Add regr_slope() aggregate function #7135
Conversation
github-actions
bot
added
logical-expr
|
Thank you @2010YOUY01 -- I'll check this out carefully over the next few days. One thing I wonder about is how large the datafusion library is getting. I was thinking maybe it would be useful to start splitting some of the more advanced / special purpose functionality into their own crates, as partly described in #7110. Curious what you think about this |
|
I am sorry I ran out of time today to review this. It is on my list for tomorrow |
There was a problem hiding this comment.
Thank you @2010YOUY01 -- I found this PR easy to read and well tested and documented. I think it is a very nice contribution
Looks like it is implemented in several other databases:
https://docs.snowflake.com/en/sql-reference/functions/regr_slope
| @@ -384,6 +385,22 @@ var_samp(expression) | |||
| - **expression**: Expression to operate on. | |||
| Can be a constant, column, or function, and any combination of arithmetic operators. | |||
|
|
|||
| ### `regr_slope` | |||
|
I don't really have any suggestions or comments for this PR so merging it in |
Which issue does this PR close?
NA
Rationale for this change
I'm planning to add PostgreSQL linear regression aggregate functions. Their implementations are similar, so I'd like to implement
regr_slope(y, x)first, and the remaining functions should be easy to extend.What changes are included in this PR?
Adding
regr_slope(y, x)aggregate function, mainly implementingAccumulatorforRegrSlopeAccumulatorThe linear regression is calculated using
COVAR_POP(x,y) / VAR_POP(x)One-pass and parallel algorithm for
covar/varreference: https://en.wikipedia.org/wiki/Algorithms_for_calculating_variance#Welford's_online_algorithmAre these changes tested?
All end-to-end tests produced the same result compared to Postgres
Are there any user-facing changes?
No