bug(pandas): TypeError: pipe() missing 1 required positional argument: 'f' when calling .seconds() on Interval

[tswast]

If I to subtract two timestamp columns to get an interval, it works and I'm able to execute on the pandas backend. If I then try to call .seconds() or .milliseconds() on that interval, the pandas backend fails with TypeError: pipe() missing 1 required positional argument: 'f'

Working code:

import ibis
import pandas


source_df = pandas.DataFrame({"timecol": [pandas.Timestamp("2020-06-30T16:00:00Z")]})
target_df = pandas.DataFrame({"timecol": [pandas.Timestamp("2020-07-01T16:00:00Z")]})

client = ibis.pandas.connect(
    {"source": source_df, "target": target_df}
)

source = client.table("source")
target = client.table("target")
joined = source.cross_join(target)
difference = joined[
    (
        target["timecol"] - source["timecol"]
    ).name("difference")
]

print(difference.execute())

Output:

$ python interval.py
  difference
0     1 days

Broken code:

import ibis
import pandas


source_df = pandas.DataFrame({"timecol": [pandas.Timestamp("2020-06-30T16:00:00Z")]})
target_df = pandas.DataFrame({"timecol": [pandas.Timestamp("2020-07-01T16:00:00Z")]})

client = ibis.pandas.connect(
    {"source": source_df, "target": target_df}
)

source = client.table("source")
target = client.table("target")
joined = source.cross_join(target)
difference_seconds = joined[
    (
        target["timecol"] - source["timecol"]
    ).seconds().name("difference")
]

print(difference_seconds.execute())

Output:

$ python interval.py
Traceback (most recent call last):
  File "interval.py", line 17, in <module>
    target["timecol"] - source["timecol"]
TypeError: pipe() missing 1 required positional argument: 'f'

[tswast]

Maybe Ibis should be calling df["difference"].dt.total_seconds() instead of whatever it's currently doing for the seconds()

[gforsyth]

Hey @tswast -- thanks for the convenient test case!

It works if we remove the method call on seconds:

import ibis
import pandas


source_df = pandas.DataFrame({"timecol": [pandas.Timestamp("2020-06-30T16:00:00Z")]})
target_df = pandas.DataFrame({"timecol": [pandas.Timestamp("2020-07-01T16:00:00Z")]})

client = ibis.pandas.connect(
    {"source": source_df, "target": target_df}
)

source = client.table("source")
target = client.table("target")
joined = source.cross_join(target)
difference_seconds = joined[
    (
        target["timecol"] - source["timecol"]
    ).seconds.name("difference")
]

We're going to remove the blanket __call__ on expressions to avoid these errors moving forward