True facts about... Extended postcodes

deals with geocoding and reverse geocoding
Post Reply
User avatar
Bernd Welter
Site Admin
Posts: 2564
Joined: Mon Apr 14, 2014 10:28 am
Contact:

True facts about... Extended postcodes

Post by Bernd Welter »

Hello together,

these days a customer requested more details about our internal handling with extended postcodes. Therefore I decided to collect some basics and to provide them here in the forum. There we go - I hope this enables you to get a better understanding of what you can expect from them (and what they can't resolve!):

Countries like the USA, Great Britain/UK, Netherlands provide different levels of postcode character lengths. If the extended postcode is available for an input address this usually enables us to have a detailed geocode just on the basis of the postcode
  • UK has 7 characters (e.g. “AB101AA”) in extended mode, 5 characters in standard mode (“AB101”)
  • USA has 9 characters  5 characters
  • Netherlands have 6 digits versus 4 digits in standard mode
If your input address is based on a standard level we need additional information for a proper geocoding: City name, Street name, … Now to understand the effects of geocoding with extended codes you need to be aware of the following facts:
  • PTV map data contains both
    • standard data (based on standard codes + city names, street names, house numbers and coordinates)
    • extended data (based on the extended postcodes and their coordinates)
  • If you enter a standard postcode we will search for it in the standard data (and only in the standard data) together with city name and street and so on. There will be no search on the extended data.
  • If you enter an extended postcode we look for a perfect match of the postcode in the extended data. City/district, street and housenumber are ignored in this step. If there is no perfect match we fall back to the standard data. For the fallback we cut of the the ending characters of the postcode.
And also very important for the global understanding:
The basic search! Always based on standard postcode length and city name and street

And here are some examples.
As you can see the search for 6 leading characters “AB101A” returns several hits with extended postcodes (see DetailLevelDescription). Those are all the postcodes and their reference points that start with AB101A. Very important: There is no AB101AC
As you can see the search for 6 leading characters “AB101A” returns several hits with extended postcodes (see DetailLevelDescription). Those are all the postcodes and their reference points that start with AB101A. Very important: There is no AB101AC
AB101AA (existing postcode) - UNIQUE hit with EXTPOSTCODE.<br />As the code AB101AA exists in the extended data we can find a perfect match and this is returned as a UNIQUE result on EXTPOSTCODE level. (A street nearby is Broad Street.)
AB101AA (existing postcode) - UNIQUE hit with EXTPOSTCODE.
As the code AB101AA exists in the extended data we can find a perfect match and this is returned as a UNIQUE result on EXTPOSTCODE level. (A street nearby is Broad Street.)
AB101AC (non existing postcode) - Multiple hits with POSTCODE. <br />As the code AB101AC does not exist in the extended data we fall back to the basic data. This is why you can also see the district names in the result list and the detail level is now POSTCODE (and no longer EXTPOSTCODE).
AB101AC (non existing postcode) - Multiple hits with POSTCODE.
As the code AB101AC does not exist in the extended data we fall back to the basic data. This is why you can also see the district names in the result list and the detail level is now POSTCODE (and no longer EXTPOSTCODE).
AB101AC with BROAD STREET - UNIQUE hit even more detailed! <br />As we do not find a perfect match during the EXTENDED search we fall back to the standard data. In the standard search we use postcode AB101 and street name Broad Street and we are able to find a good result  DetailLevel is even HNRSECTION
AB101AC with BROAD STREET - UNIQUE hit even more detailed!
As we do not find a perfect match during the EXTENDED search we fall back to the standard data. In the standard search we use postcode AB101 and street name Broad Street and we are able to find a good result  DetailLevel is even HNRSECTION
AB101AA with BROAD STREET - UNIQUE hit at EXTPOSTCODE level. <br />Once more we are successfully looking for the extended postcode and this is why we return the UNIQUE hit on EXTPOSTCODE level.
AB101AA with BROAD STREET - UNIQUE hit at EXTPOSTCODE level.
Once more we are successfully looking for the extended postcode and this is why we return the UNIQUE hit on EXTPOSTCODE level.
And some more statements straight from DEV DEPARTMENT (thanks to Jochen!):

What are ExtendedPostcodes?
The standard map providers (HERE, TOMTOM, AND) provide postcodes for each town, mostly on 4- or 5-digit level, depending on the country. However, in some countries there are also postcodes with additional letters or numbers at the end that identify an address to the street level or even further. Therese are called in this document 'Extended Postcodes'.
The data for the extended postcodes is mostly defined and provided by the national mail company and contains for each extended postcode at least a coordinate, plus sometimes additional fields like Town, Townpart, Street and HouseNumbers.
At the time of writing, extended postcode data was available for the Netherlands and Great Britain, where only the Netherlands had also the additional fields Town, Townpart, Street and HouseNumbers filled.

