· python pyarrow til

pyarrow: pyarrow.lib.ArrowNotImplementedError: Filter argument must be boolean type

I wanted to filter a table in pyarrow table recently and ran into troubles when trying to use the filter syntax that I’m used to from DuckDB. In this blog post I’ll explain my mistake and how to fix it.

First, let’s install pyarrow:

pip install pyarrow

And now we’re going to create a table that has a few countries and their corresponding continents:

import pyarrow as pa

countries = pa.Table.from_arrays(
      pa.array(['India', 'Pakistan', 'Belgium', 'Finland'], pa.string()),
      pa.array(['Asia', 'Asia', 'Europe', 'Europe'], pa.string())
    names=['Country', 'Continent']

Let’s say we want to find just the rows where the continent is Europe. I initially tried to do that using the following syntax:

countries.filter("Continent = 'Europe'")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "pyarrow/table.pxi", line 3154, in pyarrow.lib.Table.filter
  File "/Users/markhneedham/Library/Caches/pypoetry/virtualenvs/ch07-YVG4Qrie-py3.11/lib/python3.11/site-packages/pyarrow/compute.py", line 259, in wrapper
    return func.call(args, options, memory_pool)
  File "pyarrow/_compute.pyx", line 367, in pyarrow._compute.Function.call
  File "pyarrow/error.pxi", line 144, in pyarrow.lib.pyarrow_internal_check_status
  File "pyarrow/error.pxi", line 121, in pyarrow.lib.check_status
pyarrow.lib.ArrowNotImplementedError: Filter argument must be boolean type

Hmmm, that didn’t work so well. Instead we need to construct a filter predicate using some functions from the pyarrow.compute module, so let’s import that:

import pyarrow.compute as pc

And now we have (at least) two ways to write the filtering statement. We could use pc.equal like this:

countries.filter(pc.equal(countries["Continent"], "Europe"))

Or pc.field like this:

countries.filter(pc.field("Continent") == "Europe")

Either way we get the same result:

Country: string
Continent: string
Country: [["Belgium","Finland"]]
Continent: [["Europe","Europe"]]
  • LinkedIn
  • Tumblr
  • Reddit
  • Google+
  • Pinterest
  • Pocket