DynamoDB attributevalue.Integer() loses precision for large values

[BastiPaeltz]

When dealing with large integer values (timestamps in nanoseconds) that you are able to store in DynamoDB, attributevalue.Integer() will lose precision when converting to int64. An example value would be 1556893286484340854.
This is caused by the strconv.ParseFloat(value, 64) LOC that this function uses before converting to int64. When doing a strconv.ParseInt(value, 10, 64) instead, the value does not lose precision.

[diegobernardes]

This code can be used to confirm this behaviour:

package main

import (
	"fmt"
	"strconv"
	"time"
)

func main() {
	base := time.Now().UnixNano()
	baseString := strconv.FormatInt(base, 10)

	floatParse, _ := strconv.ParseFloat(baseString, 64)
	floatResult := int64(floatParse)
	intResult, _ := strconv.ParseInt(baseString, 10, 64)

	fmt.Println("origin:             ", base)
	fmt.Println("current behaviour:  ", floatResult)
	fmt.Println("proposed behaviour: ", intResult)
}

Execution result:

origin:              1570726625494837000
current behaviour:   1570726625494836992
proposed behaviour:  1570726625494837000

[diegobernardes]

Solved at #227.