a view from afar

UPDATED: corrected the last part of the expression from [0-9][ABD-HJLNP-UXYZ]{2} to [0-9][ABD-HJLNP-UWXYZ]{2}

Just doing a bit of DataMapper validation, and found myself wanting a regular expression for UK postcodes. Well it turns out the UK government specifies exactly what an address is (possibly the hardest data modelling problem I know of!!!), and also what a postcode is. What's more their XML schema gives the official regular expression for what a postcode is.

It needs a bit of translation for Ruby 1.8, because the regex library doesn't support character class subtractions, eg [A-Z-[BTV]]. So I've had to expand these. If there are any errors, it's my fault. Please check them if you're going to use the regex yourself, and let me know what you find.

I've broken the full expression onto multiple lines because it's about a mile long, but it's a one-line expression. I haven't investigated using free format mode with it. There is only one space in the expression, which occurs before [0-9][ABD-HJLNP-UWXYZ]{2}. Re-assemble before use!

1
2
3
4
5
6
7
8

POSTCODE_REGEX = /
  (GIR 0AA)|
  ((([A-PRSTUWYZ][0-9][0-9]?)|
  (([A-PRSTUWYZ][A-HK-Y][0-9][0-9]?)|
  (([A-PRSTUWYZ][0-9][A-HJKSTUW])|
  ([A-PRSTUWYZ][A-HK-Y][0-9][ABEHMNPRVWXY])))) [0-9][ABD-HJLNP-UWXYZ]{2})
/

I have a separate environment for running Cucumber features, called (unsurprisingly), features. That means I have three useful environments locally: development, test and features. More on this later.

On Friday I got really caught out, trying to run Cucumber against a features environment that was out of date. Now it was my mistake for not looking at the console more clearly, but it shouldn't be a mental overhead to remember to what state each database is in (especially with 3!). So, on the basis that there are only three numbers (0, 1 and n), I generalised the problem with these rake tasks, which I currently have sitting in Merb.root/lib/tasks/database.rake.

And, yes, process_environments is ugly, but it's partly down to making sure the development environment always appears in the list (instead of rake), and partly down to wanting each database migrated exactly once.

I have no idea how this will work with multiple DataMapper repositories, sorry.