While working on the Not So Extreme Makeover: Community Edition site, I came up with an algorithm that simplifies anything else I've ever written to deal with this condition. I'll set the scenario, explain the algorithm, share how I implemented it in PHP, and provide a modification if the scenario is a bit more complicated.

Scenario - You have two parent tables, and a child table with a many-to-one relationship with both parent tables, used to map entries in the two parent tables to each other. For this example, we'll use these three tables…

create table volunteer (
    vol_id  integer  not null,
    vol_last_name  varchar(50)  not null,
    ...etc...
  primary key (vol_id)
);

create table r_volunteer_area (
    rva_id  integer  not null,
    rva_description  varchar(255)  not null,
  primary key (rva_id)
);

create table volunteer_area (
    va_volunteer_id  integer  not null,
    va_area_id  integer  not null,
  primary key (va_volunteer_id, va_area_id),
  foreign key (va_volunteer_id) references volunteer (vol_id),
  foreign key (va_area_id) references r_volunteer_area (rva_id)
);

Algorithm - The three-step algorithm is as follows…

1. Create a comma-delimited string of IDs for the child table.
2. Delete the IDs from the child table that are not in the list.
3. Insert the IDs into the child table that are not there already.

Implementation - In PHP, if you have an array, it's easy to come up with comma-delimited list. To get an array of values back in a post, define your fields with “[]” after the name…

<input type="checkbox" name="area[]" id="chkArea1" value="1" />
<label for="chkArea1">Do Something</label><br />
<input type="checkbox" name="area[]" id="chkArea7" value="7" />
<label for="chkArea7">Do Something Else</label>

Here's the PHP code, using PHP Data Objects (PDO) as the database interface, behind a helper class that creates the statement, appends the parameters, and executes it. (The “quoting” escapes the statement to avoid potential SQL injection attacks - putting it in its own class would make the implementation here much cleaner.)

/**
 * STEP 1
 *    Create a comma-delimited list of IDs.
 */

// Quote will return the string as '2,3,4' - since we're using this
// as an IN clause of integers, we'll strip the quotes off.
$sAreas = $pdo->quote(join(",", $_POST["area"]));
$sAreas = substr($sAreas, 1, strlen($sAreas) - 1);

// Quote the volunteer ID.
$iVol = $pdo->quote($_POST["vol"], PDO::PARAM_INT);

/**
 * STEP 2
 *    Delete the IDs that are no longer in the list.
 */
$dbService->executeCommand(
    "DELETE FROM volunteer_area
    WHERE   va_volunteer_id = ?
        AND va_area_id NOT IN ($sAreas)",
    array($iVol);

/**
 * STEP 3
 *    Insert the IDs that are not yet in the list.
 */
$dbService->executeCommand(
    "INSERT INTO volunteer_area
        SELECT $iVol, rva_id
        FROM r_volunteer_area
        WHERE   rva_id IN ($sAreas)
            AND rva_id NOT IN
            (SELECT va_area_id
            FROM volunteer_area
            WHERE va_volunteer_id = ?)",
    array($iVol));

Modification - Suppose that now you accepted comments along with each of the checkboxes, so a simple two-integer insert/delete is no longer sufficient. You would still only need to break step 3 into two steps.

1. Get a list of IDs to update.
2. For each ID in the posted list
   1. If the ID exists in the update list, update it.
   2. Otherwise, insert it.

The implementation would then be able to use this list to make the decision without hitting the database every time.

// Assume this returns an associative array of IDs.
$aUpdates = $dbService->performSelect(
    "SELECT va_area_id
    FROM volunteer_area
    WHERE   va_volunteer_id = ?
        AND va_area_id IN ($sAreas)",
    array($iVol));

foreach($_POST["area"] as $iArea) {
    if (in_array($iArea, $aUpdates)) {
        // Update the table
        ...etc...
    }
    else {
        // Insert into the table
        ...etc...
    }
}

I think you'll agree that this is much better than spinning through a loop, doing a count on each ID to see if it exists, then either doing an update or an insert based on the count. And, while the implementation here is PHP, it could easily be implemented in any language that supports arrays and database access.

