feat: added stringify_unsupported()

[SiddhantSadangi]

Before submitting checklist

• Did you ask the docs owner to review all the user-facing changes?

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[Copilot]

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

Comments suppressed due to low confidence (2)

src/neptune_scale/utils.py:5

• [nitpick] The '**kwargs' parameter is defined but not used within the function. Consider removing it to simplify the function signature unless it's reserved for future extensions.

def stringify_unsupported(d: dict[str, Any], **kwargs: Any) -> dict[str, Any]:

src/neptune_scale/utils.py:44

• [nitpick] Consider simplifying new_key assignment for clarity. For example, using: new_key = key if not prefix else f"{prefix}/{key}".

new_key = f"{prefix}/{key}" if prefix else f"{prefix}{key}"

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[Copilot]

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

[Copilot]

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

[Copilot]

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

[pitercl]

What about types that are supported in some methods, but not others? E.g. File is supported in assign_files but not in log_configs. Or Histogram - supported as an element of a series, but not as an atom.

TBH I'm not entirely sure what the purpose of this function is.

[SiddhantSadangi]

This function is meant only to convert anything that can be represented to a primitive type, much like the original stringify_unsupported

[pitercl]

Ok, what worries me here:

1. This function does 2 unrelated things: flatten + stringify
2. It's called stringify_unsupported yet it stringifies also types that Neptune supports. What does it even mean that Neptune doesn't support a type?

How would you feel about splitting this into 2 methods:

# just flattening dictionaries, no change to leaf values
# and no changes to the original value
def flatten(d: dict[str: Any]) -> dict[str: Any]
   pass

PRESERVED_TYPES = [int, float, str, datetime, bool]

# Assume it's a flat dictionary.
# Any value not in the `preserved_types` list gets strigified
def stringify_values(d: dict[str, Any], preserved_types: list[Type] = PRESERVED_TYPES) -> dict[str, Any]:
   pass

[SiddhantSadangi]

What does it mean that Neptune doesn't support a certain type?

How would you feel about splitting this into 2 methods:

I can divide it into private methods internally, but from a user's perspective, the problem remains the same. You can't pass nested dictionaries to log_configs(), so these inputs will need to be:

1. Flattened
2. Any unsupported types (like tensor or np.array()) will need to be converted to a supported type.

The function stringify_unsupported combines both of these operations into a single function, which reduces the boilerplate code required by the user.

I agree that the naming could be improved. I copied the name from neptune<2.x.

@normandy7 , do you have any suggestions for a better name?

[normandy7]

convert_to_flat_stringified_dict? Verbose, but wouldn't leave anyone guessing 😅

[normandy7]

Might as well be flatten_and_stringify_unsupported or something like that, though. I don't think there's an obvious way to shorten it further without obfuscating it.

[SiddhantSadangi]

flatten_and_cast?

[normandy7]

Isn't cast ambiguous, though? I'd rather use stringify.

[SiddhantSadangi]

No strong opinions here... Will leave it to @normandy7 and @pitercl to finalize the name :)

[Copilot]

Pull Request Overview

This PR introduces the new function stringify_unsupported(), which flattens nested dictionaries and converts unsupported values to strings for logging purposes, along with comprehensive unit tests verifying its behavior.

• Added stringify_unsupported() implementation in the utils module
• Introduced multiple unit tests covering basic, collection-based, nested, mixed, edge cases, custom objects, and invalid/empty inputs

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
tests/unit/test_stringify_unsupported.py	Added extensive tests covering various input types and edge cases for the new helper.
src/neptune_scale/utils.py	Added the stringify_unsupported() function implementation with recursive flattening logic.