Distinct value count for columns in table

[kalyanm2305]

Hi, I have started using duckdb ui and i have question on the right side menu that is being displayed when a sql query is executed.
When I run a simple query select * from table, it quickly shows me all the relevant info for each column where it tells me the number of unique values for that column and for each column it shows the count of records for top 10 values in case of varchar column. So i would like to know if the count of unique that is being displayed next to the column name is an approximation or the exact value.
I've noticed that the value shown doesn't tie to the actual number in the table.

[jraymakers]

While the number of unique values is approximate (it uses approx_count_distinct), the count of records for each value in the top 10 is exact.

In general, we show in a tooltip if a value is approximate. See:

[data-igor]

Is there a way to change it from approx in settings? I haven't found it, it's really confusing for data profiling.

[jraymakers]

No, getting exact counts would often be too expensive for this case.

But you can of course write your own query to get exact counts, if that works well for your case.

[data-igor]

Of course I can, but I was hoping to use duckdb as a quick data profiling tool. Having approximate values doesn't solve that problem, which is a shame.

[jraymakers]

It's an interesting idea to enable requesting exact values. It's likely to be slower, sometimes significantly so, but for some use cases it may work well. We'll consider it.

[data-igor]

That'd be fantastic. I presented duckdb on Tableau Conference last May as a potential simple tool to run locally and do data profiling. It's a legitimate use case since the DB is so fast and easy to use.