Rename DataFrame.arrange as DataFrame.sort

[billylanchantin]

Renames:

• DataFrame.arrange -> DataFrame.sort_by
• DataFrame.arrange_with -> DataFrame.sort_with

Closes:

• Revist the names of DataFrame.arrange/2 and DataFrame.arrange_with/2 #766

[billylanchantin]

Naming

There's a naming issue I hadn't considered. So far I've named the DataFrame functions to match their Series counterparts:

Type	Series	DataFrame
Function	Series.sort	N/A
Expression	Series.sort_by	DataFrame.sort_by
Callback	Series.sort_with	DataFrame.sort_with

But this is how I think we should name them instead:

Type	Series	DataFrame
Function	Series.sort_values	N/A
Expression	Series.sort	DataFrame.sort
Callback	Series.sort_with	DataFrame.sort_with

That change would be breaking because Series.sort would change meaning. (We discussed this here.)

However, I hadn't considered "five main verbs" angle. As it stands, this is how that section of "Exploring Explorer" reads:

In Explorer, like in dplyr, we have five main verbs to work with dataframes:

• select
• filter
• mutate
• sort_by
• summarise

Now we could just change that bullet to "sort" and be done with it. But each of the other four follows the pattern:

• Explorer.DataFrame.verb (expression)
• Explorer.DataFrame.verb_with (callback)

Since we're pre-1.0, I think it's worth breaking Series.sort to remain consistent.

As a bonus, there'd be an opportunity for a DataFrame analog of Series.sort_values. Maybe something like:

DataFrame.sort_columns(asc: "column_name")

WDYT? /cc @josevalim @philss @cigrainger

[philss]

@billylanchantin I'm more inclined to the version we have in this PR - the sort / sort_by / sort_with. But I'm not against the proposed new names! And I think it's fine to break the API at this point.

[jcmkk3]

I'm just a dataframe geek that likes to keep an eye on this project, but I don't find sort_by to be a problem. There is also group_by, which is another verb that provides context to subsequent operations. I'm glad to see the change from arrange. I'm a big fan of the tidyverse, but don't view some of the names to be optimal. They were restricted by wanting to avoid explicit namespacing and not wanting to conflict with built-in functions.

[billylanchantin]

Thanks @philss! I'm gonna wait for @cigrainger to weigh in too just in case. He was who originally brought up changing the arrange choice.

[cigrainger]

Hey! Sorry for weighing in late. I don't think sort_by is a problem. It's also 100% fine to break stuff at this point in time, so don't let that bound you in. We should focus on getting it right. Thanks @jcmkk3 I agree totally. And I think the _by suffix makes sense in a dataframe.

[cigrainger]

<3

[billylanchantin]

Thanks all! If no one objects to _by, then I don't either.

Also sorry for the snub, @jcmkk3! I never saw your input. Turns out we commented within 60s of each other 😅 I think group_by is a great parallel to draw.