[Talk-GB] Import UK postcode data?

Dave F davefoxfac63 at btinternet.com
Fri Oct 4 12:47:03 UTC 2019

On 04/10/2019 01:52, ndrw6 at redhazel.co.uk wrote:
> On 04/10/2019 00:26, Dave F via Talk-GB wrote:
>> I think you're missing the point. Most contributors believe postcodes 
>> on buildings or property nodes, add quality to the OSM's database, 
>> but object to the import of codepoint as it's just not accurate 
>> enough as stated in this, & numerous other threads.
> This is incorrect. CPO/ONSPD postcodes _are_ accurate, up to date and 
> include all postcodes in the UK except NI. 

Please note: "not accurate *enough*"

> They are not complete (contain one and only one delivery point per 
> postcode), 

No. The centre point is not associated with *any* delivery point. It is 
an arbitrary mean, calculated mathematically. it could, in theory, be 
located in the middle of a park.
Even postcodes unique to one property/business aren't accurate as their 
positions are misaligned by the effect of adjacent areas.

> which is pity, but that's not a reason not to use the ones that are 
> available, which is still _far_ more that what we have in the database.

Quantity does not equate to quality.

> This may not be a perfect solution but the information CPO/ONSPD 
> contains is still extremely useful for geocoding. Search for a 
> postcode and you are _guaranteed_ to get an address in a close 
> vicinity to a place you are looking for. 

No. With an import of the centroids points you're only guaranteed to be 
given the location of the node with the postcode.

> How about not needing to start Google Maps when searching for a 
> location on the go?
>> There's no point in importing to stand alone nodes as deliveries are 
>> destined for buildings. Adding to streets is also pointless for the 
>> same reason plus they can have multiple postcodes.
> Addresses on nodes are commonly used in the UK OSM. Many mappers 
> prefer them over placing addresses on buildings. There are also many 
> cases (POIs) where nodes are objectively better than buildings. So, 
> no, there right and wrong solution here.

Allow me to clarify. I should have maybe said 'properties' which can be 
represented by nodes instead of 'buildings'.
My objection, which I thought was clear, was to "standalone nodes" with 
just a postcode tag.

> Besides, the main reason for importing these data is that we can get 
> _all_ postcodes in the database.

Again quantity /= quality. If you can't manipulate data then it's 
useless. These standalone postcode nodes will relate to nothing.

> This gives users confidence that when they search for a postcode they 
> will reliably get a result they are looking for. This is not possible 
> when merging postcodes with buildings simply because we still have 
> only a small fraction of buildings in the database.
> By the way, I'm not against merging addr:postcode with buildings, 
> that's exactly what I was doing myself when adding postcodes manually. 
> However, this is not a process that can be automated (lack of 
> buildings, single OSM buildings having more than one address/postcode).

Then add buildings.

> Based on my experience with mapping postcodes with CPO, I would 
> recommend starting with an import and merge postcodes and buildings 
> later.

Experience has shown that doesn't happen. I'm thinking US TIGER imports, 
but I'm sure there are other examples.


More information about the Talk-GB mailing list