Avoid use of SERIAL in ALTER TABLE with PostgreSQL

[asmecher]

Per email conversation with Damien -- see http://phplens.com/lens/lensforum/msgs.php?id=18150

[peterdd]

@asmecher, @dregad : The link is lost, so we can't see the discussed problems.

[peterdd]

review/compare with #73 required

[asmecher]

Unfortunately I don't have access to that conversation either. The problem as I recall it was the ADODB PostgreSQL driver using SERIAL as a data type, when PostgreSQL doesn't consider it that way.

The patch I've been recommending to Open Journal Systems users is here: https://gist.github.com/asmecher/d1013624ff5325eae97384ae5adc88c4

It's a shame the phplens content wasn't saved.

[dregad]

The link is lost, so we can't see the discussed problems.

It's a shame the phplens content wasn't saved.

Indeed, a shame that @heyjohnlim took the old phplens / adodb support forums down, and never responded to my requests to provide me with a dump.

Anyway, I found this one on the web archive : https://web.archive.org/web/20150419100241/http://phplens.com/lens/lensforum/msgs.php?id=18150 - pasted below.

---

Topic: AlterColumnSQL uses SERIAL type for PostgreSQL
author: Alec Smecher created: 15-09-2009 04:12:00 PM

Hi all,

It looks like it's possible for the AlterColumnSQL function in adodb/datadict/datadict-postgres.inc.php to generate ALTER TABLE statements using the SERIAL type. SERIAL is not an actual type (it's actually INT4, with a macro to create a sequence etc.) and this causes SQL errors, i.e.:

Query failed: ERROR: type "serial" does not exist

...from generated SQL statements such as...

ALTER TABLE signoffs ALTER COLUMN signoff_id TYPE SERIAL

Thanks,
Alec Smecher
Public Knowledge Project Team

---

author: asmecher created: 14-04-2014 06:28:25 PM

I've hacked together a serviceable workaround for this, but frankly my heart wasn't in it -- ADODB is getting pretty hard to maintain cleanly.

http://pkp.sfu.ca/support/forum/viewtopic.php?f=40&t=11896&p=45922#p45922

[asmecher]

I don't know the circumstances of the forum going down, but I will say that @heyjohnlim was always pleasant to work with and seemed to me to be doing a tremendous amount of thankless middleware work. It certainly helped get OJS off the ground in its early days.

Good sleuthing finding that wayback copy -- I googled briefly for a cached copy or something but didn't spot anything. The patch I recommended above is essentially the same as the one referenced in the quoted message.

[dregad]

For the record, my earlier comment was absolutely not meant as a criticism of John's tremendous work on ADOdb for over a decade - I was just expressing my regrets of not being able to consult the old forums anymore.