Hi

I have an object with fields that I have marked as untokenised (due to requirements as the should not be tokenised). As I understand no tokenised fields are stored in the index as they are inputted. When I run a query such as email:email I get no result (luke tells me the same result). This is obviously because in the index the token is 'EMAIL'.

I use the same analyzer as i use for indexing. What is the recommended approach for dealing this situation where the user may type in a word which may be lowercase but in the index the token is held as uppercase?


Cheers
Amin

Hi,


Yes, un-tokenized fields are indexed as they are. This does not mean that they are stored in the index. Indexing and keeping the actual indexed values are two orthogoanl concepts.


Given that you index all properties un-tokenized you are actually not using any Analyzer at indexing time. So if you are using an analyzer at search time you are actually applying modifications to the search terms which are not applied during indexing.


Either manually upper-/lowercase the data and the search terms, or use a custom analyzer which will return the whole field value as single token. Chain this "tokenizer" with a LowerCaseFilter.

--Hardy

Hi

Thanks for your reply. After digging around I found that I can use KeywordAnalyzer for un-tokenised fields (Lucene In Action). So wrote the following code for an un-tokenised field while applying the standard analyzer for all other fields (tokenised fields).



I know that the standard analyzer deals with emails but the above just depicts an example of using the KeywordAnalyzer.

Cheers
Amin