Make head(), tail() and limit() return the same number of rows for Expr, Series, DataFrame and LazyFrame

[Wainberg]

Description

Expr.head() and Series.head() return 10 rows by default, while DataFrame.head() and LazyFrame.head() return 5 rows. We should pick one ;)

We should do the same for tail() and limit().

[mcrumiller]

I like 5. The point is to see what the table looks like (columns, dtypes, reasonable samples of data). and adding more rows than that might make you scroll up in a smallish terminal.

[stinodego]

I don't think we need to pick one at all. One is a single column while the other is a 2D table. 10 rows of a dataframe is a lot of information, 10 elements of a column not so much.

There is also something to be said for having the same value across data types, but it's not so black and white as you make it out to be.

[Wainberg]

I would argue for either using 5 for all, or changing the default number of printed rows. Currently you can't see the intermediate rows with Expr.head():

>>> pl.DataFrame(range(100)).select(pl.all().head())
shape: (10, 1)
┌──────────┐
│ column_0 │
│ ---      │
│ i64      │
╞══════════╡
│ 0        │
│ 1        │
│ 2        │
│ 3        │
│ …        │
│ 6        │
│ 7        │
│ 8        │
│ 9        │
└──────────┘

[Wainberg]

Would you be open to changing the default number of printed rows from 8 to 10?

[alexander-beedie]

Would you be open to changing the default number of printed rows from 8 to 10?

Apparently so 😎 #13699

[Wainberg]

Love it!!

Ok so at this point I would definitely advocate for head(), tail() and limit() to return 10 rows instead of 5. I'm sympathetic to Stijn's argument that 10 rows of a DataFrame can be more of an 'information overload' than 5 rows, but if we're now printing 10 rows of a DataFrame by default, may as well do the same for head() etc.

[alexander-beedie]

if we're now printing 10 rows of a DataFrame by default, may as well do the same for head() etc.

Counter-argument: in those 10 rows we're displaying the equivalent of head(5) and tail(5) once the total number of rows >= 10, so it can also be argued that it is consistent to keep that correspondence (I'm relatively neutral, though lean towards keeping frame head/tail at 5 for that reason as well as @stinodego's 😉).

[mcrumiller]

@alexander-beedie very good point. No matter what n we pick for head and tail, we must show 2n+1 or 2n-1 rows in df.__str__() without an asymmetric head/tail (if we want to show the same information).

[mcrumiller]

My vote here is:

• 10 rows for head and tail
• 10 rows for limit
• first 5 and last 5 for __str__