Can't query an edge_ngram field in _all - elasticsearch

Question

So I'm setting up an index and I'd like to have a single search that would do a partial-word edge_ngram search for one field and a more normal search of the rest of the fields. From what I understand this should be easy to do by just matching on _all. However I just can't seem to make it work.
I have been able to get the desired results from a bool query that searches _all and the specific ngram field separately but that seems hackey and I'm guessing there's just something simple that I'm missing.
Here is just a minimal example to show what I'm doing and how it's not working for me.
Here is the index setup:
curl -XPUT "http://localhost:9200/test_index?pretty=true" -d'
{
"settings": {
"analysis": {
"filter": {
"edge_ngram_filter": {
"type": "edge_ngram",
"min_gram": 2,
"max_gram": 20
}
},
"analyzer": {
"edge_ngram_analyzer": {
"type": "custom",
"tokenizer": "standard",
"filter": [
"lowercase",
"edge_ngram_filter"
]
}
}
}
},
"mappings": {
"doc": {
"properties": {
"text_field": {
"type": "string",
"analyzer": "edge_ngram_analyzer",
"search_analyzer": "standard"
}
}
}
}
}'
And add a simple document:
curl -XPUT "http://localhost:9200/test_index/doc/1?pretty=true" -d'
{
"text_field": "Hello, World!"
}'
_all partial search doesn't work. It returns an empty result.
curl -XPOST "http://localhost:9200/test_index/_search?pretty=true" -d'
{
"query": {
"match": {
"_all": "hell"
}
}
}'
_all whole word search works though
curl -XPOST "http://localhost:9200/test_index/_search?pretty=true" -d'
{
"query": {
"match": {
"_all": "hello"
}
}
}'
And a partial search on the specific field works
curl -XPOST "http://localhost:9200/test_index/_search?pretty=true" -d'
{
"query": {
"match": {
"text_field": "hell"
}
}
}'
The term vector looks fine too
curl -XGET "http://localhost:9200/test_index/doc/1/_termvector?fields=text_field&pretty=true"
I really can't figure out what I'm doing wrong here. Any help would be appreciated.
Here are some details about my environment.
Elasticsearch version: Version: 2.3.3, Build: 218bdf1/2016-05-17T15:40:04Z, JVM: 1.8.0_92
Linux OS: Arch Linux
Kernel version: 4.4.3-1-custom

Answer 1

The _all field combines the original values of all fields as a string, not the terms produced for each field. So in your case, it doesn't contain the terms produced by the edge_ngram_analyzer, just the text from the text_field field. It's just like any other text field, you can specify analyzers for it, etc. In your example, it's using the default analyzer.