Add the limit on the number of expanded terms

[mayya-sharipova]

e.g wildcards, fuzzy queries

Relates to #11511

[mayya-sharipova]

This was one of the items in the Elasticsearch roadmap.

But looks like we may not need an extra limit, as fuzzy queries and wild card queries use one of the multi term query rewrite methods that already safeguard against too much expansion. For example:

• constant_score - tries disjunction of boolean clauses; and if there are too many of them builds a bitset of matching docs
• scoring_boolean - will hit too many clauses failure if it exceeds the boolean query limit (1024), and produce the following error:

"error":{  
   "root_cause":[  
      {  
         "type":"too_many_clauses",
         "reason":"maxClauseCount is set to 1024"
      }
   ]

• constant_score_boolean - similar to scoring_boolean will hit too many clauses failure if it exceeds the boolean query limit (1024).
• top_terms_N - the N controls the size of the top scoring terms to use.
• top_terms_boost_N - the N controls the size of the top scoring terms to use.
• top_terms_blended_freqs_N - the N controls the size of the top scoring terms to use.

[mayya-sharipova]

As discussed in the FixitFriday, we still will have a limit on the number of expanded terms equal to 1024, as all other Lucene checks are done later.

[mayya-sharipova]

After further investigation, looks like this feature is NOT necessary, as there are already safeguards on the Lucene level.

For example, in the case of the wildquery, the algorithm is following:

1. Collect all the terms for the field
2. Compute the intersection of the WildQuery automaton with the field terms
3. Rewrite query:
   • constant_score -> MultiTermQueryConstantScoreWrapper attempts to rewrite as a boolean query, but changes to the bitset of docs, if too many terms
   • scoring_boolean -> ScoringRewrite:rewrite before adding boolean clauses tries to collect terms in TermCollector; fails as soon as No. of terms exceeds 1024
   • constant_score_boolean -> ScoringRewrite, similar to scoring_boolean
   • top_terms_N and other top terms methods -> TopTermsRewrite uses minimum between N and 1024 for rewriting the query