Here are a few notes on DAO and ODBC with Oracle

I've been playing with Oracle's implementation for XQueries recently. Here are a few queries to demonstrate some features of XQueries. By no means are they complete, but they're meant to give a quick overwiev on some possibilities.

Summing up nodes in a hierachical query.

How to include files in SQL*Plus...

Recently, I had a company call me because they had a performance problem. After inquiring how that problem manifested itself, they told me that they needed to update an application with lots of data to a newer relase and that they used a "tool" that migrated that data. The estimates for this tool to complete the migration amounted to some 90 hours, which was too long, because it should be completed during a weekend. So, I went to see them to investigate that problem....

I have two tables that I want to join with a nested loop (instead of a hash join).

Although I usually write on Oracle here, this time, I have to use this channel because I am looking for some employees (or freelancers, although we prefer employees).

With PL/SQL, it's possible to create functions that return a nested table type of an object type. If the return value of such a function is given as the argument to the table() operator, it behaves like an ordinary table. I have already written about this in On emulating tables with nested tables and On emulating tables with PL/SQL (Part II). In these examples, the structure of the returned table (that is: the column names and their datatypes) is fixed. However, it is possible to create a function that returns a table whose data types and column names are constructed dynamically. This article tries to discuss this.

There are a few things in PL/SQL that I am really missing. One of these things is printf. I am not aware of any built-in functionality that remotely resembles the elegance of printf. Ok, lpad and rpad and to_char come to mind. But these functions are ,IMHO, tedious to use, especially when it comes to handling null values.

Therefore, I have hacked up my own printf.

In my last blog entry, I have written a little Excel macro. This macro opens a connection to an Oracle database, issues a select statement and then iterates over each record within the result set. For each record, it also iterates over each attribute within this record.

Now, there is CopyFromRecordset which does this iteration. I only have to specify the top left cell where I want my result set to appear.

I have been playing with Excel and its Visual Basic for Application (VBA) macros recently. As much as I loathe the syntax of VBA, I do like what can be done with it. Especially, I found it quite easy to connect to an Oracle Database with ADODB, execute a select statement and import the result into an Excel worksheet. Let me share my joy with you...

The as of clause in a select statement allows to see the data as of a past SCN or past point in time. So, it's possible, for example, to compare the current data in a table with a previous data set. This is what I try to do in this article.

Contexts allow to paramtrize a view. This article discusses how.

With the PL/SQL gateway, it's possible to invoke PL/SQL procedures via HTTP (or a browser, for that matter) and to return HTML pages to the browser. Since AJAX seems (or at least seemed) to be quite popular at the moment, I decided to find out if it is possible to create an AJAX web application with the PL/SQL gateway. To make the story short: it is. In this article, I am going to show how I did it. I used Oracle Express Edition 10 (Oracle XE) on Windows more or less out of the box.

Beware of select statemenst with a NOT IN where NULLs are part in the NOT IN clause.

With Oracle's regular expressions, it's relatively easy to split a string into it's words. Easy, that is, compared with a solution that doesn't use regexp_substr. It still requires some amount of setup as the following steps show...

SQL*Plus generates HTML output if it is invoked with the -m flag.

This article is going to demonstrate that.

I have recently stumbled upon something, which, after some thinking about, makes partially sense. The corpus delicti is a procedure that has an out parameter.

Until now, my blog wasn't a real blog because there was no possibility to leave a comment to a posting. I have changed that now: from now on, there will be such a possibility.

Oracle's Workspace Manager allows to travel back in time, so to speak, to see the data of that point in time to which one travelled. This is posslble because the Workspace Manager keeps different versions of data (See On creating workspaces and merging/refreshing data with Oracle's Workspace Manager).

In this article, I want to demonstrate how this is done.

On May 7th, I demonstrated how data can be versioned in workspaces and then be merged or refreshed. This worked well as long as I didn't change the same row in two different workspaces. However, when the same row is changed in two different workspaces, it creates what the Oracle documentation calls a conflict. Such conflicts cannot be resolved automatically because there is no way for Oracle to determine which version of the row is the version to keep. Manual conflict resolving is required in such cases. In this article, I am going to demonstrate that.

With Oracle's Workspace Manager it's possible to have several versions of data. That is, data can be changed, thus giving it a new version, without users looking at data of another version seeing the modified data. In this article, I try to explain this feature.

On January 8th, I have demonstrated how to emulate tables with nested tables in PL/SQL. Unfortunately, these emulated tables have a drawback: they can only store a number and a varchar2(15) per row. This article shows how to create truly dynamic tables, that is, the number of columns in the table can be dynamically controlled.

Reading a trace file is usually a three step process. First, the trace file is created with an SQL statement such as alter session set sql_trace=true. Then, the name of the trace file along with the directory that stores it, is located. Finally, the trace file is opened in an editor.

I usually don't do this sort of things, but this time, it seems real serious and I have to urge every dba to install the January 17th, 2006 critical patch update. The reason: Imperva has found a security bug that allows practically anyone to create a user with dba privileges (found via Pete Finnigan's website).

