Question: Is there a way to manually trigger a version? about logidze

[NullVoxPopuli]

Looks like this works:

 Logidze.without_logging { 
  MyRecord.connection.execute("SELECT logidze_snapshot('#{my_record.attributes.to_json}'::jsonb)")
}

[palkan]

Looks like this works:

Yep, without_logging disables only triggers.

But we don't have a procedure to update existing log_data (i.e. to add a new snapshot).
logidze_snapshot can be used only to create the initial log_data.

I think, it's a good idea to have an ability to manually create versions. Will put it on the TODO list)
Thanks!

[NullVoxPopuli]

oh. how does the updating process work?
(This is something I need for work)

maybe logidze_version instead of logidze_snapshot?

[palkan]

Take a look at the trigger function.

If you just want to add a new snapshot, you can do smth like this:

 CREATE OR REPLACE FUNCTION logidze_add_snapshot(row jsonb, log_data jsonb) RETURNS jsonb AS $body$
   DECLARE
     new_v integer;
     size integer;
     new_log_data jsonb;
   BEGIN
     -- Calculate new version number
     new_v := (row#>>'{log_data,h,-1,v}')::int + 1;
     -- Calculate the size of log
     size := jsonb_array_length(row#>'{log_data,h}');
     
     -- Add snapshot to the end of the log
     new_log_data := jsonb_set(
       row->'log_data',
       ARRAY['h', size::text],
       logidze_version(new_v, logidze_snapshot(row)),
       true
     );
     
     -- Update version number
     new_log_data := jsonb_set(
       new_log_data,
       '{v}',
       to_jsonb(new_v)
     );

     RETURN new_log_data;
   END
 $body$
 LANGUAGE plpgsql;

And then:

UPDATE users SET log_data = logidze_add_snapshot(to_jsonb(users.*), log_data) WHERE id=123

Note that the script above doesn't consider columns filtering and the history size limit.

[NullVoxPopuli]

This is what I have so far:

      CREATE OR REPLACE FUNCTION add_to_snapshot(id bigint, tbl text) RETURNS jsonb AS $body$
        BEGIN
          EXECUTE format('
            UPDATE %1$s
            SET log_data = logidze_add_snapshot(to_jsonb(%1$s.*), log_data)
            WHERE %1$s.id = %2$s
          ', tbl, id);
        END;
      $body$
      LANGUAGE plpgsql;

      CREATE OR REPLACE FUNCTION logidze_add_snapshot(obj jsonb, log_data jsonb) RETURNS jsonb AS $body$
        DECLARE
          new_v integer;
          size integer;
          new_log_data jsonb;
        BEGIN
          -- Calculate new version number
          new_v := (obj#>>'{log_data,h,-1,v}')::int + 1;
          -- Calculate the size of log
          size := jsonb_array_length(obj#>'{log_data,h}');

          -- Add snapshot to the end of the log
          new_log_data := jsonb_set(
            obj->'log_data',
            ARRAY['h', size::text],
            logidze_version(new_v, logidze_snapshot(obj)),
            true
          );

          -- Update version number
          new_log_data := jsonb_set(
            new_log_data,
            '{v}',
            to_jsonb(new_v)
          );

          RETURN new_log_data;
        END;
      $body$
      LANGUAGE plpgsql;

   # ApplicaitonRecord
  def save_version
    return false unless has_attribute?(:log_data)

    klass = self.class
    klass.connection.execute(%(
      SELECT add_to_snapshot(#{id}, '#{klass.table_name}')
    ))
  end

Currently getting this error:

ActiveRecord::StatementInvalid: PG::InvalidParameterValue: ERROR:  cannot set path in scalar
CONTEXT:  PL/pgSQL function logidze_add_snapshot(jsonb,jsonb) line 13 at assignment
SQL statement "
            UPDATE step_definitions
            SET log_data = logidze_add_snapshot(to_jsonb(step_definitions.*), log_data)
            WHERE step_definitions.id = 1410
          "
PL/pgSQL function add_to_snapshot(bigint,text) line 3 at EXECUTE
: 
      SELECT add_to_snapshot(1410, 'step_definitions')
    

not sure what's going on yet (new to this level of sql) -- but I figured I'd post what I have in case others want to see

[NullVoxPopuli]

it looks like the problem was that I'm trying to use these functions when log_data is nil. if log_data is already populated, it appears to work. So, I guess there needs to be some conditions where it creates a new snapshot if there isn't one already.

[palkan]

Yep, you should add some checks. Also you may want to add columns filtering support.