Sum aggregations for long values precision

[hackerwin7]

Create long type for InternalSum and SumAggregator to process the long field type of values without loosing precision.

relates #50503

[ghost]

Hi @hackerwin7, we have found your signature in our records, but it seems like you have signed with a different e-mail than the one used in your Git commit. Can you please add both of these e-mails into your Github profile (they can be hidden), so we can match your e-mails to your Github profile?

[hackerwin7]

retest

[hackerwin7]

ping @colings86 @jpountz

[elasticmachine]

Pinging @elastic/es-analytics-geo (:Analytics/Aggregations)

[polyfractal]

Hi @hackerwin7, thanks for the PR :) Please don't ping individual members though, when tickets are labeled it will ping the correct team to take a look.

I'm going to mark this ticket for team discussion, but I suspect we won't be able to merge it. There is some discussion about supporting more precise aggregations in this issue, for background: #46934

There are a few issues which make this problematic:

• Doubles are used everywhere in aggregations for several reasons, but one prominent reason is that it reduces the chance of overflows. With this change, a user could easily overflow the long and not realize
• Consistency. Since doubles are used everywhere, having just the sum agg switch behavior would be very confusing
• This breaks backwards compatibility, both at the wire transport level (fixable) but also in behavior. E.g. a sum agg would start to behave differently on the same data after upgrading. This is also fixable with flags and such, but again goes back to consistency... if we start to introduce "precise" aggs we should really support it for all metrics which is a much larger undertaking.
• Better, but potentially not good enough. E.g. the discussion in Add BigDecimal data type #46934 shows that some folks want/need much more robust control over metric precision. Introducing a change like this is more of a bandaid than a fix that would satisfy many use-cases, but still carries maintenance cost.

Marking as team-discuss to see what the rest of the analytics team has to say!

[polyfractal]

Related: #9545 and #43258

[hackerwin7]

@polyfractal
Thanks for your reply, after reviewing the discuss #46934 , what the future works in Elastic for support to compelete accuracy of long field type sum aggregations or just have no plan on this?

[polyfractal]

@hackerwin7 at the moment, no plans at all... just something we've been discussing. That's the main reason I didn't close this PR but marked it as team-discuss to see what the rest of the team thought. Sometimes small incremental improvements are better than nothing at all :)

[Laurianti]

Any news here?

[nik9000]

Actually yes! ESQL supports sum using long precision. FROM foo | STATS SUM(long_field) will sum a long.