Detailed boolean search engine methodology help

ppowell

My image cataloging application project is now being tasked to come up with a simple/advanced search engine, fully custom (no 3rd party). The criteria to search for will include up to 15 different fields, all conjoined by the boolean "AND", "OR" and "AND NOT" in between them all.

Based on this incredibly simple criteria, how would any of you approach this, seriously? What would you think would be the necessary steps I should take to approach, logically. Client has enormous cash flow and is willing spend wantonly, however, it is made clear to me the client will NOT accept any 3rd party source of any kind within the application, every single thing must be built by me, by hand. Call it the DIY application approach if you must, but that's it.

Everything else is custom-built by me and I am about 40% completed on the search engine, however, I am running into serious methodology problems because of the complexity of the means by which a search can be done, consider the enormity of combinations of boolean conjoinings that can take place.

If any of you ever approached anything similar to this, hopefully under the same conditions as I, how have you approached it?

Thanx
Phil

AstroTeg

Simplest thing to do is "divide and conquer".

I usually start out looking at the problem of developing a search query.

Then I look at what I need to search on.

Then I look at the search parameters. What does the user want to search on and how? Which tables need to be involved to make a particular search work?

From there I start building with the idea of keeping the query modular. A query has the SELECT clause (which fields you want the query to return), the FROM clause, the WHERE clause, and if you need to get fancy, the GROUP and ORDER clauses. Its a matter of building all your clauses and then assembling them before running your query. I find string concatenation to be a handy tool.

So, you'd probably have a search form. The user submits it. You'll have code to validate the input to make sure its groovy. Then you'll need to look at which search items the user has given you and then build your query as needed. But based off those elements the user gave you, you'll need to customize and build the query clauses. Sometimes switch() functions will do. Other times you may need to use if() or a combination of the two.

But I'd recommend starting off small. Start with a just testing 1 search parameter and make sure that works. Then working on the 2nd. Make sure it works. Make sure the 1st parameter still works with the 2nd. If this is working, figure out the pattern you used to make it work and build the rest of the parameters into your query.

mtimdog

2 suggestions from prior experience.

a drop-down on the search page for and, or, exact phrase.
a switch statement to make an array out of given values (checked to search....or just have an array if searching all 15 all the time).

foreach loop through array and dynamically make the query. Echo out the query. Query the query. Echo out the Error. Fix Error and restart the process.

ppowell

Thanx.. in my case I will have 1 - X amount of dropdowns of "AND", "OR", "AND NOT" boolean word phrases to conjoin with any number of elements, so the algorithm validation alone is extremely difficult.

Thanx!
Phil