Default fuzzy transpositions to true

[grmblfrz]

The documentation states:

0..1:: must match exactly
1..4:: one edit allowed
>4:: two edits allowed

whereas the source contains:

final int len = termLen(text);
if (len <= 2) {
    return 0;
} else if (len > 5) {
    return 2;
} else {
    return 1;
}

For backwards compatibility the documentation should be adapted to:

0..1:: must match exactly
1..5:: one edit allowed
>5:: two edits allowed

(though I personally would prefer to change the implementation - then 'Flyod' would match 'Floyd' with fuzziness 'AUTO', now I'm needing fuzziness 2)

[clintongormley]

Hi @grmblfrz

Thanks for spotting this.

(though I personally would prefer to change the implementation - then 'Flyod' would match 'Floyd' with fuzziness 'AUTO', now I'm needing fuzziness 2)

Actually, you can set fuzzy_transpositions to true, which will mean that swapping yo is a single step:

DELETE t

PUT /t/t/1
{
  "text": "Floyd"
}

GET /_search
{
  "query": {
    "match": {
      "text": {
        "query": "flyod",
        "fuzziness": 1,
        "fuzzy_transpositions": true
      }
    }
  }
}

So I wonder if the right thing is to change the cutoff from 5 to 4, or to default to fuzzy transpositions?

[clintongormley]

@mikemccand any thoughts on this?

[mikemccand]

+1 to default to fuzzy_transpositions = true.

A cutoff for 4 to allow 2 edits seems a bit too eager...

[clintongormley]

I've updated the docs to correct the string lengths, and changed this issue to set fuzzy transpositions to true by default

[grmblfrz]

Please document the fuzzy_transpositions parameter, I did not know about it before and could not find anything in the documentation.

[devinrsmith]

I also don't see documentation around "fuzzy_transpositions", nor do I see the ability to set it in the java client code.