2

By default, a multiple word search is split into files that have each word separately. How can I override this default and have whoosh search for an exact match? Even though it's most likely supported, I can't find in google/whoosh documentation.

In addition, would searching for an exact match have better or worse performance than the same multi-word search?

maged
  • 859
  • 10
  • 24

1 Answers1

1

Use phrase queries (that is, put double quotes around the words which you want to be matched in your query), e.g. :

"to be or not to be"

However, this only works if the field you're searching in is of type whoosh.fields.TEXT.

As for the performance thing, phrase searches are necessarily slower than "classic search". To do a phrase search, it is first necessary to retrieve all the documents that contain all the terms that you specified in your query (this is the "classic search" part), and then to compare the terms positions between you query and the document to check if it looks like a match.

michaelmeyer
  • 7,985
  • 7
  • 30
  • 36
  • Wouldn't a better implementation be to find all the documents containing the first (or longest) word and then check the next n directly following (or preceding) words? – maged Jun 24 '13 at 21:47
  • Actually, I see why your description would be faster in most cases, and mine only in exceptions. – maged Jun 24 '13 at 21:49
  • 1
    @maged: I haven't checked Whoosh code in detail, but it is probable that, when you issue a "phrase query", it already does the job appropriately (as you described). – michaelmeyer Jun 25 '13 at 00:57