Alter MODE to work on all datatypes and return multiple values

[richthegeek]

This alteration allows MODE to be used to get, for example, the "10 most common strings" in a dataset.

I've never written in Go before, so if I've made some silly choices let me know!

I assume it needs some tests attached, although I'm not sure how to write these.

[richthegeek]

Old function signature:

MODE(column{int, float})

New function signature:
MODE(column, [size = 1])

[maxd]

I want to clarify how the size option works.

At present MODE function return only most frequent values. For example for dataset:

count of city_id	city_id
1439	498817
1439	524901
1416	472459
1416	500096
1416	520555
1416	1486209
1416	1497543
1355	2013348
1344	542374
1338	542420

MODE function return only city_ids where count = 1493 i.e. 498817 and 524901.

Which values will be returned is I set size = 10 for this dataset?

Thanks

[richthegeek]

It will return the 10 most frequent values, in order of frequency. So for this dataset, it will return as it was supplied (it's already sorted by count). It does this by taking items from the list in descending order until it has size values.

I have infact just realised that this code changes how mode works, so if you called mode(city_id) now it would default to size=1 and thus return only 498817. I've commented on the commit with the fixes required, i'll commit those in a minute.

I think allowing any data-type in mode is a good thing, but perhaps the size option is better suited to a top(column, n) or similarly-named function?

[maxd]

I read article about MODE and I think that the size option is appropriate only for top(column, size) function.

@jvshahid What do you think?

[richthegeek]

I've implemented the suggested change on a branch in my own repository (richthegeek/influxdb@dbaaacf55b)

This still allows mode(column) to work on all column types, whilst top(column, size = 1) is also available.

There's a lot of duplicated code in there, but it works.