Normalize lists argument in MerlinDataLoader._augment_schema

[shaun0927]

Goals ⚽

Stop MerlinDataLoader._augment_schema from raising TypeError when the caller omits lists= (the common case for datasets without list-typed features).

Fixes #812.

Implementation Details 🚧

transformers4rec/torch/utils/data_utils.py normalizes cats, conts, and labels to empty lists when they default to None, but the normalization for lists was missing:

cats = cats or []
conts = conts or []
labels = labels or []
# missing: lists = lists or []

schema = schema.select_by_name(conts + cats + labels + lists)

Any caller that does not pass lists= (or passes lists=None explicitly) hits TypeError: can only concatenate list (not "NoneType") to list at the select_by_name line. The later for col in lists: loop would also raise TypeError: 'NoneType' object is not iterable.

Minimal fix — add the symmetric lists = lists or [] alongside the other three:

         cats = cats or []
         conts = conts or []
         labels = labels or []
+        lists = lists or []

         schema = schema.select_by_name(conts + cats + labels + lists)

Testing Details 🔍

Pre-fix repro (extracted from the source, no install needed):

class Schema:
    def select_by_name(self, cols):
        return self

def _augment_schema(schema, cats=None, conts=None, labels=None, lists=None):
    cats = cats or []
    conts = conts or []
    labels = labels or []
    return schema.select_by_name(conts + cats + labels + lists)

_augment_schema(Schema(), cats=["x"])
# TypeError: can only concatenate list (not "NoneType") to list

With the patch, the same call succeeds.

No unit test is added — _augment_schema is currently exercised indirectly by every MerlinDataLoader construction in the test suite that supplies lists=[...]. I am happy to add a direct test for lists=None if reviewers would prefer.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[shaun0927]

FYI — I have read CLA.md and agree to its terms for this submission. The changes in this PR are entirely my original work, made on my own behalf (not in the course of employment by any other party), and are offered under the Apache 2.0 license of the project. Happy to re-state this in any additional form the maintainers prefer.