DataFrame.groupby().groups should truncate output

[pikeas]

The output from .groups on a grouped DataFrame should be handled similarly to a large DataFrame, where beyond a certain length, the data is displayed in a summary form rather than output in its entirety.

This truncation should occur both at the group level (large number of groups) and per-group (large amount of data points in a group).

[cpcloud]

bump. What about the groups property being a DataFrame with nans filing unequal group sizes? I don't know about others, but I frequently find myself converting groups to a DataFrame.

[kokes]

Truncation on the per-group basis already works now, because indices in pandas have their own __repr__.

pd.DataFrame({'a': [random.randint(0, 100) for j in range(10000)]}).groupby('a').groups

As for the group-level truncation, that's a bit tricky, because the .groups property is a plain dictionary. What could be done is that we could create a class that inherits from dict and adds a custom __repr__ and nothing else. Something along these lines:

class DataFrameGroups(dict):
	def __repr__(self):
		if len(self) < 30:
			return super().__repr__()

		for k, v in self.items():
			pass # build representation here

[jreback]

closing as duplicate