Test failure appears to be from HEAD, not this change.

Can you please add a test case? thanks

I am not really convinced about adding this as it would mean that we would dump something that couldn't be parsed back into the object that we had before.

I share @xabbuh's concern. This PR makes the assumption that the string representation of an object is meant for serialization which is not necessarily the case.

How about putting this behind an option flag as with other similar dumping alternatives?

An option flag might work, WDYT @xabbuh?

I am still not convinced that supporting to dump something that cannot be parsed back into the initial structure is a good idea. But in any case making this an opt-in feature would be a must-have.

Closing as 2 core team members expressed valid concerns about this PR.