[
https://issues.apache.org/jira/browse/LUCENE-9365?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17104233#comment-17104233
]
Mark Harwood commented on LUCENE-9365:
--------------------------------------
bq. Maybe we should disallow prefix == term.text().length() for FuzzyQuery? It
is sort of strange to use FuzzyQuery in this way
Following that logic it's probably a lower threshold than that -
{code:java}
prefixLength + editDistance == term.text().length()
{code}
...would be "mis-use" too. e.g. given term `abcd` and prefix=2, edit=2 the
characters `cd` are wholly redundant offerings and no more informative to the
search than providing the search term `ab` with the same parameters.
This probably comes down to how user-friendly we want to be. I would have
leaned towards a more lenient approach - the Query objects the user creates
capture the high-level intent and the rewrite phase is the opportunity for that
Query class to figure out if a term, prefix-with-fixed-length or FuzzyQuery is
the preferred mode of execution.
> Fuzzy query has a false negative when prefix length == search term length
> --------------------------------------------------------------------------
>
> Key: LUCENE-9365
> URL: https://issues.apache.org/jira/browse/LUCENE-9365
> Project: Lucene - Core
> Issue Type: Bug
> Components: core/query/scoring
> Reporter: Mark Harwood
> Priority: Major
>
> When using FuzzyQuery the search string `bba` does not match doc value `bbab`
> with an edit distance of 1 and prefix length of 3.
> In FuzzyQuery an automaton is created for the "suffix" part of the search
> string which in this case is an empty string.
> In this scenario maybe the FuzzyQuery should rewrite to a WildcardQuery of
> the following form :
> {code:java}
> searchString + "?"
> {code}
> .. where there's an appropriate number of ? characters according to the edit
> distance.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
