4.2 Index and Summarize

Let’s go back to the example grades_2020() data defined before:

grades_2020()
name grade_2020
Sally 1.0
Bob 5.0
Alice 8.5
Hank 4.0

To retrieve a vector for name, we can access the DataFrame with the ., as we did previously with structs in Section 3:

function names_grades1()
    df = grades_2020()
    df.name
end
JDS.names_grades1()
["Sally", "Bob", "Alice", "Hank"]

or we can index a DataFrame much like an Array with symbols and special characters. The second index is the column indexing:

function names_grades2()
    df = grades_2020()
    df[!, :name]
end
JDS.names_grades2()
["Sally", "Bob", "Alice", "Hank"]

Note that df.name is exactly the same as df[!, :name], which you can verify yourself by doing:

julia> df = DataFrame(id=[1]);

julia> @edit df.name

In both cases, it gives you the column :name. There also exists df[:, :name] which copies the column :name. In most cases, df[!, :name] is the best bet since it is more versatile and does an in-place modification.

For any row, say the second row, we can use the first index as row indexing:

df = grades_2020()
df[2, :]
name grade_2020
Bob 5.0

or create a function to give us any row i we want:

function grade_2020(i::Int)
    df = grades_2020()
    df[i, :]
end
JDS.grade_2020(2)
name grade_2020
Bob 5.0

We can also get only names for the first 2 rows using slicing (again similar to an Array):

grades_indexing(df) = df[1:2, :name]
JDS.grades_indexing(grades_2020())
["Sally", "Bob"]

If we assume that all names in the table are unique, we can also write a function to obtain the grade for a person via their name. To do so, we convert the table back to one of Julia’s basic data structures (see Section 3.3) which is capable of creating mappings, namely Dicts:

function grade_2020(name::String)
    df = grades_2020()
    dic = Dict(zip(df.name, df.grade_2020))
    dic[name]
end
grade_2020("Bob")

5.0

which works because zip loops through df.name and df.grade_2020 at the same time like a “zipper”:

df = grades_2020()
collect(zip(df.name, df.grade_2020))
("Sally", 1.0)
("Bob", 5.0)
("Alice", 8.5)
("Hank", 4.0)

However, converting a DataFrame to a Dict is only useful when the elements are unique. Generally that is not the case and that’s why we need to learn how to filter a DataFrame.



Support this project
CC BY-NC-SA 4.0 Jose Storopoli, Rik Huijzer, Lazaro Alonso