🐛 Potentially missing extra in `ibis` dependency

[alexkolo]

I had to install the following packages to make icanexplai work for me, which were required by the ibis module

- pyarrow~=18.0.0
- pyarrow_hotfix~=0.6

I encounter this in Win10 with Python 3.12.3 and in Linux (WSL) with Python 3.11.10.
I can't say whether this is an individual problem or a general one. But I thought I should let you know in case you want to check.

Using icanexplain==0.3.0.

Possible fix

Change this Line:

https://github.com/carbonfact/icanexplain/blob/2b5b1730d5aa16c07891530f5619e2fe2ea9062e/pyproject.toml#L10

to this:

ibis-framework = {extras = ["duckdb"], version = "^9.5.0"}

According to this file, it would install the missing modules.

How to reproduce

• pip install icanexplai==0.3.0

import pandas as pd
import icanexplain as ice
data: list[dict[str, int]] = [
    {"year": 2019, "n_bookings": 1_000, "revenue_per_booking": 200},
    {"year": 2020, "n_bookings": 1_000, "revenue_per_booking": 220},
    {"year": 2021, "n_bookings": 1_500, "revenue_per_booking": 220},
    {"year": 2022, "n_bookings": 1_700, "revenue_per_booking": 225},
]
revenue: pd.DataFrame = pd.DataFrame(data=data)
explainer = ice.SumExplainer(fact="revenue_per_booking", period="year", count="n_bookings")
explanation: pd.DataFrame = explainer(revenue)

Traceback

{
	"name": "ModuleNotFoundError",
	"message": "No module named 'pyarrow'",
	"stack": "---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
Cell In[4], line 2
      1 explainer = ice.SumExplainer(fact=\"revenue_per_booking\", period=\"year\", count=\"n_bookings\")
----> 2 explanation: pd.DataFrame = explainer(revenue)

File ..\site-packages\\icanexplain\\__init__.py:76, in Unpacker.__call__(self, table)
     75 def __call__(self, table):
---> 76     explanation = self._explanation(table)
     77     explanation_fmt = self._format(explanation)
     78     if is_pandas_dataframe(table):

File ..\site-packages\\icanexplain\\__init__.py:42, in coerce_table.<locals>._impl(self, table)
     39 @functools.wraps(method)
     40 def _impl(self, table):
     41     if is_pandas_dataframe(table):
---> 42         return method(self, ibis.memtable(table[self._necessary_columns]))
     43     if is_polars_dataframe(table):
     44         return method(self, ibis.memtable(table[self._necessary_columns]))

File ..\site-packages\\ibis\\expr\\api.py:462, in memtable(data, columns, schema, name)
    457 if columns is not None and schema is not None:
    458     raise NotImplementedError(
    459         \"passing `columns` and schema` is ambiguous; \"
    460         \"pass one or the other but not both\"
    461     )
--> 462 return _memtable(data, name=name, schema=schema, columns=columns)

File ..\site-packages\\ibis\\common\\dispatch.py:140, in lazy_singledispatch.<locals>.call(arg, *args, **kwargs)
    137 @functools.wraps(func)
    138 def call(arg, *args, **kwargs):
    139     impl = dispatcher.dispatch(type(arg))
--> 140     return impl(arg, *args, **kwargs)

File .\site-packages\\ibis\\expr\\api.py:475, in _memtable(data, columns, schema, name)
    465 @lazy_singledispatch
    466 def _memtable(
    467     data: pd.DataFrame | Any,
   (...)
    471     name: str | None = None,
    472 ) -> Table:
    473     import pandas as pd
--> 475     from ibis.formats.pandas import PandasDataFrameProxy
    477     if not isinstance(data, pd.DataFrame):
    478         df = pd.DataFrame(data, columns=columns)

File ..\site-packages\\ibis\\formats\\pandas.py:20
     18 from ibis.formats import DataMapper, SchemaMapper, TableProxy
     19 from ibis.formats.numpy import NumpyType
---> 20 from ibis.formats.pyarrow import PyArrowData, PyArrowSchema, PyArrowType
     22 if TYPE_CHECKING:
     23     import polars as pl

File ..site-packages\\ibis\\formats\\pyarrow.py:5
      1 from __future__ import annotations
      3 from typing import TYPE_CHECKING, Any
----> 5 import pyarrow as pa
      6 import pyarrow_hotfix  # noqa: F401
      8 import ibis.expr.datatypes as dt

ModuleNotFoundError: No module named 'pyarrow'"
}