Using Julia to expand a list of dicts fields in dataframe

user_17039578 0
Viewed 33

I have a column in a dataframe that is a list of dictionaries:

[{"key1": value1, "key2": "value2", "key3": "value3", "key4": "value4"}, {"key1": value1, "key2": "value2", "key3": "value3", "key4": "value4"}]

Is there a way to expand this column to get something like this:

key1    key2     key3      key4
value1  value2   value3    value4
value1  value2   value3    value4

Note: key_ can be any string, value_ can be any value.

ijulia-notebook julia
1 Answers

This is quite easy:

julia> df = DataFrame(col=[Dict("key1"=>"value1", "key2"=>"value2", "key3"=>"value3", "key4"=>"value4"),
                           Dict("key1"=>"value1", "key2"=>"value2", "key3"=>"value3", "key4"=>"value4")])
2×1 DataFrame
 Row │ col
     │ Dict…
─────┼───────────────────────────────────
   1 │ Dict("key2"=>"value2", "key3"=>"…
   2 │ Dict("key2"=>"value2", "key3"=>"…

julia> select(df, :col => AsTable)
2×4 DataFrame
 Row │ key2    key3    key1    key4
     │ String  String  String  String
─────┼────────────────────────────────
   1 │ value2  value3  value1  value4
   2 │ value2  value3  value1  value4

The only limitation is that the order of resulting columns is undefined as Dict does not guarantee key order. You would need to re-order them in the second step e.g. like this:

julia> select(select(df, :col => AsTable), string.("key", 1:4))
2×4 DataFrame
 Row │ key1    key2    key3    key4
     │ String  String  String  String
─────┼────────────────────────────────
   1 │ value1  value2  value3  value4
   2 │ value1  value2  value3  value4

Another approach would be:

julia> select(df, :col .=> [ByRow(x -> x["key$i"]) => "key$i" for i in 1:4])
2×4 DataFrame
 Row │ key1    key2    key3    key4
     │ String  String  String  String
─────┼────────────────────────────────
   1 │ value1  value2  value3  value4
   2 │ value1  value2  value3  value4

it is a bit more complex conceptually and requires you to know what keys you want to extract, but the benefit is that you do the operation in one shot.

user_12695670
Related