bug: casting Timestamps to timezone works for scalars but not columns

[NickCrews]

What happened?

import datetime
import ibis

t = ibis.memtable(
    {"ts": [datetime.datetime.fromisoformat("2023-10-01 12:00:00+00")]},
    schema=ibis.schema([("ts", "timestamp('UTC', 6)")]),
)

def view(ts):
    print(ts.type())
    print(ts.to_sql())
    print(ts.to_pyarrow())
    print("\n\n")

scalar = ibis.timestamp("2023-10-01 12:00:00+00").cast("timestamp('UTC', 6)")
column = t.ts

view(scalar)
view(column)
view(scalar.cast("timestamp('America/Anchorage')"))
view(column.cast("timestamp('America/Anchorage')"))

Results in (note how the needed cast is missing in the final block)

timestamp('UTC', 6)
SELECT
  CAST(MAKE_TIMESTAMPTZ(2023, 10, 1, 12, 0, 0.0, 'UTC') AS TIMESTAMPTZ) AS "Cast(datetime.datetime(2023, 10, 1, 12, 0, tzinfo=tzutc()), timestamp('UTC', 6))"
2023-10-01 12:00:00+00:00



timestamp('UTC', 6)
SELECT
  *
FROM "ibis_pandas_memtable_5eusqqjk3rcsrnmoeobf7clqy4" AS "t0"
[
  [
    2023-10-01 12:00:00.000000Z
  ]
]



timestamp('America/Anchorage')
SELECT
  CAST(CAST(MAKE_TIMESTAMPTZ(2023, 10, 1, 12, 0, 0.0, 'UTC') AS TIMESTAMPTZ) AS TIMESTAMPTZ) AS "Cast(Cast(datetime.datetime(2023, 10, 1, 12, 0, tzinfo=tzutc()), timestamp('UTC', 6)), timestamp('America/Anchorage'))"
2023-10-01 04:00:00-08:00



timestamp('America/Anchorage')
SELECT
  CAST("t0"."ts" AS TIMESTAMPTZ) AS "Cast(ts, timestamp('America/Anchorage'))"
FROM "ibis_pandas_memtable_5eusqqjk3rcsrnmoeobf7clqy4" AS "t0"
[
  [
    2023-10-01 12:00:00.000000Z
  ]
]

What version of ibis are you using?

main

What backend(s) are you using, if any?

duckdb

Relevant log output

Code of Conduct

• I agree to follow this project's Code of Conduct

[NickCrews]

hmm, in test_temporal.py, I don't even see any tests for this sort of thing. Is this an officially supported feature? IS there some other API for doing this?

[contang0]

Thanks for raising this, was going nuts trying to convert a timezone. I thought from_unit is supposed to handle conversion, but it doesn't work either.