Bug in File Reader node

KNIME Analytics Platform
1

I have i file which contains a column with 19 digits identification number (only numbers). KNIME File reader node identifies it as Double. And that would be OK, but the number is not the same as in my source. For example (real cases):

Source file:

ID
1000000000082440469
2000000000203430482
1000000000205390188

But in KNIME then i get:

ID

1000000000082440450
2000000000203430400
1000000000205390210
 

As you see the last 2-3 digits are complete miss. But if i manually change the ID format to String, i get everything correctly.

So, if KNIME is not able to deaql with Long numbers, why, while importing, it can not automaticallly choose format String if it founds Long numbers. And then, what is the maximum Double the KNIME can deal with?

Thanks

2

Hi Audrius,

Just to answer your last question: It really depends on the numbers. For numbers with large gaps, double can store values up to ~3*10308. To represent integers, you can use 52 bits + 1 sign, so the largest representable integer (without loss of precision) is: 4,503,599,627,370,495. Your numbers are unfortunately are above of this

Cheers, gabor

3

But then, why File Reader node, while scanning file could not assign String to such field. If the lengh is more than 52 bits then let it be String. And the String would be OK, as it is just ID field.