I did a few full text search experiments. As a baseline I used the full text of the
ICD-10.
Searching a random words, or a word at the very start or end - as long as longer as 5
characters - with a simple Boyer-Moore command line utility, was always"instant" - no
perceivable delay.
That makes me wonder whether static text like this is better dealt with in terms of
preserving ressources by just storing the whole thing as a blob and writing a short
stored procedure implementing the Boyer-Moore. Would debulk our database backbone a
lot.
As we will only use one coding base and just map all the others to it, we might as
well just store all of them with the exception of the master coding base in a blob,
degine rules how to extract the code, and do the searching un-indexed with Boyer-Moore
full text. Would free some MB Ram, and I can't see any problems with referential
integrity etc.
Are there any known drawbacks? Has anybody solved this in a similar manner before?
Horst
