head() and tail() to Peek at the Triangle

[kennethshsu]

Sometimes the triangles are quite large, would it be helpful to have head() and tail() to peek at the top or bottom origin exposures?

Obviously this would only work for a 1D triangle. Not sure how to handle the columns (development) though.

[jbogaardt]

We do have a head and tail that works on the index, identical to pandas. Are you wanting one for the origin axis or more generally supporting an axis arg?

[kennethshsu]

I mean something like this:

cl.load_sample("raa").head()
cl.load_sample("raa").tail()

This isn't a good example, but if we have a lot of origin years, the triangle display will quickly clog on the whole screen.

[jbogaardt]

If you want modify dataframe displays globally, it will cascade to triangles:

import pandas as pd
pd.set_option("display.max_rows", 10)
pd.set_option("display.max_columns", 10)

import chainladder as cl
tri = cl.load_sample('prism')['Paid']
tri.sum()

As for explicit functionality, there is only a head/tail for the index axis:

tri.head(5) # returns head of index only
tri.tail(5) # returns head of index only

We do not currently have a named methods for head/tail for origin/development axes. I usually just iloc for this:

tri.sum().iloc[..., -5:, :] # origin tail
tri.sum().iloc[..., -5:] # development tail

If none of these feels great, then we could add an arg to head/tail to support an axis. Something like this:

# This functionality currently doesn't exist
tri.sum().head(5, axis=2) # origin head
tri.sum().head(5, axis=3) # development head

On this last one, we can even make the default axis be the first axis that has a length > 1, (similar to how sum works when axis is omitted).