MySql Order By Value equivalent in ElasticSearch 5.6 - elasticsearch

Question

ElasticSearch Version: 5.6
I have imported MySQL data in ElasticSearch and I have added mapping to the elastic search as required. Following is one mapping for the column application_status.
Mappings:
{
"settings": {
"analysis": {
"analyzer": {
"case_insensitive": {
"type": "custom",
"tokenizer": "keyword",
"filter": ["lowercase"]
}
}
}
},
"mappings": {
"lead": {
"properties": {
"application_status": {
"type": "string",
"analyzer": "case_insensitive",
"fields": {
"keyword": {
"type": "keyword"
}
}
}
}
}
}}
On the above mapping, I am able to do simple sorting (asc or desc) using following query:
{
"size": 50,
"from": 0,
"sort": [{
"application_status.keyword": {
"order": "asc"
}
}]}
which is MySql equivalent of
select * from <table_name> order by application_status asc limit 50;
Need help on following problem:
I have MySQL query which sorts based on application_status:
select * from vLoan_application_grid order by CASE WHEN application_status = "IP_QUAL_REASSI" THEN application_status END desc, CASE WHEN application_status = "IP_COMPLE" THEN application_status END desc, CASE WHEN application_status LIKE "IP_FRESH%" THEN application_status END desc, CASE WHEN application_status LIKE "IP_%" THEN application_status END desc
Please help me write the same query in ElasticSearch. I am not able to find order by value equivalent for strings in ElasticSearch. Searching online, I understood that, I should use sorting scripts but not able to find any proper documentation.
I have following query which just does simple sort.
{
"size": 500,
"from": 0,
"query" : {
"match_all": {}
},
"sort": {
"_script": {
"type": "string",
"script": {
"source": "doc['application_status.keyword'].value",
"params": {
"factor": ["IP_QUAL_REASS", "IP_COMPLE"]
}
},
"order": "desc"
}
}}
In the above query, I am not using params section as I am not aware how to use it for type: string
I believe I am asking too much. Please help or any relevant documentation links would be greatly appreciated. Hope question is clear. I'll provide more details if necessary.

Answer 1

You have two options:
the most performant one is to index at indexing time another field that should be a number. This number (your choice) will be the numerical representation of that status. Then at search time, you simply sort by that number and not by the status
at search time use a script that will do almost the same thing as the first option, but dynamically, and less performant (but still quite fast)
Below you have the second option:
"sort": {
"_script": {
"type": "number",
"script": {
"source": "if (params.factor[0].containsKey(doc['application_status.keyword'].value)) return params.factor[0].get(doc['application_status.keyword'].value); else return 1000;",
"params": {
"factor": [{
"IP_QUAL_REASS":1,
"IP_COMPLE":2,
"whatever":3
}
]
}
},
"order": "asc"
}
}
If you also want things like LIKE WHATEVER%, my suggestion is to consider an indexing time change, rather than search time because the script gets more complex. But, this is the one for wildcard matches as well:
"sort": {
"_script": {
"type": "number",
"script": {
"source": "if (params.factor[0].containsKey(doc['application_status.keyword'].value)) return params.factor[0].get(doc['application_status.keyword'].value); else { params.wildcard_factors[0].entrySet().stream().filter(kv -> doc['application_status.keyword'].value.startsWith(kv.getKey())).map(Map.Entry::getValue).findFirst().orElse(1000)}",
"params": {
"factor": [
{
"IP_QUAL_REASS": 1,
"IP_COMPLE": 2,
"whatever": 3
}
],
"wildcard_factors": [
{
"REJ_": 66
}
]
}
},
"order": "asc"
}
}