Evgenij Ryazanov

I saw it only once, but I run `TestAll` only few times after recent changes in LOB cleanup code. On the other hand, I didn't see it in recent CI...

Unfortunately, H2 doesn't have user-defined data types, H2 has only domains. Undocumented `CREATE TYPE` command shouldn't be used, this stuff is for some limited compatibility with PostgreSQL, maybe we should...

I think it shouldn't be possible to create a domain-based collection type or row value data type. Not only because the SQL Standard doesn't allow it, but also because it...

Domain in the Standard is not a data type. > A domain is a set of permissible values. A domain is defined in a schema and is identified by a...

In the Standard, domains are expected to be used when the same check constraint is needed for values of multiple columns. ```SQL CREATE TABLE TEST(A INT CHECK (A > 0),...

This method incorrectly creates regular LOB object even if data size is smaller than value of `MAX_LENGTH_INPLACE_LOB` setting, persistent databases are also affected, but it isn't that critical for them.

`LobStorageMap.createBlob(InputStream)` should be able to create `LobDataInMemory` for small objects, but implementation without performance regressions for normal use cases (persistent databases with relatively large objects) isn't that simple.

There are two remaining problems: 1. Methods without length limit should be able to create in-place LOBs (both in-memory and persistent databases are affected). I wrote a draft implementation, but...

I don't care too much about in-memory cases, but we have a minor problem with persistent databases too. H2 should always use in-place LOBs for objects smaller than `MAX_LENGTH_INPLACE_LOB` value,...

1. Cache size setting for `CacheLongKeyLIRS` doesn't set the actual memory limit, there was a discussion about that somewhere. Actual size is larger, especially when small objects are cached. From...