BooleanTransformer.reverse_transform sometimes crashes with TypeError

[katxiao]

Environment Details

Please indicate the following details about the environment in which you found the bug:

• RDT version: 0.5.2.dev0
• Python version: 3.8
• Operating System: macOS 10.15.7

Error Description

When sampling data from a boolean column, we sometimes receive the error TypeError: Need to pass bool-like values from the BooleanTransformer.

This happens when the BooleanTransformer.reverse_transform method receives inputs that round to values outside of [0, 1] (the valid boolean float representations). When we cast those values to 'boolean', it throws this type error. We had previously been rounding to type 'bool', which would cast these out-of-bound floats to True. Now that we are rounding to 'boolean', we should clip the rounded values to [0, 1].

Steps to reproduce

We can reproduce the error with the following code snippet:

from sdv.demo import sample_relational_demo
from sdv.relational import HMA1

metadata, tables = sample_relational_demo(size=30)

model = HMA1(metadata)
model.fit(tables)

new_data = model.sample()

The above code sometimes crashes with:

  File "/SDV/sdv/relational/base.py", line 184, in sample
    return self._sample(table_name, num_rows, sample_children)
  File "/SDV/sdv/relational/hma.py", line 471, in _sample
    self._sample_table(table, num_rows, sampled_data=sampled_data)
  File "/SDV/sdv/relational/hma.py", line 430, in _sample_table
    self._sample_children(table_name, sampled_data, table_rows)
  File "/SDV/sdv/relational/hma.py", line 355, in _sample_children
    self._sample_children(child_name, sampled_data, child_rows)
  File "/SDV/sdv/relational/hma.py", line 352, in _sample_children
    self._sample_child_rows(child_name, table_name, row, sampled_data)
  File "/SDV/sdv/relational/hma.py", line 335, in _sample_child_rows
    table_rows = self._sample_rows(model, table_name)
  File "/SDV/sdv/relational/hma.py", line 318, in _sample_rows
    sampled = model.sample(num_rows)
  File "/SDV/sdv/tabular/base.py", line 442, in sample
    return self._sample_batch(num_rows, max_retries, max_rows_multiplier)
  File "/SDV/sdv/tabular/base.py", line 299, in _sample_batch
    sampled, num_valid = self._sample_rows(
  File "/SDV/sdv/tabular/base.py", line 235, in _sample_rows
    sampled = self._metadata.reverse_transform(sampled)
  File "/SDV/sdv/metadata/table.py", line 640, in reverse_transform
    reversed_data = self._hyper_transformer.reverse_transform(data)
  File "/lib/python3.8/site-packages/rdt/hyper_transformer.py", line 248, in reverse_transform
    reversed_data = transformer.reverse_transform(columns_data)
  File "/lib/python3.8/site-packages/rdt/transformers/boolean.py", line 88, in reverse_transform
    return np.round(data).astype('boolean').astype('object')
  File "/lib/python3.8/site-packages/pandas/core/generic.py", line 5546, in astype
    new_data = self._mgr.astype(dtype=dtype, copy=copy, errors=errors,)
  File "/lib/python3.8/site-packages/pandas/core/internals/managers.py", line 595, in astype
    return self.apply("astype", dtype=dtype, copy=copy, errors=errors)
  File "/lib/python3.8/site-packages/pandas/core/internals/managers.py", line 406, in apply
    applied = getattr(b, f)(**kwargs)
  File "/lib/python3.8/site-packages/pandas/core/internals/blocks.py", line 595, in astype
    values = astype_nansafe(vals1d, dtype, copy=True)
  File "/lib/python3.8/site-packages/pandas/core/dtypes/cast.py", line 919, in astype_nansafe
    return dtype.construct_array_type()._from_sequence(arr, dtype=dtype, copy=copy)
  File "/lib/python3.8/site-packages/pandas/core/arrays/boolean.py", line 279, in _from_sequence
    values, mask = coerce_to_array(scalars, copy=copy)
  File "/lib/python3.8/site-packages/pandas/core/arrays/boolean.py", line 158, in coerce_to_array
    raise TypeError("Need to pass bool-like values")
TypeError: Need to pass bool-like values