How to use REGEXP_LIKE or REGEXP_INSTR in a query

Question

Hello All,
I would like to do a query on a column of a table to see if it has any combination of ALL of up to 5 words. So for example, if I search for (Apple, Banana, Blueberry), I would like to see the following data returned
Apples are better than Bananas and Blueberrys.
Blueberry recipes contain apples and bananas.
Bananas can be baked into bread with Apples but not Blueberrys.
So the criteria I would like to meet are
1. All three words are in the data returned
2. The three words can be in any order
3. There can be any or no other text in between the three words.
4. The query is case insensitive.
So far I have come up with this
    select * from hc_work_items where REGEXP_LIKE(wki_name, '(Apple)', 'i') AND REGEXP_LIKE(wki_name, '(Banana)', 'i') AND REGEXP_LIKE(wki_name, '(Blueberry)', 'i') This does the trick but I am wondering if it looks ok (I am new to REGEXP and also tuning queries for efficiency). I did also try
    select * from hc_work_items where REGEXP_INSTR(wki_name, '(Apples|Blueberrys|Bananas)') > 0 but this was returning only an OR selection of the words, not all three.
Thank you for any advice !
Edited by: 991003 on Feb 28, 2013 8:32 AM
Edited by: 991003 on Feb 28, 2013 8:34 AM

Answer 1

Hi,
Welcome to the forum!
991003 wrote:
Hello All,
I would like to do a query on a column of a table to see if it has any combination of ALL of up to 5 words. So for example, if I search for (Apple, Banana, Blueberry), I would like to see the following data returned
Apples are better than Bananas and Blueberrys.
Blueberry recipes contain apples and bananas.
Bananas can be baked into bread with Apples but not Blueberrys.It doesn't seem like you're really looking for words. In most of these cases, the text you are looking for (e.g. 'Apple') is not a separate word, but is a sub-string embedded in a longer word (e.g., 'Apple<b>s</b>').
What if someone uses the correct plural of 'Blueberry', that is, 'Blueberr<b>ies</b>? You might have to instruct your users to look for only the common part; in this case 'Blueberr'.
So the criteria I would like to meet are
1. All three words are in the data returned
2. The three words can be in any order
3. There can be any or no other text in between the three words.
4. The query is case insensitive.
So far I have come up with this
select * from hc_work_items where REGEXP_LIKE(wki_name, '(Apple)', 'i') AND REGEXP_LIKE(wki_name, '(Banana)', 'i') AND REGEXP_LIKE(wki_name, '(Blueberry)', 'i')
Yes, I think you'll have to do separate searches for each of the 3 targets.
Regular expressions might not be the most efficient way. INSTR or LIKE will probably be faster, e.g.
WHERE     UPPER (wki_name)   LIKE '%APPLE%'
AND     UPPER (wki_name)   LIKE '%BANANA%'
AND     UPPER (wki_name)   LIKE '%BLUEBERRY%'Oracle is smart enough to "short crcuit" compound conditions like this. For example, if if looks for 'Apple' first, and doesn't find it on a given row, then it doesn't waste time looking for the other targets on the same row.
This does the trick but I am wondering if it looks ok (I am new to REGEXP and also tuning queries for efficiency). I did also try
select * from hc_work_items where REGEXP_INSTR(wki_name, '(Apples|Blueberrys|Bananas)') > 0 but this was returning only an OR selection of the words, not all three.Exactly. You could look for any of the 6 possible permutations, but that's really ugly, inefficient, and unscalable. (If you ever need 4 targets, there are 24 permutations; with 5 targets there are 120.) You were better off the first time, with 3 separate conditions.
Oracle Text is a separate product that was designed for jobs like this. It's a separate product, that requires a separate license, and it has Text