There are lots of “how to” articles on sharing data between Oracle and SQL Server. Most of these involve installing Oracle's code base on the SQL Server machine, then using that instance to link tables within Oracle. This technique does not require that, thanks to a product from Oracle called Oracle Instant Client.

To set up the Oracle piece, download the packages for “Basic” and “ODBC Supplement”, and follow the instructions for installation, on the machine with SQL Server. (This is not an “install” per se - it's basically an unzip.) Next, you'll need to provide a TNSNAMES.ORA file - this can be any valid file, including a simple shell with an “ifile=” statement pointing to a common TNSNAMES.ORA file. Finally, set the environment variable TNS_ADMIN to point to the directory where this TNSNAMES.ORA file resides.

Now, you can easily create a DTS script through SQL Server to push or pull data however you'd like. Oracle Instant Client will appear in the drop-down list of providers, and you'll be able to specify your connection the way you normally do (i.e., “DB01.WORLD”).

Happy migrating!

Linking databases in Oracle make it easy to share data, and can be useful for replication. However, there is a limitation in Oracle that prevents Character Large Objects (CLOBs) from coming across these links. The following technique uses stored procedures and a temporary table to pull CLOBs across a database link.

First, you'll need the temporary table, which will hold a sequence number, the primary key for the table where you'll want to reconstruct the CLOB, and some text. This table can reside in the source or destination database, but must be linked from the other one. For our purposes, it looks like this…

create table clob_xfer_area
(
  cxa_pk      number(12),
  cxa_number  number(12),
  cxa_text    varchar2(4000 byte)
);
alter table clob_xfer_area add
(
  constraint pk_cxa_id
    primary key (cxa_pk, cxa_number)
);

Second, you'll need the procedure in the source database that breaks the CLOB apart and populates the temporary table.

set serveroutput on size 1000000
set lines 1000
set pages 0
set tab off
set feedback on
create or replace
procedure break_clobs_apart
is
  v_line_number   number(3);
  v_text_piece    varchar2(4000);
  v_total_length  number(12);
  cursor clob_cur is
    select twc_pk, twc_clob_field
    from   table_with_clob;
begin /* { */
  for clob_rec in clob_cur loop /* { */
    v_total_length := 1;
    v_line_number  := 0;
    while (v_total_length <=
           DBMS_LOB.GETLENGTH(clob_rec.twc_clob_field)) loop /* { */
      v_line_number := v_line_number + 1;
      v_text_piece := DBMS_LOB.SUBSTR(clob_rec.twc_clob_field,
        3999, v_total_length);
      v_total_length := v_total_length + 3999;
      insert into clob_xfer_area (
        cxa_pk,
        cxa_number,
        cxa_text
      )
        values (
          clob_rec.twc_pk, -- cxa_pk
          v_line_number,   -- cxa_number
          v_text_piece     -- cxa_text
        );
    end loop; /* } of while */
  end loop; /* } of clob_cur */
end; /* } of procedure break_clobs_apart */

Third, you'll need a procedure in the destination database that puts the CLOB back together, and deletes the data from the temporary table.

set serveroutput on size 1000000
set lines 1000
set pages 0
set feedback on
set tab off
create or replace
procedure put_clobs_together
is
  v_new_clob   clob;
  cursor pk_cur is
    select distinct cxa_pk
    from   clob_xfer_area;
  cursor piece_cur(p_cxa_pk number) is
    select cxa_text
    from   clob_xfer_area
    where  cxa_pk = p_cxa_pk
    order by cxa_number;
begin /* { */
  for pk_rec in pk_cur loop /* { */
    DBMS_LOB.CREATETEMPORARY(v_new_clob, TRUE);
    DBMS_LOB.OPEN(v_new_clob, DBMS_LOB.LOB_READWRITE);
    for piece_rec in piece_cur(pk_rec.cxa_pk) loop /* { */
      DBMS_LOB.WRITEAPPEND(v_new_clob, LENGTH(piece_rec.cxa_text),
        piece_rec.cxa_text);
    end loop;  /* } of piece_cur */
    DBMS_LOB.CLOSE(v_new_clob);
    update dest_table_with_clob
      set  migrated_clob = v_new_clob
      where dtwc_pk = pk_rec.cxa_pk;
  end loop; /* } of pk_cur */
  delete from clob_xfer_area;
end; /* } of procedure put_clobs_together */

