mankowitz
asked on
Best way to handle SETS in SQL SERVER 2008 R2
I have two tables. One represents a person, and another is a one-to-many child table that has certain characteristics of that person. Each person can have 0, 1 or many characteristics.
So here is the question. I have a web page where people update the characteristics and the results are submitted in a form using a multi-select. When users make changes, I delete all the old values and replace them with the new ones, like this
This strikes me as painfully inefficient. Does SQL Server have any other way to manage a set of boolean values? I understand that if I was truly enterprising, I could assign each characteristic a power of two and then store a sum, but I am looking for an extensible (i.e. easy) way of doing this.
Here's a fiddle http://sqlfiddle.com/#!3/df792/1
create table people (id int, name varchar(25));
insert people (id, name) VALUES (1, 'Scott'), (2, 'Joe');
create table characteristics (characteristic int, characteristic_text varchar(25));
insert characteristics (characteristic, characteristic_text) VALUES
(1, 'Fat'), (2, 'Skinny'), (3, 'Ugly'), (4, 'Toothsome');
create table people_characteristics (person int, characteristic int);
insert people_characteristics (person, characteristic) VALUES
(1, 1), (1, 3), (2, 4);
So here is the question. I have a web page where people update the characteristics and the results are submitted in a form using a multi-select. When users make changes, I delete all the old values and replace them with the new ones, like this
delete from people_characteristics where person=1;
insert people_characteristics (person, characteristic) VALUES
(1, 1), (1, 2);
This strikes me as painfully inefficient. Does SQL Server have any other way to manage a set of boolean values? I understand that if I was truly enterprising, I could assign each characteristic a power of two and then store a sum, but I am looking for an extensible (i.e. easy) way of doing this.
Here's a fiddle http://sqlfiddle.com/#!3/df792/1
It might make more sense to use updatable recordset objects in your front-end, and then just update, instead of an DELETE-INSERT.
What front-end are you using, and I recommend adding that zone to this question.
>Does SQL Server have any other way to manage a set of boolean values?
Nothing optimized for boolean data types.
What front-end are you using, and I recommend adding that zone to this question.
>Does SQL Server have any other way to manage a set of boolean values?
Nothing optimized for boolean data types.
ASKER
for the time being, the scope of the project is pretty small. It includes the editing of various characteristics (as above) and running a monthly report showing the aggregate number of each of the characteristics in the community. In other words, how many fat people, how many ugly people for each month.
The front end is PHP, but I'm not sure that helps. For example, If I stored an object in the master table, I could accomplish my changes with a single update
UPDATE people SET characteristics='3,6,22,66 ' WHERE id=3
The problem with that is that my monthly query would be slower - something like this:
SELECT ....
WHERE characteristics like '%'+c.characteristic+'%'
FROM people p CROSS JOIN characteristics c
The front end is PHP, but I'm not sure that helps. For example, If I stored an object in the master table, I could accomplish my changes with a single update
UPDATE people SET characteristics='3,6,22,66
The problem with that is that my monthly query would be slower - something like this:
SELECT ....
WHERE characteristics like '%'+c.characteristic+'%'
FROM people p CROSS JOIN characteristics c
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
I was hoping for an easy analog of MySQL's SET type, but no luck.
any other method is not really normalized, but then, it might not be the most efficient in regards to queries ...
now, in your "update" you "delete all + insert all" ... you could consider to delete only those that are not set anymore, and insert new ones.
however, in terms of efficiency, if you have that table in sql server with a clustered index on the personid, this will touch 1 page block at max, and hence be the most efficient solution for that part.
for the queries (to find all persons matching a certain "profile", you might consider adding a denormalized version of those settings on the person table, using for example some XML field (which could be indexed eventually)
it depends on the usage of those "settings/properties"