How to partition by quantity with write_to_dataset

[phpsxg]

Describe the usage question you have. Please include as many useful details as possible.

write_to_dataset partition by field, error is reported: pyarrow.lib.ArrowInvalid: Fragment would be written into 32768 partitions. This exceeds the maximum of 1024, is there a better way to partition ? Or can it be partitioned by quantity? Thanks

pq.write_to_dataset(table,
  root_path=file_path,
  partition_cols=cls.partition_cols,
  existing_data_behavior='delete_matching',
  use_legacy_dataset=False
)

• python=3.10
• pyarrow=10.0.0

Component

Python

[assignUser]

Without knowing what your data looks like it is hard to say how to better partition it. But you can raise the maximum number of partitions via max_partitions which defaults to 1024 or you could just limit the number of rows per file if you don't have a more specific partitioning scheme with max_rows_per_file:
https://arrow.apache.org/docs/python/generated/pyarrow.dataset.write_dataset.html#pyarrow.dataset.write_dataset

Feel free to close the issue if this answers your question.

[westonpace]

We typically operate in batches of size 32768. So if you are getting that message it means you are trying to partition on a column that has a different value for every single row! As @assignUser pointed out, you can increase max_partitions to support that but this is probably a sign that you are not picking a good partition column. What is the data type of the column you are trying to partition on?

[phpsxg]

Without knowing what your data looks like it is hard to say how to better partition it. But you can raise the maximum number of partitions via max_partitions which defaults to 1024 or you could just limit the number of rows per file if you don't have a more specific partitioning scheme with max_rows_per_file: https://arrow.apache.org/docs/python/generated/pyarrow.dataset.write_dataset.html#pyarrow.dataset.write_dataset

Feel free to close the issue if this answers your question.

Thank you, ds.write_dataset meets my needs, but I have two more questions

1. Is there any difference between pq.write_to_dataset and ds.write_dataset?
2. How to implement dynamic filtering with ds.write_dataset, if the filters I get according to different parameters are a list

For example, there are two filters, which is fine
df = dataset.to_table(filter=( ds.field('c').is_null()) & (ds.field('b') > 0)).to_pandas()

But if the filter is a list, how to pass the parameters

filters = []
if ...:
    filters.append(ds.field('c').is_null())
if ....:
    filters.append(ds.field('b') > 0)

[github-actions]

This issue has been marked as stale because it has had no activity in the past 365 days. Please remove the stale label or comment below, or this issue will be closed in 14 days. If this usage question has evolved into a feature request or docs update, please remove the 'Type: usage' label and add the 'Type: enhancement' label instead.