asked
Exclude file type from fulltext index
I have a fulltext index created on a varbinary(max) filestream column. I have both PDF's and XML's being stored in this filestream, and consequently both are being indexed. However, I really don't want to index the XML files, because I'm afraid that over time they'll bloat my index and slow it down. So what I would really like to do is only have my fulltext index on the PDF files instead. Is there a way to do this? Can I somehow disable the XML filter, perhaps? Thanks in advance for any help that I receive on this!
If anything is going to slow down your full text searches, that would be the iFilter for PDF, not XML. I believe that xml iFilter is pretty efficient specifically when compared to PDF. It is designed to ignore the xml markup, so you should not be worried about the bloating. PDF iFilter is also designed to ignore the (PDF) markup, but the horrendousness of the latter should be a bigger concern. Just my 2 cents.
@Oleg - thanks for your two cents. I had read on quite a few different websites (forums and such) that the performance of the Adobe PDF iFilter was abysmal, but I haven't really noticed it myself. I'm indexing in the vicinity of 2500 PDF's, with some of them being hundreds of megs in size, and the full-text search doesn't ever seem to take more than a second.
@Jagd Oh, I see. This actually makes sense. The possibility of a poor performance only affects the engine ***while*** it is indexing the data as it needs to consider a blob, feed it to the PDF IFilter which allows extraction of useful text (that would be a time comsumming part), index the text and store index data. Once this is done, the format of the original blob is irrelevant, the search queries are applied directly against the FT index. As a mattter of fact, if original PDF contains a bunch of big images not carrying any text information then these bytes are ignored and the volume of useful information can be considerably smaller than the size of the original PDF blob. This means that while the engine might have to work harder while indexing PDF rather than other document types, the final result is still the same, the actual FT search is just as fast :)
Seems that you cannot just by reading [BOL][1] You might want to isolate indexable data into its own table and create the fulltext index there instead. [1]:
http://technet.microsoft.com/en-us/library/ms187317.aspx
