Type conversion with NA or null instead of throwing errors

[Burtan]

Hi,

when trying to convert values like String to Double and there is an invalid value, an error is thrown. Maybe change this behaviour to convert invalid values to NA or null, or add functions that have different behaviours like toDoubleOrNull() or toDoubleOrNa().

Thanks!

[koperagen]

Hi! You can use with { it.toDoubleOrNull() }. But generally speaking, yes, throwing errors in column-wise operations is not very helpful.
Let me ask a couple of questions to better understand what we could do. What would you do with these toDoubleOrNull() or toDoubleOrNa()? Filter/replace nulls out (so it'll remain in the code like a part of a pipeline) or try to understand why certain values cannot be converted to double (only use for debug, remove later)?

[Burtan]

It would be part of a pipeline. E.g. the values my pcr machine puts out is either a Double or "Undetermined". I would like to treat this as NA or null. So I can check if certain tests that should give a value succeeded or failed.

[Kopilov]

@koperagen but what do you think about using logging inside DataFrame? For example, to print warning on some data loss (such as converting specific values to string or the situation above). Will it be good idea to add slf4j-api or any another dependence?

[koperagen]

@koperagen but what do you think about using logging inside DataFrame? For example, to print warning on some data loss (such as converting specific values to string or the situation above). Will it be good idea to add slf4j-api or any another dependence?

Let's discuss it in #147

[zaleslaw]

@Burtan if it's actual, and you keep the example with the invalid input, could you please share the code/data with us?