Databricks Connections

AACAAC supports managed tables and external delta tables from Databricks. For external delta tables, AACAAConly supports append and truncate actions.

You can't query materialized views with job clusters.

Due to a Databricks limitation, only use date values after October 1, 1582. Date values before October 1, 1582 result in an error.

Typecasting issues during integer division expression. The output becomes Float even if you choose Integer for the output column data type. For example, the output of the expression [total_weight]/[quantity], where both columns are Integer data types, becomes a Float data type.

Precision error with float data types. For example, the expression [weight]-1 => 1.099999 in Databricks SQL where {int weight=2.1}. The same expression using Photon as the engine results in 1.1.

There is no specific way to order rows except when you specify the order_by clause. This affects the 行IDツール and ユニオンツール.

For JSON/MAP data in inputs - the SQL pushdown might not work. DatabricksSpark would be used as fallback.

DateTime output ignores the millisecond values of the input DateTime value. For example, 2021-12-31 23:59:59.999 → 2021-12-31 23:59:59.000. Import jobs also ignore millisecond values when creating a dataset.

The substring function might give incorrect results when the value position and length are very large.

Precision issues with Double and Float data types. For example:

select  687399551400673279 + 5.35
returns 687399551400673284.35

select  687399551400673279 + double(5.35)
returns 687399551400673300

select  double(687399551400673279) + double(5.35)
returns 687399551400673300

The exp function returns infinity in cases with very large Float values.

The rpad function produces empty results as opposed to null. For example:

SELECT rpad('hi', 500000000000000000000000000000000, 'ab');
returns ""