Cosmos: DISTINCT results with JOIN and ORDER BY - azure

Question

I'm trying to write a query that uses a JOIN to perform a geo-spatial match against locations in a array. I got it working, but added DISTINCT in order to de-duplicate (Query A):
SELECT DISTINCT VALUE
u
FROM
u
JOIN loc IN u.locations
WHERE
ST_WITHIN(
{'type':'Point','coordinates':[loc.longitude,loc.latitude]},
{'type':'Polygon','coordinates':[[[-108,-43],[-108,-40],[-110,-40],[-110,-43],[-108,-43]]]})
However, I then found that combining DISTINCT with continuation tokens isn't supported unless you also add ORDER BY:
System.ArgumentException: Distict query requires a matching order by in order to return a continuation token. If you would like to serve this query through continuation tokens, then please rewrite the query in the form 'SELECT DISTINCT VALUE c.blah FROM c ORDER BY c.blah' and please make sure that there is a range index on 'c.blah'.
So I tried adding ORDER BY like this (Query B):
SELECT DISTINCT VALUE
u
FROM
u
JOIN loc IN u.locations
WHERE
ST_WITHIN(
{'type':'Point','coordinates':[loc.longitude,loc.latitude]},
{'type':'Polygon','coordinates':[[[-108,-43],[-108,-40],[-110,-40],[-110,-43],[-108,-43]]]})
ORDER BY
u.created
The problem is, the DISTINCT no longer appears to be taking effect because it returns, for example, the same record twice.
To reproduce this, create a single document with this data:
{
"id": "b6dd3e9b-e6c5-4e5a-a257-371e386f1c2e",
"locations": [
{
"latitude": -42,
"longitude": -109
},
{
"latitude": -42,
"longitude": -109
}
],
"created": "2019-03-06T03:43:52.328Z"
}
Then run Query A above. You will get a single result, despite the fact that both locations match the predicate. If you remove the DISTINCT, you'll get the same document twice.
Now run Query B and you'll see it returns the same document twice, despite the DISTINCT clause.
What am I doing wrong here?

Answer 1

Reproduced your issue indeed,based on my researching,it seems a defect in cosmos db distinct query. Please refer to this link:Provide support for DISTINCT.
This feature is broke in the data explorer. Because cosmos can only
return 100 results per page at a time, the distinct keyword will only
apply to a single page. So, if your result set contains more than 100
results, you may still get duplicates back - they will simply be on
separately paged result sets.
You could describe your own situation and vote up this feedback case.