[Python] Dictionary has out-of-bound index when creating DictionaryArray from Pandas with NaN

[asfimport]

Minimal reproducer:

import pandas as pd
import pyarrow as pa

pa.array(pd.Categorical(['a', 'b', 'c'], categories=['a', 'b']))

Reporter: Keith Kraus / @kkraus14
Assignee: Wes McKinney / @wesm

PRs and other links:

• GitHub Pull Request #2670

_{Note: This issue was originally created as ARROW-3374. Please see the migration documentation for further details.}

[asfimport]

Wes McKinney / @wesm:
You have to pass from_pandas=True

In [4]: pa.array(pd.Categorical(['a', 'b', 'c'], categories=['a', 'b']), from_pandas=True)
Out[4]: 
<pyarrow.lib.DictionaryArray object at 0x7fd2064810a8>

-- dictionary:
  [
    "a",
    "b"
  ]
-- indices:
  [
    0,
    1,
    null
  ]

[asfimport]

Wes McKinney / @wesm:
Actually, I think in this case we can safely assume the user's intent is to respect pandas's nulls. I will fix

[asfimport]

Wes McKinney / @wesm:
Issue resolved by pull request 2670
#2670