Question: add TOP x column values and distribution.

[diegodewilde]

Hi,

I was looking at this project and I must say: it's awesome and something that dbt docs currently is missing.

One thing got in my mind is the question why there's not an option to add the TOP x column values and their distribution? Is there any other reason to not include this in the docs?

Like in this example where you show TOP 2 for example:

Column Name	Top 1 Value	Distribution	Top 2 Value	Distribution
Column 1	Value 1 A	0.50 ("number"/"total")	Value 1 B	0.20 ("number"/"total")
Column 2	Value 2 A	0.50 ("number"/"total")	Value 2 B	0.30 ("number"/"total")
Column 3	Value 3 A	0.10 ("number"/"total")	Value 3 B	0.05 ("number"/"total")
Column 4	Value 4 A	0.10 ("number"/"total"	Value 4 B	0.05 ("number"/"total")

Looking forward to your thoughts!

[stumelius]

@diegodewilde I've thought about adding a "mode" (most common value) profiling metric to the package but never around to implementing it. This proposal expands the mode concept into N most common values and I think it's a good idea.

Just throwing thoughts here:

• What would be a sensible default for the number of top values? 1, 2, 3?
• How should we name the columns? top_1_value, top_1_value_proportion, top_2_value, top_2_value_distribution, etc?
• Is there a better way to display the distributions than the (value, proportion) pairs for each top value?

Would you be interested in implementing this? :)

[diegodewilde]

Hi stumelius,

• It would make sense to make this dynamic, so you can choose the amount of top values you want to see in your docs. Not sure if that's possible?
• Sounds like a good suggestion, I actually think this goes hand-in-hand with the visualization you would like to see here.

[stumelius]

@diegodewilde Circling back to this. Is this feature still in your interests and if so, would you like to contribute? :)