How to do group by count distinct with multivalued dimension?

I have a pretty simple use case with the following data:

the columns are id, time, sessionId, funnelLevel and segments

"1", "2018-07-17T00:00:00.001-06:00", "session-1", "1", "a"
"2", "2018-07-17T01:00:00.001-06:00", "session-1", "1", "a|b"

I am trying to group by segments and count distinct session ids per segment.

If I lay out the same data without multivalued segment dimension

"1", "2018-07-17T00:00:00.001-06:00", "session-1", "1", "a"
"2", "2018-07-17T01:00:00.001-06:00", "session-1", "1", "a"
"3", "2018-07-17T01:00:00.001-06:00", "session-1", "1", "b"

This query can be used:

{

“queryType” : “groupBy”,

“dataSource” : “s8”,

“intervals” : [ “2018-07-01/2018-08-01” ],

“granularity” : “all”,

“dimensions” : [ “segments” ],

“aggregations” : [ {

“type” : “cardinality”,

“name” : “distinct_count”,

“fields” : [ “segments”, “sessionId” ],

“byRow” : true

} ]

}

and it produces the expected result:

{

“version” : “v1”,

“timestamp” : “2018-07-01T00:00:00.000Z”,

“event” : {

“distinct_count” : 1.0002442201269182,

“segments” : “a”

}

}, {

“version” : “v1”,

“timestamp” : “2018-07-01T00:00:00.000Z”,

“event” : {

“distinct_count” : 1.0002442201269182,

“segments” : “b”

}

} ]

The same query with the multivalued data produces this output. Note that segment “a” has a count of 2 instead of th

[ {

“version” : “v1”,

“timestamp” : “2018-07-01T00:00:00.000Z”,

“event” : {

“distinct_count” : 2.000977198748901,

“segments” : “a”

}

}, {

“version” : “v1”,

“timestamp” : “2018-07-01T00:00:00.000Z”,

“event” : {

“distinct_count” : 1.0002442201269182,

“segments” : “b”

}

} ]

I posted a bit prematurely :).

How can I get the expected result use multivalued dimension?

Thanks,

Patrick