With the extreme popularity of pandas, I think a lot of Python programmers would be amazed how clean and easy-to-read SQL queries look like, compared to the (downright) mess that's being written to query pandas dataframes.
Depends on the transformation. Dropping a couple of columns in a wide dataset is extraordinarily ugly in sql vs anything else, for example.
...how is it ugly in SQL? Don't SELECT those columns. Or, if you need to do multiple things with the dataset sans those columns, use a CTE of a query that doesn't SELECT those columns. Or a view.
It seems like SQL is only uglier if you're used to "SELECT *"ing, which is asking for trouble--in SQL or when doing the equivalent with other row/column data manipulation tools.
I don’t see what the issue is with using something like spark’s drop over using select.
Just don't pick them in the first place?
Though I'll grant you that extremely wide data is a lengthy experience to write SQL for.
Agreed! Pandas is great but at this point instead of using logic to solve data structuring tasks I often find myself googling for an optimized built-in Pandas method to help me out. Leads to less elegant code -- not sure if it is less readable though.
Well R has packages that buy you similar functionality eg dplyr.
Re:Medium-hard SQL interview questions
by alexilliamson on Sunday May 03, 2020 @02:09PM
(#23060776)
Dplyr is IMO the best thing to come from R. The design is so much simpler than pandas and the operations mirror SQL operations so closely.
It is so usefull, reliable and does not change every year.