Decimal multiplied by Float produces incorrect results

[andygrove]

Describe the bug
Decimal multiplied by Float produces incorrect results

To Reproduce

This is fine:

❯ select cast(400420638.54 as decimal(12,2));
+-----------------------+
| Float64(400420638.54) |
+-----------------------+
| 400420638.54          |
+-----------------------+
1 row in set. Query took 0.000 seconds.

This is not fine:

❯ select cast(400420638.54 as decimal(12,2)) * 1.0;
+------------------------------------+
| Float64(400420638.54) * Float64(1) |
+------------------------------------+
| 10377.38413171004264709            |
+------------------------------------+
1 row in set. Query took 0.000 seconds.

More info:

❯ explain select cast(400420638.54 as decimal(12,2)) * 1.0;
+---------------+--------------------------------------------------------------------------------------------------+
| plan_type     | plan                                                                                             |
+---------------+--------------------------------------------------------------------------------------------------+
| logical_plan  | Projection: Decimal128(Some(1037738413171004264709),38,17) AS Float64(400420638.54) * Float64(1) |
|               |   EmptyRelation                                                                                  |
| physical_plan | ProjectionExec: expr=[Some(1037738413171004264709),38,17 as Float64(400420638.54) * Float64(1)]  |
|               |   EmptyExec: produce_one_row=true                                                                |
|               |                                                                                                  |
+---------------+--------------------------------------------------------------------------------------------------+
2 rows in set. Query took 0.001 seconds.

Expected behavior

Looks like type coercion needs to cast both sides to decimal.

❯ select cast(400420638.54 as decimal(12,2)) * cast(1.0 as decimal(12,2));
+------------------------------------+
| Float64(400420638.54) * Float64(1) |
+------------------------------------+
| 400420638.54                       |
+------------------------------------+
1 row in set. Query took 0.000 seconds.

Additional context
Add any other context about the problem here.

[andygrove]

Here is some more odd behavior:

❯ select cast(400420638.54 as decimal(32,15)) * cast(0.1 as decimal(32,15));
+--------------------------------------+
| Float64(400420638.54) * Float64(0.1) |
+--------------------------------------+
| 40042063.854000005133107             |
+--------------------------------------+
1 row in set. Query took 0.001 seconds.

❯ select cast(400420638.54 as decimal(32,16)) * cast(0.1 as decimal(32,16));
+--------------------------------------+
| Float64(400420638.54) * Float64(0.1) |
+--------------------------------------+
| -791820.1765126158512068             |
+--------------------------------------+
1 row in set. Query took 0.000 seconds.

Postgres for reference:

postgres=# select cast(400420638.54 as decimal(32,16)) * cast(0.1 as decimal(32,16));
                 ?column?                  
-------------------------------------------
 40042063.85400000000000000000000000000000
(1 row)

[andygrove]

@liukun4515 are you familiar with this area of code?

[liukun4515]

@liukun4515 are you familiar with this area of code?

are you familiar with this area of code?

@andygrove

sorry for the later reply.

Recently, I am busy for some internal business.
Will be back soon. 😁