[
https://issues.apache.org/jira/browse/CALCITE-2679?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16730473#comment-16730473
]
Andrei Sereda commented on CALCITE-2679:
----------------------------------------
[~Functor10], I believe a bug was introduced by this change.
Please check the following test cases (which currently fail):
{code:java}
// combine any_value with other aggregation functions (eg. max)
CalciteAssert.that()
.with(newConnectionFactory())
.query("select cat1, any_value(cat2), max(val1) from view group by
cat1")
.returnsUnordered("cat1=a; EXPR$1=g; EXPR$2=1.0",
"cat1=null; EXPR$1=g; EXPR$2=null",
"cat1=b; EXPR$1=h; EXPR$2=7.0");
CalciteAssert.that()
.with(newConnectionFactory())
.query("select max(val1), cat1, any_value(cat2) from view group by
cat1")
.returnsUnordered("EXPR$0=1.0; cat1=a; EXPR$2=g",
"EXPR$0=null; cat1=null; EXPR$2=g",
"EXPR$0=7.0; cat1=b; EXPR$2=h");
CalciteAssert.that()
.with(newConnectionFactory())
.query("select any_value(cat2), cat1, max(val1) from view group by
cat1")
.returnsUnordered("EXPR$0=g; cat1=a; EXPR$2=1.0",
"EXPR$0=g; cat1=null; EXPR$2=null",
"EXPR$0=h; cat1=b; EXPR$2=7.0");
{code}
Probable reason is that GroupValue is added multiple times resulting in more
rows than necessary.
> Group by without aggregation function cannot be translated to correct JSON in
> Elasticsearch Adapter
> ---------------------------------------------------------------------------------------------------
>
> Key: CALCITE-2679
> URL: https://issues.apache.org/jira/browse/CALCITE-2679
> Project: Calcite
> Issue Type: Bug
> Components: elasticsearch-adapter
> Affects Versions: 1.18.0
> Reporter: Siyuan Liu
> Assignee: Julian Hyde
> Priority: Major
> Labels: easyfix
> Fix For: 1.18.0
>
>
> The Elasticsearch Adapter of the current master branch has some problems when
> querying for group by clause without aggregation function. For example, when
> I query the following SQL:
> {code:sql}
> select state, city from zips group by state, city
> {code}
> After translation, adapter should return the following JSON:
> {code:json}
> {
> "_source": false,
> "size": 0,
> "aggregations": {
> "g_state": {
> "terms": {
> "field": "state",
> "missing": "__MISSING__"
> },
> "aggregations": {
> "g_city ": {
> "terms ": {
> "field ":"city ",
> "missing":"__MISSING__ "
> }
> }
> }
> }
> }
> }
> {code}
> But it returns the following JSON now:
> {code:sql}
> {
> "_source": ["state", "city"]
> }
> {code}
> The reason for this problem is that there is a missing condition for judging
> the aggregation query.
> In addition, there is the other associated problem. After building
> aggregation JSON, the code of current version will remove empty aggregation
> block in ElasticsearchTable class, the code just like this:
> {code:java}
> JsonNode agg = query;
> while (agg.has(AGGREGATIONS) && agg.get(AGGREGATIONS).elements().hasNext()) {
> agg = agg.get(AGGREGATIONS);
> }
> ((ObjectNode) agg).remove(AGGREGATIONS);
> {code}
> But if input the JSON like above, this code will not work out.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
