Adds the parent name in the flatten operation #378
Conversation
|
Don't forget to regenerate docs, any build/assemble call will do so. Maybe it's a good idea to make this behavior optional? I can see many cases where flattening should result in all child columns with their actual name, not including the parent name. |
|
Could you please share the case, I will add it to the tests and could make ot optional |
|
Just take any Dataframe containing some column groups where the children (value- or frame-columns) names are unique and flatten it. This should produce a new data frame with all value- and frame-columns regardless of whether they were nested in a column group or not |
|
In SQL, it's done by default, you know all this stuff: "MAX (name)" and so on. |
|
@zaleslaw "keepParentNameForColumns" is a bit long isn't it? Also, "forColumns" is a bit redundant IMO. I'd consider something like "renameByPath". |
| @@ -29,7 +30,8 @@ internal fun <T, C> DataFrame<T>.flattenImpl( | |||
| .into { | |||
| val targetPath = getRootPrefix(it.path).dropLast(1) | |||
| val nameGen = nameGenerators[targetPath]!! | |||
| val name = nameGen.addUnique(it.name()) | |||
| val preferredName = if (keepParentNameForColumns) "${it.name()}.${it.parentName}" else it.name() | |||
There was a problem hiding this comment.
Looks like this is the wrong way around, should be parentName.name instead.
Fixes #229