Creating a DataFrame from Dicts can lead to missing columns in the DataFrame or an error depending on the order of the data

[stensmo]

using DataFrames

car1=Dict("id"=>"f3eb12eb-3135-4a0b-a2fd-4ccbb981974e", "name"=>"Fiat","engine"=>Dict("id"=>"0e2e1a0d-c7ea-446a-b4c0-6c1fc441bd73", "cylinders"=>4, "horsepower"=>125))
car2=Dict("id"=>"8ee4e187-5890-4953-98cf-b97bb819befb", "engine"=>Dict("id"=>"da8fc4c5-ba57-40d9-84ba-097f23070e55", "cylinders"=>6, "horsepower"=>170))


cars=[car2;car1]

carsDF_temp=DataFrame(cars)

In this case, the column "name" is missing from the DataFrame.

using DataFrames

car1=Dict("id"=>"f3eb12eb-3135-4a0b-a2fd-4ccbb981974e", "name"=>"Fiat","engine"=>Dict("id"=>"0e2e1a0d-c7ea-446a-b4c0-6c1fc441bd73", "cylinders"=>4, "horsepower"=>125))
car2=Dict("id"=>"8ee4e187-5890-4953-98cf-b97bb819befb", "engine"=>Dict("id"=>"da8fc4c5-ba57-40d9-84ba-097f23070e55", "cylinders"=>6, "horsepower"=>170))


cars=[car1;car2]

carsDF_temp=DataFrame(cars)

This throws an error.

The order of the data is the only difference

[bkamins]

As you can see from the stack trace this issue is not in DataFrames.jl but in Tables.jl. I agree this is unfortunate. I am transferring the issue to Tables.jl to decide there what should be the behavior.