[OSM-dev] 0.6 API hack-a-thon Date

Andy Allan gravitystorm at gmail.com
Thu Oct 16 11:52:05 BST 2008


On Thu, Oct 16, 2008 at 11:37 AM, Tom Hughes <tom at compton.nu> wrote:
> Andy Allan wrote:
>
>> Actually, we won't. We still desperately need more unit and functional
>> tests for the code base - I've been writing unit tests for utf-8
>> handling, and finding truncation errors in parts of the code. I also
>> need to write tests so that invalid utf-8 byte sequences are handled
>> properly. There's not much point in *not* getting things like these
>> sorted.
>
> On the subject of UTF8 I noticed that you made a commit on the api06 branch
> yesterday that will break when used against the live database.

Yeah. I figured out that without that setting the whole chars/bytes
thing was impossible to resolve. At the moment my philosophy is to
make a 'correct' version of the code, so that everyone can develop
against a fully working 'perfect' system. And then figure out how to
special-case the transition to it on the live server. But I'll defer
on this if we want to do it differently.

> Or are we are planning to do a complete reload of the database to fix the
> UTF encoding mess during the 0.6 transition?

I'm not sure there's many other options available, and I think we'd
discussed it that way before. The sanest way seems to be to suck it
out of the database with a latin-1 connection (as now) and stuff it
back in through a proper utf-8 connection - unless anyone wants to
come up with SQL to give the same end result.

Cheers,
Andy




More information about the dev mailing list