Question

I have a rather complex DELETE query that I need to break down into batches, but I can’t do that until I can determine how many rows would be deleted before performing the query. Here is the query that I want to get a count on:

DELETE parts, binaries FROM parts LEFT JOIN binaries ON binaries.ID = parts.binaryID LEFT JOIN releasenfo rn ON rn.binaryID = binaries.ID WHERE binaries.procstat IN (4, 6) AND (rn.binaryid IS NULL OR (rn.binaryid IS NOT NULL AND rn.nfo IS NOT NULL))) OR binaries.dateadded < '2013-01-04 22:01:17' - INTERVAL 36 HOUR;

Here is what I thought might work, but it’s not valid:

Select COUNT(*) FROM (DELETE parts, binaries FROM parts LEFT JOIN binaries ON binaries.ID = parts.binaryID LEFT JOIN releasenfo rn ON rn.binaryID = binaries.ID WHERE binaries.procstat IN (4, 6) AND (rn.binaryid IS NULL OR (rn.binaryid IS NOT NULL AND rn.nfo IS NOT NULL))) OR binaries.dateadded < '2013-01-04 22:01:17' - INTERVAL 36 HOUR) AS countme;

Can someone please point me in the right direction?

To put this into context, I want to use it with the following php pseudo code:

$limit = 10000; // Do in batches to give user status            
$totalRows = $db->query("Count query goes here");

while($rowsLeft > 0)
{
    echo "$rowsLeft remaining to be deleted\n";
    $db->query("DELETE ...... LIMIT ".$limit);
    $rowsLeft -= $limit;
}   

Answer 1

You can simply try this:

SELECT * FROM parts p 
LEFT JOIN binaries b ON b.ID = p.binaryID AND b.procstat IN (4, 6) 
LEFT JOIN releasenfo rn ON rn.binaryID = b.ID 
WHERE rn.binaryid IS NULL OR (rn.binaryid IS NOT NULL AND rn.nfo IS NOT NULL) OR 
        b.dateadded < '2013-01-04 22:01:17' - INTERVAL 36 HOUR;

DELETE p, b FROM parts p 
LEFT JOIN binaries b ON b.ID = p.binaryID AND b.procstat IN (4, 6) 
LEFT JOIN releasenfo rn ON rn.binaryID = b.ID 
WHERE rn.binaryid IS NULL OR (rn.binaryid IS NOT NULL AND rn.nfo IS NOT NULL) OR 
        b.dateadded < '2013-01-04 22:01:17' - INTERVAL 36 HOUR;