fround(x)
This is what the function is for: the result is a 32-bit value stored in 64-bit form. Anything computed from it afterwards starts from about 7 significant digits.
def('single', (a) => { const t = T(a[0], 'single'); return out(t.shape, Float64Array.from(t.data, Math.fround)); }, 'single(A)');int8, int16, int32, int64 and the unsigned forms: each value rounded to a whole number.
x ← round(x); NaN → 0
All eight names do the same thing: Math.round with no range limit and no sign check for the unsigned names. Halves round toward +∞. Values above 2⁵³ are already whole and pass through.
for (const name of ['int32', 'int64', 'int16', 'int8', 'uint8', 'uint16', 'uint32', 'uint64']) def(name, (a) => K.map(T(a[0], name), (x) => (x !== x ? 0 : Math.round(x))), `${name}(A)`);One step stores or exchanges 32-bit values that reach a number a person reads. It is listed here with its open item, and the step is either moved to 64-bit or the app will warn where the number is read.