Hi, I don't want to steal Kevin Minder's work, but one of our important customers got serious performance issues due to this bug.
This ticket is untouched for 7 years, so if you don't mind, I would hand it over. Reminder about the bug: When we run a select on a table, Beeline has an option to do extra coloring and formatting in the console output if we use the default output format. It wants to highlight the columns that are part of the primary key. Unfortunately, it does it in a way by querying the metadata for each column and each row: https://github.com/apache/hive/blob/master/beeline/src/java/org/apache/hive/beeline/Rows.java#L92 And it can lead to serious performance problems. I believe the problem can be solved in two separate ways: - Firstly, as a quick solution, coloring is disabled by default. So, we can completely get rid of those isPrimaryKey calls in case of no coloring required. I started to work on a draft to do that: https://github.com/apache/hive/blob/master/beeline/src/java/org/apache/hive/beeline/Rows.java#L92 (note it is not the final version, just the validation). - Secondly, as a long-term solution, we can query the primary keys only once during a select. And use that information later, at the actual coloring. And one note: In my personal opinion, I can accept that there are people who prefer coloring the output. I see no value in this at all. What do you think about removing this option from Beeline at all? Note: there is a screenshot from my test. That simple test had 4 calls:[image: image.png] Thanks, Zsolt Miskolczi
