diqiu50 opened a new issue, #13040:
URL: https://github.com/apache/gravitino/issues/13040
### Version
main branch
### Describe what's wrong
Two columns in the PostgreSQL catalog are reported with types that do not
describe the data the source column can hold.
1. A `NUMERIC` declared without precision and scale is reported as
`decimal(38,18)`. PostgreSQL accepts up to 131072 digits before and 16383
digits after the decimal point for such a column, and its precision and scale
vary per row. The reported type claims 20 integer digits and 18 fractional
digits that the source never declared, so a consumer that trusts it truncates
or overflows on data PostgreSQL accepts. Constrained `NUMERIC(p, s)` is handled
correctly.
2. Array columns are reported with `containsNull: false`. PostgreSQL array
elements always accept NULL and cannot be declared otherwise, so `containsNull:
false` is never a true statement about a PostgreSQL array column.
### Error message and/or stacktrace
No exception is raised. Both types are reported silently:
```
["c_numeric_nodef", "decimal(38,18)"]
["c_int_array",
{"type":"list","containsNull":false,"elementType":"integer"}]
```
### How to reproduce
Against PostgreSQL 16.6 with a `jdbc-postgresql` catalog:
```sql
CREATE TABLE type_probe (c_numeric_nodef NUMERIC, c_int_array INTEGER[]);
INSERT INTO type_probe VALUES
(123456789012345678901234567890.123456789012345678901234567890,
ARRAY[1,NULL,3]);
```
PostgreSQL stores both values intact. Load the table through the catalog and
compare the reported column types with the values above.
### Additional context
Observed on 1.3.0-rc2. The rest of the PostgreSQL type map is in good shape:
unmappable types return `external` carrying their source spelling, and
precision is preserved on every constrained numeric.
Reporting unconstrained `NUMERIC` as the external type `numeric` requires
the Trino, Spark and Flink connectors to read an external type as a string,
otherwise a single such column fails the type conversion of the whole table.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