The bug boils down to the fact that SQL*Plus sends an alter session set nls... statement to Oracle when it initiates a connection. This statement is executed by Oracle with the rights of sys. Now, this statement can easily be replaced with another one. More details can be found at impervas link.

I verified the severity of this bug with two steps. In the first step, I intercepted the communication between SQL*Plus and the Oracle Server in order to find out how the alter session statement in question is transmitted. In the second step, I actually injected two SQL statements, the first of which creates a new user and the second of which grants dba to the newly created user.

In PL/SQL, tables can be emulated with nested tables. A nested table type is created with the create type type_name as table of dependand_type_name command. Basically, this command creates a type (here with the name type_name) that can store 0, 1 or more instances of dependand_type_name. To make theory a bit more clear, this is demonstrated in this article.

Oracle employs lazy evaluation for decode, but eager evaluation for nvl and nvl2. This article tries to prove this fact.

When autotrace is set within SQL*Plus, statistics about a DML statement will be shown when the statement has finished. This is quite a usable feature to compare different SQL statements regarding their performance.

However, I miss the ability to compare multiple SQL statements. statspack comes closer, but still is not what I want. Therefore, I have written a package to do what I need.

I like programming Perl a lot. When I first used Perl some ten years ago, I immediatly fell in love with this language because it was so easy to use regular expressions (I don't know much about today's trendy languages like Ruby or Python, but I guess, these languages make it equally simple to use regular expressions).

Now, Oracle 10g comes with four regular expressions supporting sql functions: regexp_instr, regexp_substr, regexp_replace and regexp_like. Yet, using them is not nearly as simple as using Perl's regular expressions. Specifically, I am missing an easy way to get the values matched by paranthesis. So, I had to write a package to do that for me.

Suppose, you have a table that stores some text along with a number:

create table elim_lowest_highest (
  txt varchar2(10),
  num number
);

Now, you want to select the average num grouped by txt from this table. However, you realize that the lowest and highest values per group (txt) are outliers and you want to eliminate those.

The row_number() analytic function comes in handy for this task.

In my last article, I demonstrated how to store hierarchical data and how to select from it. This article builds on that last article and shows how triggers can be used to keep summary data in the nodes of hierarchical data in sync.

The canonical approach to store hierachical data is to create a self referencing table...

create table hierarchy_tbl (
  id        number primary key,
  parent_id references hierarchy_tbl,
  data      varchar2(30)
);

...and then to use Oracle's start with ... connnect by idiom to select from it.

Unfortunatly, this approach causes a problem if a subtree can be assigned to more than one node. For example, let's assume you sell three products: prod_1, prod_2 and prod_3. In order to produce these products, you buy things (I'll call these things items) from other vendors. These items are then assembled into compositions. These compositions can then be assembled into further compositions. Such a composition might not only consist of other compositions but also of other items. Finally, the compositions (possibly again along with items) are then assembled into a product.

On September 25, 2005, I presented desc_table which is a package that can be used to extract various information about a table or view, such as column names, their types, primary key and foreign key constraints, as well as comments made on the table and its columns.

In this article, I try to show how this package can be used to create Entity Relationship Diagrams (ERD) along with neato. neato is a open source software that can be used to create graphs (directed or undirected).

I am a bit disappointed with describe. When I use it to describe a table (or view), it only reports column names, their types and if they have not null constraints. But I also want it to print the comments on a table and its columns. And, I'd appreciate if it told me which columns make a primary key and which tables are referenced by their foreign keys. It would be nice if describe told me, if the described thing is a table or a view. Lastly, I want to know the owner of the table (or view).

Because describe lacks all those little -but imho useful- features, I decided to write my own. It consists of two parts: a package and an SQL script. The script is basically a wrapper called from SQL*Plus and wrapps the call of the procedure in the package.

Oracle doesn't have a built-in aggregate function that multiplies its arguments (like sum adds expressions).

However, using a well known mathematical property, it is possible to emulate such an aggregate function.

Say, you encounter the following SQL statement:

             select * from some_table;
             

Without knowing further details, you cannot be totally sure that there is actually a table named some_table. This is the case if there is a synonym named some_table that actually points to a table with a different name.

If you want to find out which table and schema some_table stands for, dbms_utility.name_resolve helps do that.

• Each row has no duplicate values in its cells
• Each column has no duplicate values in its cells
• The grid is split into 9 non-overlapping 3x3 blocks. These blocks also have no duplicate values.

PL/SQL does not have function pointers like for example C. In C, function pointers are used to (sort of) pass an arbitrary function to an algorithm of any kind. The algorithm can then call the passed function without knowing what this function does.

This article tries to show a way how to emulate such a behaviour

Oracle has a feature called Change Data Capture (CDC) which allows to track changes (coming from inserts, updates, deletes) in a table or in multiple tables.

This article demonstrates a tiny bit of CDC