Sequel 5.89.0 Released

[jeremyevans]

Sequel 5.89.0 has been released!

New Features

• A query_blocker extension has been added, for blocking queries
  inside a block:

    DB.extension :query_blocker
    DB[:table].all # works
    DB.block_queries do
      DB[:table].all # raises
    end

  To handle concurrency, you can specify a scope for the block:

    DB.block_queries(scope: :thread) do
      # Block queries for current thread
    end
  
    DB.block_queries(scope: :fiber) do
      # Block queries for current fiber
    end

  In some cases, you may want to block queries in general, and only
  allow them in specific places. The query blocker extension
  supports this:

    DB.block_queries do
      # Queries blocked
      DB.allow_queries do
        # Queries allowed
      end
      # Queries blocked
    end

• The alter_table add_primary_key and add_unique_constraint methods
  now support a :using_index option on PostgreSQL, to add the
  constraint using an existing index, instead of building a new
  unique index to enforce the constraint.

• A :compare_connections_by_identity Database option is now
  supported, which can be set to false to not use compare_by_identity
  on hashes keyed by connections. This should only be used to work
  around bugs in other libraries or ruby implementations.

Other Improvements

• All anonymous classes and modules created by Sequel now have
  temporary names set when using Ruby 3.3+. This makes debugging
  and introspection easier. Example:

    class Foo < Sequel::Model(:foo)
      def_column_alias :a, :a
      puts ancestors[0..3]
    end

  Previous and when running on Ruby < 3.3 Output:

    Foo
    #<Module:0x00000846cf717aa0>
    #<Class:0x00000846cf718040>
    Sequel::Model
  

  New output when running on Ruby 3.3+:

    Foo
    Foo::@overridable_methods_module
    Sequel::_Model(:foo)
    Sequel::Model
  

• The connection_validator extension now handles exceptions
  raised by Database#valid_connection?, which shouldn't happen, but
  would result in the thread/fiber being assigned the connection
  permanently in that case.

• The mysql2 adapter now handles invalid statement handles when
  closing prepared statements. This only affected cases where
  you were changing the definition of already prepared statement.

Thanks,
Jeremy