Elasticsearch: match exact keywords with special characters - elasticsearch

Question

I am storing tags as an array of keywords:
...
Tags: {
type: "keyword"
},
...
Resulting in arrays like this:
Tags: [
"windows",
"opengl",
"unicode",
"c++",
"c",
"cross-platform",
"makefile",
"emacs"
]
I thought that as I am using the keyword type I could easily do exact search terms, as it is not supposed to be using any analyser.
Apparently I was wrong! this gives me results:
body.query.bool.must.push({term: {"_all": "c"}}); # 38 results
But this doesn't:
body.query.bool.must.push({term: {"_all": "c++"}}); # 0 results
Although there are obviously instances of this tag, as seen above.
If I use body.query.bool.must.push({match: {"_all": search}}); instead (using match instead of term) then "c" and "c++" returns the exact same results, which is wrong as well.

Answer 1

The problem here is that you are using _all - Field, which uses an analyzer (standard by default). Make a small test with your data to be sure:
Test 1:
curl -X POST http://127.0.0.1:9200/script/test/_search \
-d '{
"query": {
"term" : { "_all": "c++"}
}
}'
Test 2:
curl -X POST http://127.0.0.1:9200/script/test/_search \
-d '{
"query": {
"term" : { "tags": "c++"}
}
}'
In my test second query returns documents, first not.
Do you really need to search with multiple fields? If so, you can override the default analyzer of _all field - for a quick test I put an index with settings like this:
{
"settings": {
"number_of_shards": 1,
"number_of_replicas": 0
},
"mappings": {
"test" : {
"_all" : {"type" : "string", "index" : "not_analyzed", "analyzer" : "keyword"},
"properties": {
"tags": {
"type": "keyword"
}
}
}
}
}
Or you can create Custom _all Field.
Solutions like Multi Field query, that allow to define list of fields to be searched over would rather behave like your example with body.query.bool.must.push({match: {"_all": search}});.