Decorate namedtuple's default repr with `@reprlib.recursive_repr`

[bswck]

Feature or enhancement

Proposal:

For

from __future__ import annotations

from typing import NamedTuple


class Biz(NamedTuple):
    our_final_destination: list[Foo]


class Bar(NamedTuple):
    even_longer_attribute_name_presumably: Biz


class Foo(NamedTuple):
    very_long_attribute_name: Bar


foo = Foo(Bar(Biz(Bar(dest := []))))
dest.append(foo)

print(foo)

Before decorating:

Foo(very_long_attribute_name=Bar(even_longer_attribute_name_presumably=Biz(our_final_destination=Bar(even_longer_attribute_name_presumably=[Foo(very_long_attribute_name=Bar(even_longer_attribute_name_presumably=Biz(our_final_destination=Bar(even_longer_attribute_name_presumably=[...]))))]))))

after decorating:

Foo(very_long_attribute_name=Bar(even_longer_attribute_name_presumably=Biz(our_final_destination=Bar(even_longer_attribute_name_presumably=[...]))))

(Note that if our_final_destination pointed to Foo, e.g. via a mutable object that didn't have recursive repr support but included foo in its repr, printing foo would fail with a recursion limit excess. It's not a bug of CPython by any means though.)

The decorator is already available in the module as _recursive_repr , so it's just a matter of application.

Has this already been discussed elsewhere?

This is a minor feature, which does not need previous discussion elsewhere

Links to previous discussion of this feature:

No response

[bswck]

Now that could be backported, but I don't think it's worth the effort or even classifies.

[ericvsmith]

Could you explain what problem this solves?

[rhettinger]

Long ago we decided to apply recursive_repr to OrderedDict and ChainMap but not Counter and namedtuple. The use cases made it unlikely that the latter two would ever need it. Real world deployment for 15+ years has borne out that judgment.

Also there is a performance consideration for named tuples. Its repr is fine-grained with one repr call per tuple. That adds up in the common case where users are displaying lists of named tuples. The situation is different from OrderedDict and ChainMap which are course-grained and the recursive_repr is applied just once to the entire structure rather than for every entry in the structure.

[ericvsmith]

I'm going to close this. If the OP has s a compelling reason to re-think the earlier decision, we can re-open this.