Function `to_long()` returns unexpected value when input is float with decimals

[abitmore]

As mentioned in bitshares/bitshares-ui#3545 (comment), the function to_long() in SerializerValidation.js may return unexpected result if the input is a float with decimals (E.G. 12.999999999999998). Code:

bitsharesjs/lib/serializer/src/SerializerValidation.js

Lines 103 to 116 in c962b80

	to_long(value, field_name = "", unsigned = false) {
	if (this.is_empty(value)) {
	return value;
	}
	if (Long.isLong(value)) {
	return value;
	}
	this.no_overflow64(value, field_name, unsigned);
	if (typeof value === "number") {
	value = "" + value;
	}
	return Long.fromString(value, unsigned);
	},

> parseFloat("0.00013")*100000
12.999999999999998
> Long.fromString(""+parseFloat("0.00013")*100000)
Long { low: 1150981118, high: 30, unsigned: false }
> Long.fromString(""+parseFloat("0.00013")*100000).toString()
129999999998

To fix this, IMHO, as a library, we shouldn't change user input, but we should require the input to be valid.

[christophersanborn]

For a little extra context, it looks like Long (from the bytebuffer library), will correctly handle the first five digits after the decimal point, whereafter it seems to just append digits.

> var Long = require("bytebuffer").Long;
undefined
> Long.fromString("12.12345")
Long { low: 12, high: 0, unsigned: false }
> Long.fromString("12.123456")
Long { low: 126, high: 0, unsigned: false }
> Long.fromString("12.1234567")
Long { low: 1267, high: 0, unsigned: false }
>