Feature-like bug in File reader: wrong autodetection of missing value pattern

[Experimenter]

I have a CSV file:
‘country’; ‘cus_code’
‘UK’;‘K118’
‘IT’;‘K118’
‘DE’;‘44558’
‘DE’;‘44556’
‘DE’;‘11992’

Using file reader node to read it:

What! Missing values? Maybe changing format to string helps?

Nope. Turns out, File reader wrongly autodetected the missing value string:

I guess this was intended as feature, however brings a little, unexpected bug. It appears that if there is one non-number value in number column, it is assumed as missing value pattern.

[izaychik63]

I can assume that integer field cannot have K118 filler for missing. It is string type

[Experimenter]

Well, it is string in the CSV, enclosed by quotes. Anyhow, KNIME should not be guessing “K118” as missing value pattern, it does not make sense in any case.

[izaychik63]

I would recommend to use File Reader node instead CSV Reader. It is more robust and allow you to assign a field type using header in preview.

[ipazin]

Hello @Experimenter,

I have checked it and feels wrong to do such auto detection. Have opened up a ticket and let’s see. (Internal reference: AP-15491)

However to avoid this you can use new CSV Reader (Labs) or (also new) Simple File Reader node.

Br,
Ivan

[Experimenter]

Thank you. Yeah, I checked that Simple reader handles the matter correctly. Just lately I am using drag-n-dropping file from Windows Explorer to the Knime workflow, and File reader is the default option offered.