Here's a sticky problem. How do you build a query that gives you distinct records from one table based on multiple records from another table and order by a date found in the second table. In my real world example, I have project tracks in "Items" and comments or notes in "events". I want a distinct list of "items" that have been updated in the past 10 days. That seems easy right? Well... not as easy as it seems on the surface.Read More
You probably already know about the GROUP BY clause in SQL - but have you ever tried "HAVING"? The group by clause is great for getting aggregate groups of information together. Let's say you want a count of how many times a keyword appears in a table. You have the words car, auto, van and bus all in a keywords table.Read More
Here's a rule of thumb for you. When you are writing a query that includes a sub query always use an alias for the table and qualify the columns - or use the full column name if you are so inclined. If you don't you may very well end up with unexpected results. Consider this query:Read More
SQL provides many ways of grouping data. It also has many ways to select data. But if you've ever tried to remove duplicates from a database you might have needed 2 or 3 trips to the liquor store to figure it out. I've seen routines that match and compare and order and update and delete willy nilly - all just to find and fix duplicate rows in a database. I've never had to do this to any of my databases of course, because all my constraints and applications are perfect (gah!). Still, for those of you with imperfect databases I'm happy to report there is a magic bullet. Here's the big secret.Read More
About every 2 months or so someone asks about reliably returning the primary key record from an insert query. The problem they are trying to solve usually has to do with additional inserts into related tables. For example, if you are adding a new user and you want to set group permissions as well - but group permissions require inserting into another table. One way to do it is to do an insert, then do a second query that pulls back the "max(id)" and uses another qualifier - like an email address. This requires 2 connections to the database, but it is a very common method. If you are using SQL server and your primary key is an "identity" type field then you have another option. You can insert and get back the identity value in the same query. This is preferred because of SQL treats a single query statement as an implicit transaction - meaning you are assured of data integrity, and that you will return the right value. Here's the way it works.Read More
First let me say that it is usually not a good idea to store lists in Database tables. Many novice web developers falls into the trap of treating the data in the database the way they treat it in their web application. They are used to using lists and list functions in logic loops to qualify users or statements, and they try to do the same thing in the Database. But SQL doesn't have any native list functionality. The equivalent of a "list" to a database is a table with a single row.
A while back I saw a security scheme implemented by storing a list of group ids in a character column of the users table. The code grabbed the groupIds column from the users table on login, then it selected from the "groups" table where group_id IN (1,2,4,8...). This actually works fine when dealing with individual users. But what if you wanted to grab all the users from a particular group? You end up with code that looks like "where groupIds LIKE '%,2,%' Or groupIds like '2,%' or groupIds LIKE '%,2' or groupIs = '2'. Why the 3 statements? Because you have to account for a group id of 20 and a case where 2 is alone or at the beginning or the end. There are other ways around this - none of them pretty (like storing beginning and ending comma's for example). Not to mention this sort of design makes DBA's pull out what little hair they have.
Now, having warned you against being "listy" in your database design, I'm going to show you one way of working with a list that might help in a case like that above. I know, I know, a little knowledge is a dangerous thing. Let's just assume you are forced to deal with a legacy schema that you have no power to change (ha).Read More
Are you using Union queries yet? If not, you should get up to speed. A union query is an extremely useful method for returning records from different tables in the same recordset. You just have to remember that datatypes of the columns must match in the same order they are referenced. Here's an example:Read More
Ever wonder why a table is only allowed one "clustered" index? It's because the clustered index is the actual sort order that of the table. When you implement a clustered index you are "rearranging" the rows so that they are actually ordered by that index. That makes this index faster. Choosing the right clustered index can have a major impact on the speed of searches in your table. Wait a minute - isn't the Primary Key always the clustered index? Well... yes, when you use enterprise manager to set a primary key you will find that it also designates that column or columns as the default clustered index. It doesn't have to be that way however.Read More