Multiple tokens in a synonym analyzer within a multiplexer

gistfile1.txt

curl -X DELETE "localhost:9200/companies"

curl -X PUT "localhost:9200/companies" -H 'Content-Type: application/json' -d'
{
  "settings": {
    "analysis": {
      "analyzer": {
        "company_name_analyzer": {
          "type":      "custom", 
          "tokenizer": "standard",
          "filter": ["company_name_filter"]
        }
      },
      "filter": {
        "company_name_filter": {
          "type": "multiplexer",
          "filters": ["company_name_abbreviations"]
        },
        "company_name_abbreviations": {
          "type": "synonym",
          "synonyms": ["co-op => cooperative", "corp => corporation"]
        }
      }
    }
  }
}'

# This will not work. Returns ["co op"]
curl -X GET "localhost:9200/companies/_analyze" -H 'Content-Type: application/json' -d'
{
  "analyzer": "company_name_analyzer",
  "text": "co-op"
}'

# This will work. Returns ["corporation"]
curl -X GET "localhost:9200/companies/_analyze" -H 'Content-Type: application/json' -d'
{
  "analyzer": "company_name_analyzer",
  "text": "corp"
}'

This is fixed by removing the "multiplexer" wrapper and directly setting "filters": "company_name_abbreviations" to the analyzer.

This example is distilled to show the bug. My full use case requires a multiplexer.

This is annoying, it's because SynonymGraphFilter can't handle incoming graphs yet, which is a lucene-level issue. I do plan on working on a fix for this at some point, but it's not a trivial problem unfortunately.

Bummer, thanks.

Apologies for bringing this up again. The nuanced problem above is blocking an ES 7 upgrade. I am unsure what you meant by the graph filter not handling incoming graphs, as my problem is exposing what this warning describes:

multi-word synonym token filters will not function normally when they are declared in the filters array because they read ahead internally which is unsupported by the multiplexer

Is this the same problem as not supporting an incoming graph?

Currently, my best option is to pre-process the text in my code.