How to add parameters to a scripted similarity?

[fsanzestevez]

I'm trying to make a parametrized scripted similarity function but I can't get it to work.
I have created the script and have it stored.

However, when I try to implement it in the similarity score, I keep getting errors because of the symbol [params].

^ I know my script looks very similar to BM25, but I need to parametrized certain values during the query.

{
  "error" : {
    "root_cause" : [
      {
        "type" : "script_exception",
        "reason" : "compile error",
        "script_stack" : [
          "... f = doc.freq/(doc.freq + params['k1'] * (1 - param ...",
          "                             ^---- HERE"
        ],
        "script" : "double boost = 2.2 * doc.freq; double tf = doc.freq/(doc.freq + params['k1'] * (1 - params['b'] + params['b'] * field('dl_avgdl'))); double idf = Math.log(1 + ((params['docs_count'] - term.docfreq + 0.5)/(term.docFreq + 0.5))); return query.boost * tf * idf * norm;",
        "lang" : "painless",
        "position" : {
          "offset" : 64,
          "start" : 39,
          "end" : 89
        }
      }
    ],
    "type" : "script_exception",
    "reason" : "compile error",
    "script_stack" : [
      "... f = doc.freq/(doc.freq + params['k1'] * (1 - param ...",
      "                             ^---- HERE"
    ],
    "script" : "double boost = 2.2 * doc.freq; double tf = doc.freq/(doc.freq + params['k1'] * (1 - params['b'] + params['b'] * field('dl_avgdl'))); double idf = Math.log(1 + ((params['docs_count'] - term.docfreq + 0.5)/(term.docFreq + 0.5))); return query.boost * tf * idf * norm;",
    "lang" : "painless",
    "position" : {
      "offset" : 64,
      "start" : 39,
      "end" : 89
    },
    "caused_by" : {
      "type" : "illegal_argument_exception",
      "reason" : "cannot resolve symbol [params]"
    }
  },
  "status" : 400
}

Is this possible to do with Elasticsearch 8.1?

[stu (Stuart Tettemer)]

Hi @fsanzestevez, unfortunately params is not available for similarity scripts. So the script will need to embed the constants.

See the similarity context docs for all the variables available to similarity scripts.