HHH-19331: query plan cache should take parameter type into accound for cast

[victornoel]

My understanding of the problem is that the query plan cache has some mechanisms to know if a cached query plan is invalid. Until now it handled two cases in relation to query parameters:

• if the value of the parameter was used directly in the rendered SQL, then the same query with different parameters cannot reuse the cached SQL
• if the parameters are related to limit/offset, for select queries, then we are not so much interested in the value of the limit/offset parameters but with the fact they are specified or not to know if we can reuse the cached SQL.

My change add one more case: when a parameter type is involved in the rendered SQL, precisely here if we render a cast that was not explicit in the HQL query, then the cached SQL cannot be reused with a different parameter type. I haven't found out if there were more cases of this problem, other than cast.

I didn't go all the way into remembering the type of the parameter used during the SQL rendering but just on the presence of the parameter in appliedParameters. Most certainly it should be possible to optimize this but I didn't want to make things too complex.

I went with refactoring things a bit to 1) remove duplicate code between JdbcOperationQuerySelect and AbstractJdbcOperationQuery. I think it's equivalent but I may have missed something. I think it also remove chances of NPE (as mentioned in a TODO I removed) when appliedBinding is null.

---

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license
and can be relicensed under the terms of the LGPL v2.1 license in the future at the maintainers' discretion.
For more information on licensing, please check here.

---

https://hibernate.atlassian.net/browse/HHH-19331

[victornoel]

Making this a draft based on the discussion in Zulip and the fact that

• this is really corner case (it ONLY applies to those "select ?1" kind of query where the hibernate parser cannot infer the expected type of the parameter)
• it should be improved even if we went forward