Datasets manipulations (tranformations/change) API.

[AndreiKingsley]

Problem №1

We have a lot of data transformations methods inside the plot builder that modify dataset (such as groupBy(){ },statBin(){ }):

plot {
   line { ... }
   groupBy(..) { 
      line { ... }
   }
   statBin(..) {
      line { ... }
   }
}

There will also be others in the future (for example sortBy() { }).

The problem is that they are not isolated in any way, they “break” the context of plot builder, being on the same level as creating layers.

Problem №2

We have no API for dataset changing. Sometimes, we want to use different datasets for different layers:

dataset1.plot {
   line { 
       x("column_of_dataset_1")
    }
   `function that change dataset`(dataset2) {
       line {
          x("column_of_dataset_2")
       }
    }
}

Possible solution

Both of these problems are related to the design of the dataset manipulation inside plot builder. A possible solution is to put all such manipulations under one object inside the plot builder. It could go something like this:

df.plot {
    line {}
    Data.use(otherDF) {
        line {}
    }
    Data.groupBy() {
        line {}
    }
    Data.statBin() {
        line {}
    }
    
    Data.groupBy() {
        Data.statBin() {
            bars {}
        }
    }
    
}

[zaleslaw]

For me it's a first time when I see an object as a namespace in KotlinDSL? Did you see any examples in Kotlin ecosystem in that approach?

[AndreiKingsley]

It's not really an object, it's a property. In fact it's just like enum.