How are ExtendedPostcodes used when searching for addresses?
Extended postcodes are used only in the town and singlefield searches. The input is scanned for possible long postcodes which are then used to search for exact matches in the extended postcode database. Note that only the postcode is used to find the results, all other input words are ignored.
A record is only returned if the input postcode matches a database entry exactly. However, minor typos are allowed:
  • the comparison is case insensitive
  • the comparison does not care about spaces and dashes
When an exact match is found, the extended postcode database entry is used to fill the result record fields postcode, city, city2, street and house number. Additionally, some fields related to the country are filled from the other map data.
When an extended postcode record is returned, the normal classification algorithm is executed, taking also the other input fields into account. If there exists no data for town or street return fields, these fields are assumed exact.
If no suitable extended postcode match is found, the normal search is executed to find a result.

Where are ExtendedPostcodes stored?
As said before extended postcode data is supplied by another provider than the standard map and is thus not included in the normal map distribution. Instead you have to abtain a separate postcode index file, that can be copied into the map directory.

There is one postcode index file per integration unit ('country') which is named <country>.gpi and must be copied into the directory of that integration unit. The integration unit directory is the directory where e.g. the files <country>.lv1, <country>.gcd and/or <country>.bdg are. If your map contains instead of those only files <country>.gp, then you have to create a new folder named <country> next to the file <country>.gp and copy the <country>.gpi file into it.

How do know that my gpi files have been loaded successfully?
When loading the map, the geocoder prints for each integration unit a log message to the log output stating whether or not the extended postcode data was found and loaded successfully.

How do gpi files relate to the mapserver ezi files?
It is true that the two solutions seem similair, but the gpi files are not compatible with the old ezi files. There was only one ezi file per map, while there is one gpi file per integration unit. The file format and the structure of the records have changed, so mapserver will not load gpi files and gpGeocoder will not load ezi files.

How can I create my own gpi files?
This distribution contains also a tool for importing gpi files from a CSV format. See Extended Postcode Importer Tool for more information.

Best regards Bernd
Joost
Posts: 307
Joined: Fri Apr 25, 2014 1:46 pm

Re: True facts about... Extended postcodes

Post by Joost »

Some extra information about Dutch extended postcodes:

Our default map data providers TomTom and Here have extended postal codes available on a segment level. After searching is done our geocoder can retrieve the extended postcode and enrich the answer. In other words: the geocoder cannot search on 6 digit, but can return them as part of the result. If you want to search using extended postcodes you need a GPI file like Bernd described above.

Example:
example.JPG
Complete address data is available, so the geocoder will use the postcode (but only the first digits), city , street and house number to find a match. After the result have been found it will see that their is a extended postcode define and fill this in the result. The classifying is done on the complete result.

Example 2:
example2.JPG
Only extended postcode and house number is given in the input. The geocoder can only search on the 4 digits, so it will return a hit based on this level. If you want to geocode on extended postcode and house number only (which in NL maps to an unique addresses) you need a GPI file of the dutch postcodes.
ajg
Posts: 2
Joined: Tue Mar 27, 2018 3:04 am
Location: Melbourne, Australia

Re: True facts about... Extended postcodes

Post by ajg »

I have been looking at XServer (version 1.24) xLocate using XServer Internet and have a question about the DetailLevelDescription response value for addresses containing full UK postcodes.

I am testing addresses using SOAP request with the URL ‘https://xlocate-eu-n-test.cloud.ptvgrou ... ws/XLocate’.

The xLocate API Documentation for DetailLevelDescription states:
A greater value indicates a higher data level.
EXTPOSTCODE2The candidate was found on global postcode level.
POSTCODE5The candidate was found on postcode level.
When I search using the address ‘AB23 8XP, GBR’, I am getting a response back saying that it has found the full UK postcode (which in most cases is better than ‘STREET = 6’), but only returns ‘EXTPOSTCODE = 2’.
Is this always the case with UK postcodes? Or, what is required to get a response ‘POSTCODE = 5’?
Andrew Gibbons
Software Developer (Product Integration)
PTV Group Australia
User avatar
bocajo
Posts: 44
Joined: Tue Mar 01, 2016 3:05 pm

Re: True facts about... Extended postcodes

Post by bocajo »

Normally the DetailLevel for extended postalcodes should be higher than for postalcodes. For some reasons when the xLocate API was defined the decision was made that the enumeration for extended postalcodes is 2. And because of compatibility the enumaration of the DetailLevel can't be changed.
Jochen Anderer
Manager Engineer
PTV GROUP GERMANY
User avatar
Bernd Welter
Site Admin
Posts: 2564
Joined: Mon Apr 14, 2014 10:28 am
Contact:

Re: True facts about... Extended postcodes

Post by Bernd Welter »

Sounds like we should remove the statement "A greater value indicates a higher data level." from the API if it is no longer matching the desired behaviour.
Bernd Welter
Technical Partner Manager Developer Components
PTV Logistics - Germany

Bernd at... The Forum,LinkedIn, Youtube, StackOverflow
I like the smell of PTV Developer in the morning... :twisted:
Post Reply