SPOC-283: Reallow conversions

[mason-sharp]

• In 2.2, it was possible to use snowflake.convert_sequence_to_snowflake() for cases like this.

CREATE SEQUENCE favorite_seq;
CREATE TABLE t4 (x integer DEFAULT nextval('favorite_seq'::regclass));

This was no longer possible in 2.3, add back such support.

• Also reallow conversions of plain sequences in case applications are using sequences not associated with any tables.

• The function snowflake.convert_sequence_to_snowflake() is updated to use consistent variable naming conventions.

• Prepare code for a future 2.4 release

[danolivo]

I think you can patch version 2.3 directly, if it hasn't been released yet. At least, it would be easier to see the changes.

[mason-sharp]

Good point! We did create a tag though. Let me check with a couple of people.

[mason-sharp]

We will go with 2.4 after all.

[danolivo]

What is the reason? Someone officially released 2.3?

[mason-sharp]

The v2.3 has been sitting out there a while. No big deal to just go to 2.4 instead of messing someone up who is using the repo.

[danolivo]

I'd say we should be more strict about release cuts in this case. A new version of an extension is always a massive pain for developers and a source of multiple bugs.
If you still insist on the new version, please don't forget to attach snowflake--2.4.sql.

[mason-sharp]

There is no need for a snowflake--2.4.sql file, there is an ugrade path file to 2.4. CREATE EXTENSION will execute snowflake--2.3.sql and then snowflake--2.3--2.4.sql.

[danolivo]

Please correct me if I'm wrong, but we have versions like snowflake--2.3 and spock--5.0.0.sql. This is a common practice. The reason behind it is that the update scripts can be messy for developers and prone to errors. Furthermore, users generally prefer to see the complete script rather than just the transient parts.