postgresql_info: fix SQL error of issue #790

[betanummeric]

fixing #790

The dynamic query construction fails because the column type is bytes in python, but the code expects str. The python conversion from bytes to str addess extra characters which break the SQL syntax of the following query.
I guess psycopg 2 vs 3 behave different regarding types, which was unexpected in the code.

This fix using the quote_ident function converts the type in SQL which should result in str in python. By the way it adds support for arbitrary column names (although this is rather unlikely to become relevant in a table of pg_catalog).

postgres=# SELECT pg_typeof(quote_ident(column_name)), pg_typeof(column_name) FROM information_schema.columns WHERE table_schema = 'pg_catalog' AND table_name = 'pg_subscription' limit 1;
 pg_typeof |             pg_typeof             
-----------+-----------------------------------
 text      | information_schema.sql_identifier

[betanummeric]

I didn't test the change yet, I can't promise to do that soon. If someone wants to take this PR over from here - feel free.

[Andersson007]

I didn't test the change yet, I can't promise to do that soon. If someone wants to take this PR over from here - feel free.

@betanummeric thanks for the PR! i think it's OK to wait generally:) thanks for being explicit

[Andersson007]

@betanummeric i played with it manually, and it doesn't work
If you look at possible solutions, it'd be great:

• [DRAFT] postgresql_info: fix failure when col names are returned as binary #804
• [DRAFT] postgresql_info: handle encoding #805

[Andersson007]

@betanummeric it turned out that the issue is hard to catch. I'll close this PR as i did with related mine. Feel free to re-open if you think it's relevant. Thanks for investigating and submitting it!