You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
B2T2 should give missing values a little more attention --- in the examples for sure, and perhaps in the datasheet. We don't want to proscribe how to handle NULL because it's a very difficult question, but we certainly want to highlight consequences of reasonable alternatives.
Two important scenarios to illustrate (from @ramanshah , thanks!):
count with missing data.
The sum of the output of count should be equal to the number of rows in the input table. This would require a representation of null on a par with non-null data, which is the messy thing about a given type system that B2T2 might keep people honest about.
groupBy with columns A and B, where one or both columns have missing values.
Or look at column A grouped by column B, but your groupby can't handle the case where a cell in column B is null, so you have to painstakingly tack on duplicative code to study this subset of rows. You can imagine that such situations are the norm. Organizations routinely sustain damage when an incautious analyst or data scientist counts or groups by naively, but the cases where column B was null were material (or even the whole story), so there goes $N million. I'd hope that B2T2 could elucidate this.
The text was updated successfully, but these errors were encountered:
B2T2 should give missing values a little more attention --- in the examples for sure, and perhaps in the datasheet. We don't want to proscribe how to handle NULL because it's a very difficult question, but we certainly want to highlight consequences of reasonable alternatives.
Two important scenarios to illustrate (from @ramanshah , thanks!):
count
with missing data.groupBy
with columns A and B, where one or both columns have missing values.The text was updated successfully, but these errors were encountered: