Danferno commented on issue #47022:
URL: https://github.com/apache/arrow/issues/47022#issuecomment-3637584537
The only solution I've found for this so far is catching any
pa.lib.ArrowTypeError and then using this manual parquet -> pyarrow -> pandas
code that recasts the field type. It might break upon negative values in the
indices?
```
def readParquetFileWithUnsignedDictionaryKeys(path, fs=None):
import pyarrow as pa
table = pa.parquet.read_table(path, filesystem=fs)
for field, column in zip(table.schema, table.columns): # fiter =
iter(zip(table.schema, table.columns)); field, column = next(fiter)
if pa.types.is_dictionary(field.type):
# Convert unsigned types to their signed variant
index_type = field.type.index_type
if pa.types.is_uint8(index_type):
new_index_type = pa.int8()
elif pa.types.is_uint16(index_type):
new_index_type = pa.int16()
elif pa.types.is_uint32(index_type):
new_index_type = pa.int32()
elif pa.types.is_uint64(index_type):
new_index_type = pa.int64()
else:
new_index_type = index_type
new_type = pa.dictionary(new_index_type, field.type.value_type)
table =
table.set_column(table.schema.get_field_index(field.name), field.name,
column.cast(new_type))
return table.to_pandas()
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
