Nested Query Against Different Index Mappings using ignore_unmapped

[ymao]

Related to this discussion: Nested query and ignore_unmapped option problem

Hi,

Using ES 5.6.1, I am trying to make a single query to multiple indices that have slightly different index mappings for a field. One index uses a nested mapping, the other does not. I am using the ignore_unmapped option in my nested query but am getting failures due to: "[nested] nested object under path [enrichments.urls] is not of nested type"

Here is a partial mapping for index1 (field is not nested):

{
    "top_level_field": {
        "properties": {
            "field": {
                "properties": {
                    "sub_field": {
                        "type": "keyword"
                    }
                }
            }
        }
    }
}

Here is a partial mapping for index2 (field is nested):

{
    "top_level_field": {
        "properties": {
            "field": {
                "type": "nested",
                "properties": {
                    "sub_field": {
                        "type": "text",
                        "analyzer": "custom_analyzer",
                        "fields": {
                            "analyzed1": {
                                "type": "text",
                                "analyzer": "custom_analyzer2"
                            },
                            "analyzed2": {
                                "type": "text",
                                "analyzer": "another_analyzer"
                            }
                        }
                    }
                }
            }
        }
    }
}

Here is a sample query targeting /index1,index2/_search

{
    "query": {
        "bool": {
            "should": [{
                "query_string": {
                    "fields": ["top_level_field.field.sub_field"],
                    "query": "pizza"
                }
            }, {
                "nested": {
                    "path": "top_level_field.field",
                    "ignore_unmapped": true,
                    "query": {
                        "query_string": {
                            "fields": ["top_level_field.field.sub_field", "top_level_field.field.sub_field.analyzed1", "top_level_field.field.sub_field.analyzed2"],
                            "query": "pizza"
                        }
                    }
                }
            }]
        }
    }
}

According to the docs, it seems like the ignore_unmapped option should allow this query to successfully execute against two fields in different indices with different mappings. Is that a correct assumption? Is there another way I should be structuring this query?

Thanks

Ying

[jlapp]

I am seeing the same behavior with 5.6.0. It seems like a bug, unless ignore_unmapped wasn't intended to handle this case. I have queries structured similarly to those provided by @ymao, although I've tried other variations with no luck.

I guess the bottom line is, should Elasticsearch allow me to query a field across two indices, where in one index the field is nested and in the other it is searchable but not nested?