Finally, you'll need a procedure that controls the whole thing. We'll assume that this procedure is loaded in the destination database, and the source database is linked with the name “source”.

set lines 1000
set pages 0
set feedback on
set tab off
create or replace
procedure xfer_clobs
is
begin /* { */
  break_clobs_apart@source;
  put_clobs_together;
end; /* } */

(This does not include a commit - the changes will not be persistent unless they are committed.)

Of course, these processes could (and, to be useful, likely would) be integrated into other procedures and scripts. But, this framework will successfully transfer CLOBs across linked databases in Oracle.

I was finally able to resolve my problems with Apache and MySQL. When I decided to mount my FAT32 drive under /home/summersd, I inadvertently caused myself some problems. From talking to a Linux guy at work, I found that no processes that weren't running under my user ID could access those files. The reason is that Linux looks up the entire directory tree, back to /, to determine if you can access the file. So, although I had -rwxrwxrwx summersd summersd on every file, /home/summersd was drwx------ summersd summersd, and /home was drwxr-xr-x. The permissions on /home/summersd were keeping Apache from seeing /home/summersd/drive_d/wwwroot, and MySQL from seeing or writing to /home/summersd/drive_d/mysql/data. I moved the drive to /mnt/drive_d, with the mount point being owned by “root”, still mounting the drive with my user name, and everything worked.

In the process of reconfiguring Thunderbird, I believe I may have found out how to share the address book across operating systems. The file ~/.thunderbird/default.[something]/prefs.js has a listing of all the preferences and settings. I modified this file to change the location of my mail files, and there is a setting there for an address book (which isn't shown in the configuration dialog - after all, it is 0.7.3…) I'll play with that later - right now I'm just elated to have Apache and MySQL working again.

At work, we use an editor called Visual SlickEdit (VSlick). It's got a lot of features, and supports color-coding for many different languages. I decided that I'd give wine another shot, as we only have the Windows version of this program. I installed wine and winesetuptk, used winesetuptk to configure the installation, then ran the installation program. Everything installed, and the program ran up to a point, when it started complaining about a missing DLL. I booted to WXP, found the DLL, copied it to the FAT32 drive, rebooted to Linux, and copied the DLL into the “fake windows” system directory. Soon, it was working great! I can't believe it - success with wine!

I also have made little headway towards getting Apache and MySQL to working. I changed the process that Apache uses to run as “summersd”, and I was able to see pages (although any pages that relied on a database didn't work). I still haven't figured this one out yet…

I'm still getting kernel panics from time to time, and it seems to be whenever I access networking. A suggestion from one of the folks on the WBEL users list was to download the Ultimate Boot CD, filled with diagnostic programs. I downloaded it, burned it, and ran some memory checks. Those checked out, so I'm going to run a “CPU Burn-In” program to see if it can detect errors from the CPU. It runs for up to 7 days, but I think I'll just run it overnight - folding@home didn't take nearly that long to crash it before.

Today, I reinstalled WBEL 3.0. I was able to compile ndiswrapper (as I kept that on my FAT32 drive), and get the network card working smoothly very quickly. (In fact, it seems to be more reliable under Linux than WXP!) With the network up, it was easy to download Firefox, Thunderbird, and OpenOffice, and installing them was a breeze. (I decided to put them under /opt this time, trying to stick with the FHS.) I decided to mount my FAT32 drive under my home directory, as /home/summersd/drive_d. E-mail works fine, but Apache gives me a 403 (Permission Denied) error. MySQL doesn't seem to be working either - I'll have to play with that later.