Understanding an "explain"

Nick_Hoffman · October 13, 2011, 5:01pm

Hey guys. Is there anything that I can read to help me understand the output
from an "explain"?

One of my searches is returning a document that I don't think should be
returned, and I only understand part of the explanation:

gist.github.com

https://gist.github.com/nickhoffman/1284790

gistfile1.js

// curl -X GET 'localhost:9200/development_products/_search?pretty=true' -d '
{
  explain: true,
  query: {
    dis_max: {
      queries: [
        { field: { "name":                  "grmlock" } },
        { field: { "catalog.name":          "grmlock" } },
        { field: { "items.name":            "grmlock" } },
        { field: { "properties.character":  "grmlock" } },

This file has been truncated. show original

Line 47 says that the field "items.name" matches the search. But what does "ock
in 198" mean?

kimchy · October 14, 2011, 12:21pm

It means that the (part of the) query, which is items.name:ock matched a doc
number 198 (internal Lucene numbering of docs). What confuses you? the part
that its ock? Are you doing any special analysis on that field (items.name)?

On Thu, Oct 13, 2011 at 7:01 PM, Nick Hoffman nick@deadorange.com wrote:

Hey guys. Is there anything that I can read to help me understand the
output from an "explain"?

One of my searches is returning a document that I don't think should be
returned, and I only understand part of the explanation:
gist:1284790 · GitHub

Line 47 says that the field "items.name" matches the search. But what does
"ock in 198" mean?

Nick_Hoffman · October 14, 2011, 1:23pm

Ah, gotcha re: "in 198" and "(stuff in parens)".

I understand the "ock": it's an nGram that matched.

A custom nGram analyzer named "3_8_ngram" is being used for index and search
analysis:

gist.github.com

https://gist.github.com/nickhoffman/6e935e022c759fbc8d61

1-settings.js

// curl localhost:9200/development_products/_settings?pretty=true
{ 
  "development_products" : {
    "settings" : {
      "index.analysis.filter.3_8_ngram.max_gram": "8",
      "index.analysis.filter.3_8_ngram.type":     "nGram",
      "index.analysis.filter.3_8_ngram.min_gram": "3",

      "index.analysis.analyzer.ascii_std.type":       "custom",
      "index.analysis.analyzer.ascii_std.tokenizer":  "standard",

This file has been truncated. show original

2-mapping.js

// curl localhost:9200/development_products/_mapping?pretty=true
{
  "development_products" : {
    "product" : {
      "properties" : {

        "id" : {
          "index" : "not_analyzed",
          "type" : "string"
        },

This file has been truncated. show original

The rationale for "3_8_ngram" is that my users will search for misspellings
or partial words, and I need to match those. Is there a better approach for
solving that problem, though?

Thanks again!
Nick