Chris, each change to the seq field should be followed by a resequencing of entries to use positive integers within that section and issue (see resequencePublishedArticles). The increment or decrement by 1.5 is used to move an entry between two other entries, and the subsequent resequencing will take care of making them integers again in the correct order. I'm not sure how those negative numbers got into the DB, but OTOH they shouldn't cause problems.

Alec, Thanks for letting me know about resequencePublishedArticles. I found its description in your Doxygen documentation and it looks like it should work. But it does not work in my installation. I made some minimal changes to code in my installation and none of them came even close to seq processing. I can review my changes again but I wonder if you see any reason why resequencePublishedArticles would not be triggered? Is there a way for me to trace the problem? Thanks, Chris

Chris, the easiest way to debug in PHP is simply to use "echo" statements to output IDs and status info as the code executes. The call to resequencePublishedArticles happens pretty much immediately after the sequence field update. Of course, your error log is probably also a good place to look first.

Please have a look at the following piece of code from
classes/article/PublishedArticleDAO.inc.php.
I used echo to determine that the loop 
for ($i=1; !$result->EOF; $i++)
did not execute. So I changed the loop to
for ($i=1; $i <= 5; $i++)
and once again I used echo to determine that the modified loop
did execute. But seq values are still half-integers
and negative numbers. Any suggestions?

Thanks,
Chris
        /**
         * Sequentially renumber published articles in their sequence order.
         */
        function resequencePublishedArticles($sectionId, $issueId) {
                $result =& $this->retrieve(
                        'SELECT pa.pub_id FROM published_articles pa, articles a WHERE a.section_id = ? AND a.article_id = pa.article_id AND pa.issue_id = ? ORDER BY pa.seq',
                        array($sectionId, $issueId)
                );

//              for ($i=1; !$result->EOF; $i++) {
                for ($i=1; $i <= 5; $i++) {
                        list($pubId) = $result->fields;
                        $this->update(
                                'UPDATE published_articles SET seq = ? WHERE pub_id = ?',
                                array($i, $pubId)
                        );

                        $result->moveNext();
                }

                $result->close();
                unset($result);

                $this->flushCache();
        }

I did more tests. I replaced
for ($i=1; !$result->EOF; $i++)
with
for ($i=1; $i <= 5; $i++)
I also replaced 
array($i, $pubId)
with
array($i, $i)
Then the seq values became consecutive integers, as expected.
So the problem seems to be that $result and $pubId
do not get the right values.

Chris

Chris, the resequence operation only occurs within the provided $sectionId and $issueId. Are you sure you're not watching for changes to "seq" outside of the $sectionId and $issueId that is being resequenced?

Alec, I am testing a new installation of OJS 2.3.4. There are only 5 articles:
http://depts.washington.edu/ejpecp/OJS2_3/ojs-2.3.4/index.php?journal=ejp
They are all published in only one section and one issue. The table "published_articles" lists all 5 published articles. Recall that the problem is not only that the values of seq are unusual (half-integers, negative numbers) but on the journal page, resequencing (sometimes) fails.

Created attachment 3444 [details]
Patch against OJS 2.3.4ish

Chris, try the patch below. The resequence function was being called with an empty section ID.

(In reply to comment #8)
> Created attachment 3444 [details]
> Patch against OJS 2.3.4ish
> 
> Chris, try the patch below. The resequence function was being called with an
> empty section ID.

It works - thank you!  Chris

Created attachment 3453 [details]
Patch against OJS pre-2.3.5

(Committed against ojs-stable-2_3)