$string(number) capped at precision-15

[grant-d]

We have some integers (below the safe limit of 2^53 - 1) that are being rounded to precision 15 once we use $string() on them.
This seems due to the following line of code, specifically the positive unary case, ? Number(val.toPrecision(15))

jsonata/src/functions.js

Line 134 in d7790e8

return (typeof val !== 'undefined' && val !== null && val.toPrecision && isNumeric(val)) ? Number(val.toPrecision(15)) :

I can submit a PR but my question is whether there is a good reason for the limit. I would not want to regress something else.

Repro:

$string(1e15 - 1) //  999...999 (15 digits, correct result)
$string(1e15    ) // 1000...000 (16 digits, correct result)
$string(1e15 + 1) // 1000...000 (16 digits, wrong result - notice no trailing 1)

// For completeness, Number.MAX_SAFE_INTEGER = 2^53-1 = 9,007,199,254,740,991 = ~1e16
$string(9007199254740991) // 9007199254740990 (16 digits, wrong result - notice rounding on last digit)

[andrew-coleman]

The precision is limited to avoid the inaccuracy of decimal arithmetic showing up in the string (see #194 (comment)). This even shows up occasionally for toPrecision(16), e.g with the expression Account.Order.$string($sum(Product.(Price* Quantity))) on the exerciser sample data.

To work around this, since you know you are dealing with integers, use the $formatInteger() function instead of $string().

E.g. $formatInteger(9007199254740991, '0') => "9007199254740991"