GP: location stats and grouping

[MartinKolarik]

Just an idea for discussion: when multiple locations are specified for measurement by a user, they might be looking for some comparison of the two locations/regions. It might be useful to:

1. Use a different separator and a label for a section of results that matched a particular location (only if more than one is specified), e.g. https://dl.dropboxusercontent.com/scl/fi/2xj47iyhivffvzjmx7d5c/2023-11-26_00-57-56.png?rlkey=ld5h9qz2fp9j5jk24rfmfepip&dl=0
2. Provide some summary per-location stats. Could be the same that we have for each test, but averaged for all probes in the location. Could be displayed either above each location section, or at the very top as a summary.

[jimaek]

Wouldn't the table mode address this? Maybe we could simply expand it with missing data?

[MartinKolarik]

The table mode gives you the number with less output, but it still doesn't offer any per-location overview, does it? For 5 EU probes and 5 US probes, I'll still see 10 rows with numbers.

What I'm thinking about here is a) some visual indicator within the results to visually separate them (might very well be just a slightly different background, etc) and b) 1 extra row for average EU and 1 extra row for average US.

[jimaek]

I guess it depends on what you mean with overview. What if the user selected "World"?

Maybe an extra section at the end of every measurement with all kinds of summaries?

[MartinKolarik]

In this case, I meant specifically when you request multiple locations separated by a comma. There, the grouping is clear from the request, and I'd like to see a short overview at the very top, not the bottom.

I also considered the case where you run 100 "World" locations. Some overview might be useful there as well, but I think that's a different case and a bit more complex. For World it would make sense to group by continents, probably, but if I request 100 in one region or country, what then? Per city, or show nothing? Allow the user to choose?

[jimaek]

It sounds like what I did at perfops with grouping of time series data based on any location type. But not sure if we need to make it so complicated.

A smart overview that shows what it thinks is best sounds interesting but designing it to handle all possible edge cases could be a pain.

Another option is to make separate tools aimed at a specific use case where we could add grouping