Highlight not working for multi-fields if each subfield is not stored, "main subfield" stored, and _source disabled

recreation.sh

set -v
# Delete an eventual previous index
curl -XDELETE 'localhost:9200/bugindex'


# Create an index
curl -XPUT 'localhost:9200/bugindex' -d '{
	"settings" : {
		"index" : {
			"number_of_shards" : 1,
			"number_of_replicas" : 0
		}
	}
}'


# Define a type
curl -XPUT 'localhost:9200/bugindex/bugtype/_mapping' -d '{
	"bugtype" : {
		"_source" :	{ "enabled" : false },
		"_all" :	{ "enabled" : true },
		"properties" : {
			"_all" :		{ "type" : "string" ,		"index" : "analyzed" ,	"store" : "no" ,	"analyzer" : "whitespace"	},
			"in_all" :		{ "type" : "string" ,		"index" : "analyzed" ,	"store" : "yes" ,	"analyzer" : "keyword"		},
			"simple" :		{ "type" : "string" ,		"index" : "analyzed" ,	"store"	: "yes" ,	"analyzer" : "whitespace"	},
			"multi" :		{ "type" : "multi_field" ,
				"fields" : {
					"multi" :	{ "type" : "string" ,	"index" : "no" ,		"store" : "yes" ,	"include_in_all" : "no"	},
					"exact" :	{ "type" : "string" ,	"index" : "analyzed" ,	"store" : "no" ,	"analyzer" : "whitespace" ,	"term_vector" : "with_positions_offsets"	},
					"english" :	{ "type" : "string" ,	"index" : "analyzed" ,	"store" : "no" ,	"analyzer" : "english" ,	"term_vector" : "with_positions_offsets"	}
				}
			}
		}
	}
}'


# Index a single test doc
curl -XPUT 'localhost:9200/bugindex/bugtype/1' -d '{
	"in_all" : "key word 1",
	"simple" : "ab bc cd",
	"multi" : "out of the box"
}'


# Make it avaiable for search
curl -XPOST 'localhost:9200/bugindex/_optimize?refresh=true&flush=true'




# Works as intended
curl -XGET 'localhost:9200/bugindex/bugtype/_search?pretty=1&fields=*' -d '{
	"query" : {
		"term" : {
			"simple" : "bc"
		}
	},
	"highlight" : {
		"fields" : {
			"simple" : {
				"number_of_fragments" : 0
			}
		}
	}
}'


# Works as intended
curl -XGET 'localhost:9200/bugindex/bugtype/_search?pretty=1&fields=*' -d '{
	"query" : {
		"term" : {
			"in_all" : "key word 1"
		}
	},
	"highlight" : {
		"fields" : {
			"in_all" : {
				"number_of_fragments":0
			}
		}
	}
}'


# Search so so (depends on the following), highlight works only if :
#  - either _source is enabled
#  - or     multi.exact is stored
# Neither of the two is acceptable (index size is already a bit too large,
#  we have _source disabled and all fields stored,
#  as disk seeks are not so problematic with SSD)
curl -XGET 'localhost:9200/bugindex/bugtype/_search?pretty=1&fields=*' -d '{
	"query" : {
		"term" : {
			"multi.exact" : "box"
		}
	},
	"highlight" : {
		"fields" : {
			"multi.exact" : {
				"number_of_fragments":0
			}
		}
	}
}'


# Search ok, highlight INEXISTENT
# (searching against multi (same field as the highlighted one)
#  won't work because it's not indexed)
curl -XGET 'localhost:9200/bugindex/bugtype/_search?pretty=1&fields=*' -d '{
	"query" : {
		"term" : {
			"multi.exact" : "box"
		}
	},
	"highlight" : {
		"fields" : {
			"multi" : {
				"number_of_fragments":0
			}
		}
	}
}'




# It feels like we should:
#  - Use a multi-term to analyze a single field in different manners
#  - Index only the subfields
#  - Store only the "main" subfield,
#    as it would result in storing multiple times the exact same data otherwise
#  - Query against one subfield (not the "main" subfield)
# However with highlighting we would like to
#  - highlight against the multi-field (ie the "main" subfield)
#  - group highlights of searches against all the subfields
#    (I agree everybody may not whish this though)
# A solution would consist of either:
#  - Accesing the "main" field's value (either stored of from source)
#    when trying to access to the subfield's value.
#  - Not using multi-fields and using special analyzers that would
#    multiplex the terms from multiple sub-analyzers
# Implementing this last solution may be beneficial also for:
#  - Text search where you may wish to index exact, normalize and stemmed
#    terms in the same field index.
#  - Group analyzed terms from multiple fields into a kind of _all_terms field
#    which would leverage differenciated analyzers and rapidity of searching
#    against a single field index.