Monday 1 June 2015

OS Locator is to be withdrawn

I just received an email from Ordnance Survey, telling me that OS Locator, part of the OS Open Data, is to be withdrawn in a year's time. They say that OS Open Names has been published and that replaces OS Locator. They hint that it may replace CodePoint Open too, though there's no notice of withdrawal for that yet. I thought I should look at OS Open Names to see what it offers.

The first thing I saw was that the data is broken into the OS large-scale grid squares, which break the country into a 13x7 grid with 55 squares with actual data in them. Each one of these needs to be downloaded separately to cover GB. That is a pain to start with. I requested SE and TA which are both needed to cover the village I live in. After downloading them the .zip file contains each area broken down into 100 separate sections as I would expect, but I quickly realised that there were only 25 in SE and 10 in TA. Much of TA covers the North Sea, so there should be fewer sections, but clearly many were still missing. I downloaded another area and again the same 75 sections were missing. I emailed customer service at OS to point this out, with no answer yet.

I pressed on to look at the data in the sections that were there. The data is in a CSV format (there was another choice). There is a summary of column headings, but the data is a strange muddle of three types of data. Much of the data is a url to an OS website with the data summarised on a page for each column of each line of the text. It appears that there is a record type for place names, a record type for postcode centroids and a record type for named roads all freely mixed up throughout the file. There doesn't seem to be a field to simply identify what the record type is for each line. The first field is an id field. For place names it seems to start with osgb and has a number after that, for postcodes a postcode type id, with spaces removed, is in the id field and for road names an id that looks like a GUID is in the field. It doesn't seem to be a pattern for why the data is in the order that it is.

I'm going to throw some code together to disentangle these record types and see what useful data is then available. Hopefully we will not have lost anything useful in this process, though it does look as though processing this open data is going to be a bit harder than it used to be.

Anyone would think OS don't want to release open data.

Edit: There are two fields which distinguishes these different record types.

No comments: