Trailing zeros in Metrics

So I did some digging and this has nothing to do with dashboards but it's to do with the value of your metrics. If you want to have a single decimal point in your numbers you need to use the round() function to truncate the decimals of the value in your datset, a format function is merely doing for display purposes but not behind the scenes.

I couldn't find a way to do this in a prepare recipe, there doesn't seem to be a built-in formula language function that can round a number to specific decimals. The existing round() function only allows to round a number to the nearest integer. So I created a Custom Python Function which does the rounding:

# Modify the process function to fit your needs
import pandas as pd
def process(row):

    number = float(row['Double_Number'])
    number_rounded = round(number, 1)
    
    return number_rounded

This is not going to be very fast as it will run this Python code on each individual row but since you said this is a single row dataset it should be fine.

Another way to do it is to format the number (which returns a string) and then convert it back to a number using this formula:

toNumber(format("%.1f", Double_Number))

This will probably be a lot faster. Alternatively if you are using custom Python metrics probes you can use Python's round() function there. Now will this solve all your number format issues? Well no, turns out if the rounded number ends with trailing zeros Dataiku will remove them. So in my result set I did two custom Python probes and showed the min (12334.6) and the max (4325468.0) of Double_Number_Rounded and this is what you get:

You could argue Dataiku is doing the right thing, a trailing zero is superfluous and has value. However you could also argue that if you want to see numbers all with the same precision Dataiku shouldn't be messing with them. In this case I sit more close to the latter than the former since I tend to prefer flexibility of deciding how I want things rather than programs deciding for me.