Comments on: A close look at three Rails 2.1 bugs

By: Science

Science — Sat, 30 Aug 2008 05:55:04 +0000

There’s a minor typo in your post – the file is “databases.rake” not “database.rake” – totally trivial but if someone is searching for it, it might hang them up. Thanks for the post.

By: Nathan Zook

Nathan Zook — Mon, 04 Aug 2008 19:21:39 +0000

I don’t know enough about the organization of the team to know, but with the separation of some of the dbs from the core, the API for the db adapters has to become brittle. At a minimum, that interface has to be tested to the point that any change will be immediately noted for review. Much beyond that, and I get back to my entire question about why the split was a good idea in the first place.

By: Mario Briggs

Mario Briggs — Fri, 01 Aug 2008 12:43:42 +0000

Nathan,

>>
predicted that this would happen when the non-free dbs were pushed out
<<
I guess you missed the big banner on this blog – Get DB2 for free
DB2 is FREE.

Frederick, tmcedo,
Antonio mentioned about the API team doing testing with EdgeRails to have a better handle on these. We are open to do this for Rails itself and have it benefit the community. thoughts?

By: tmacedo

tmacedo — Mon, 28 Jul 2008 11:28:44 +0000

@Frederick cheung:

The ability to assert SQL has recently been commited to the trunk.

@Nathan Zook:

I agree that removing the commercial adapters from the core probably means that those adapters aren’t being tested but the adapter specific tests are still in the test suite and rails releases aren’t that frequent so even though those adapters are not on the core they should be tested before making a new release.

By: Antonio Cangiano

Antonio Cangiano — Thu, 17 Jul 2008 10:59:37 +0000

I agree entirely, Nathan.

By: Nathan Zook

Nathan Zook — Wed, 16 Jul 2008 19:18:58 +0000

By all means stick to the standard! (And by the way, exactly how is it that after this much time, the rails test suite isn’t ensuring this?) But my point is that these escapes are the natural result of moving the commercial db adapters out of the core. The standards compliance issue is not the problem per se. The per se problem is that a version of Rails was released that preferentially broke some of the commercial db adapters.

If Rails is to become accepted by the enterprise, it is going to have to provide first-class support for the databases that the enterprise uses.

By: Antonio Cangiano

Antonio Cangiano — Wed, 16 Jul 2008 02:33:19 +0000

I encouraged IBM’s API team to work with Edge Rails and to do daily testing for this exact reason, Nathan. That said, I think that Rails contributors should still try to stick to the SQL Standard as much as possible. It’s akin to saying that speeding is a bad idea, but speeding without a seat belt on is an even worse one. SQL standard can be that seat belt.

By: Nathan Zook

Nathan Zook — Mon, 14 Jul 2008 21:52:03 +0000

I kinda predicted that this would happen when the non-free dbs were pushed out of the Rails core. The fact that the error this time is that that Rails went non-standards while the commercials were standards is really irrelevant. It could just as easily happened the other way.

Not tested == broken!
Not tested == broken!
Not tested == broken!

Therefore, any change that is introduced into the Rails core must be assumed to break all of the non-core dbs until it has been demonstrated otherwise. That is: any update to Rails must be assumed to break the non-core dbs until the maintainers of those db adapters have run the complete test suite against their adapters.

By: Frederick cheung

Frederick cheung — Mon, 14 Jul 2008 12:39:45 +0000

Oops, that non standard sql was my bad 🙂 Damn those overly permissive databases (ie sqlite, mysql, postgres)!

It would be interesting if the rails tests could assert that the sql produced were valid. I suspect a better approach would be to revive rails’ continuous integration server and have it run the tests against more databases (or find someone to loan machine/licenses for that).

That bug could have been found and fixed back in february (when that change was introduced) if someone had been looking for it.