Real tables arrive untidy. These are the words for that.
messy <- data.frame(
raw = c(" ann marie ", " bob "),
n = c(7L, 99L),
stringsAsFactors = FALSE
)
messy |>
add(name = trim(raw)) |>
add(first = split_text(name, " ", 1),
size = characters(name),
fixed = replace_text(name, "a", "A"))| raw | n | name | first | size | fixed |
|---|---|---|---|---|---|
| ann marie | 7 | ann marie | ann | 9 | Ann mArie |
| bob | 99 | bob | bob | 3 | bob |
messy = pd.DataFrame({
"raw": [" ann marie ", " bob "],
"n": [7, 99],
})
(messy
>> add(name = trim(col.raw))
>> add(first = split_text(col.name, " ", 1),
size = characters(col.name),
fixed = replace_text(col.name, "a", "A")))| raw | n | name | first | size | fixed |
|---|---|---|---|---|---|
| ann marie | 7 | ann marie | ann | 9 | Ann mArie |
| bob | 99 | bob | bob | 3 | bob |
split_text says which piece it wants, counting from 1, because every value in the grammar is one value. Where there is no such piece the answer is empty text rather than missing, so bob gives back bob.
replace_text looks for the text itself, not for a pattern, so nothing in it is special. characters counts characters; it is not called length, because R’s length counts the elements of a vector and a word that reads as one thing and does another is the trap this vocabulary is built to avoid.
lower and upper fold a value’s case. They are ordinary functions on a value, so they go anywhere a value goes.
messy |> add(shout = upper(trim(raw)), quiet = lower(trim(raw))) |> pick(shout, quiet)| shout | quiet |
|---|---|
| ANN MARIE | ann marie |
| BOB | bob |
messy >> add(shout = upper(trim(col.raw)), quiet = lower(trim(col.raw))) >> pick(col.shout, col.quiet)| shout | quiet |
|---|---|
| ANN MARIE | ann marie |
| BOB | bob |
The same two words also work on a column’s name rather than on its contents, which is how you match names whose capitalization you cannot rely on. That is Chapter 22, and it needs no second spelling of either word.