feat: (Preview) Support arithmetics between dates and timedeltas #1413
Conversation
| @@ -740,6 +740,21 @@ def timestamp_sub_op_impl(x: ibis_types.TimestampValue, y: ibis_types.IntegerVal | ||
| return x - y.to_interval("us") | ||
|
|
||
|
|
||
| @scalar_op_compiler.register_binary_op(ops.date_diff_op) | ||
| def date_diff_op_impl(x: ibis_types.DateValue, y: ibis_types.DateValue): | ||
| return (x.delta(y, "day") * UNIT_TO_US_CONVERSION_FACTORS["d"]).floor() # type: ignore |
There was a problem hiding this comment.
what is the floor() needed for? don't we already have an integer?
There was a problem hiding this comment.
Good call!
UNIT_TO_US_CONVERSION_FACTORS["d"] has float type even though the value itself is an integer, so the multiplication gives us an Ibis.FloatValue. Calling floor() over the value was to cast it into Ibis.IntValue.
That said, floor() does seem unnecessary here. I used int(..) to cast the factor value instead, as it simplifies the generated SQL a bit.
|
|
||
| @scalar_op_compiler.register_binary_op(ops.date_add_op) | ||
| def date_add_op_impl(x: ibis_types.DateValue, y: ibis_types.IntegerValue): | ||
| return x.cast("timestamp") + y.to_interval("us") # type: ignore |
There was a problem hiding this comment.
why do we convert to interval datatype? I would rather just use https://cloud.google.com/bigquery/docs/reference/standard-sql/date_functions#date_add with an integer.
There was a problem hiding this comment.
We need the interval conversion because this is how Ibis generates date_add expressions. In other words, Ibis only recognizes DateValue + IntervalValue to get DATE_ADD(Date, Interval ...)
|
|
||
| @scalar_op_compiler.register_binary_op(ops.date_sub_op) | ||
| def date_sub_op_impl(x: ibis_types.DateValue, y: ibis_types.IntegerValue): | ||
| return x.cast("timestamp") - y.to_interval("us") # type: ignore |
There was a problem hiding this comment.
same here, would rather avoid casting to interval type
There was a problem hiding this comment.
Same reason as above.
Supported operations:
Test cases
date series + timedelta seriesanddate series - timedelta seriesare not supported by Pandas until 2.1.0, but our test setup uses version 1.5.3. However, I cannot increase the pandas version accordingly in test env 3.12 because it would introduce more breaking changes in irrelevant tests.That being said, operations between a series and a literal are still covered, and we will fill the gap once our Pandas dependency is increased to 2.1.0