pgjdbc / r2dbc-postgresql

When selecting many fields, a lot of CPU is consumed because the fields are looped.

In one case I encountered, with a select of millions of records with about 700 fields, I forced the query to stop because it was consuming 100% CPU for several hours.

At this time, I checked Java's stack trace many times, and most of them were processing the following for statement.

r2dbc-postgresql/src/main/java/io/r2dbc/postgresql/PostgresqlRow.java

Lines 168 to 174 in d047276

    
           for (int i = 0; i < this.fields.size(); i++) { 
        
               RowDescription.Field field = this.fields.get(i); 
        
               if (field.getName().equalsIgnoreCase(name)) { 
        
                   return i; 
        
               } 
        
           }

Would it be possible to prepare field name and order mappings in advance?

This is possible. PGJDBC uses a lazy approach to create the column name to index map, see https://github.com/pgjdbc/pgjdbc/blob/master/pgjdbc/src/main/java/org/postgresql/jdbc/PgResultSet.java#L3108-L3136

Happy to merge a pull request.

	for (int i = 0; i < this.fields.size(); i++) {
	RowDescription.Field field = this.fields.get(i);

	if (field.getName().equalsIgnoreCase(name)) {
	return i;
	}
	}

Performance issue with PostgresqlRow.getColumn(String name) when select many columns