Hello everybody, I’m quite new to editing on OSM but I wish to add more addresses to buildings in OSM in Canada considering some cities lack them entirely.
I do not know the appropriate way to do this, thus here are my questions :
Should we tag on buildings when possible or should we use a sepperate address point ?
2.Is The Open Database of Addresses reliable enough to use as reference ?
3.Seeing the scale of the issue, how would you feel using a script to automate associating addresses in the database and buildings in OSM ?
Welcome! To be frank, I would not suggest making this your first project. Data imports are a complicated topic in OSM and address data is often a complicated topic as well.
Tag on building is preferable in cases where the building has one address
I have not examined this dataset so I cannot say. Is it reliable enough to use as a reference? The person suggesting the import would normally evaluate the data source. If you can’t evaluate the data, you probably should not be importing it.
Have you seen Automated Edits code of conduct - OpenStreetMap Wiki and Import/Guidelines - OpenStreetMap Wiki? OSM is generally very wary of using scripts to automate things, especially to automate adding lots of data. It might well be a good idea, but on the scale of the entire country it would have to be planned with a lot of detail and executed carefully.
Setting aside the topic of importing data, if you live near any buildings without addresses I would recommend checking & adding them manually as a decent way to get into the project.
You can do this pretty effectively using the StreetComplete app. I usually pop it open when I’m on a walk and add house numbers / street information to the buildings I pass, since this approach lets you gather and add the addresses at the same time. If you do this for a while, the completeness of the addresses near you will improve dramatically
As far as how, I usually put addresses directly on the buildings unless one of the following is true:
There are tons of them
The building is very large (and so knowing the location of the corresponding entrance is important)
The building has addresses corresponding to multiple streets (rare, but it happens sometimes when a building with several entrances is on a street corner)
In those cases, I prefer to manually add entrance nodes to the building outline and give them their corresponding addresses. You can do this for buildings with only one address, and it does offer slightly more information, but it’s not really that important.
If the address doesn’t correspond to a specific entrance (or you don’t know where the entrances are), you can also put an address node anywhere inside of the building. There’s a lot of valid ways to handle addresses, some of this comes down to personal / community preference.
“Mapping addresses in Canada” is a huge undertaking, we’re the size of a continent. I would approach this from a bottom-up perspective, finding and focusing on municipalities that are missing addresses, rather than from a top-down, country-wide approach.
Usually I add them to building ways, rather than separate points. I will add them to separate entrances if I figure out what’s what. For retail buildings like strip malls, where each ‘unit’ has an entirely different address (i.e. not just “Unit 1, #### Whatever Street”, “Unit 2, #### Whatever Street”, but rather “1234 Whatever Street”, “1236 Whatever Street”, “1238 Whatever Street”, etc.) I will omit an address from the building and add it to the businesses as separate nodes instead.
There are honestly lots of ways of doing this, few them explicitly ‘wrong’.
Taking a quick peek at the sources of the data, it seems these were aggregated in January 2021, making the data already almost five years old at this point. I don’t think that’s a great starting point.
I think you need to define the scale and scope of the issue first, which is why I suggested the bottom-up approach above. Sure, there are some cities that don’t have addresses; however, there are many that do already have them mapped.
Personally I think the greater issue is a lack of addresses in rural areas, which this “Open Database of Addresses” does little to help ameliorate, looking at the sources within it.
I’d like to echo some of the sentiments already shared here against doing a bulk import for address data. Addressing is more complicated than most people think. While we’d like there to a common standard and universal way of doing them, the reality is that everywhere has its own challenges.
To give you an example, here in Regina, we still use cardinal suffixes on the house number. So you can have something like 100-E Victoria Ave, which might be different from 100 Victoria Ave E. (We’re one of the last cities still using that style, if I remember correctly.) This is a nuance that could be really easy to miss if you don’t know about it, especially for automated edits.
If you’re looking to help with addressing, you honestly would be doing a service by looking around where you live for missing or incorrect addresses. You might be surprised by how many unit addresses (“A, B”, “1,2,3”, etc) are entirely unmapped. Or areas where the addresses were interpolated incorrectly (houses that were estimated to count up by some step, but they actually don’t).
Please do not import anything in Quebec, since we are currently trying to get official Adresses Québec and property assessment role data to be allowed for osm import.
In Quebec, the Open Database of Addresses is incomplete and is an aggregation of multiple sources, with different precision and validation made on the data. It is hard to evaluate its general quality.
Hi @elwwan! I’m also looking to import address data into OSM, though elsewhere (initial post). I’d be interested to learn about your current progress in this effort.
Hello @Cyphase, the progress has been decent. I’m not doing massive imports since I manually check a bunch of addresses to make sure everything is correct. I’m currently using the Vancouver’s open data portal for reference as it is the most up to data database.
Are you doing this manually? As in, looking at the data on the Vancouver portal and copying it over into OSM? Or are you using an exported data set, loading that into e.g. JOSM, and going from there?
I have tools I’ve written that take exported data sets, cleans up the data, chunks them into neighbourhoods, and spits out OSM-compatible GeoJSON. It would be trivial to run them over the Vancouver data set if you were interested in that.
On data cleanups … the addresses in the Vancouver open data portal are typical: bespoke field names (“std_street”, e.g.) filled with their own local blend of abbreviates, so that what is tagged as West 10th Avenue on OSM appears as “W 10TH AV” in their dataset. Ugh.
But with all that cleaned up automatically, the remaining work is to load a community/neighbourhood of points in JOSM, align the points over builds where necessary (Vancouver places address points on the centroid of the land parcel .. this is not the best, but also not the most annoying, choice…), and then (assuming you have the buildings helper plugin loaded in JOSM) selecting an area and hitting Ctrl+M to merge the addresses into the buildings.
I’ve done a number of Canadian towns and cities this way over the last few years and while a bit arduous, it’s the fastest method that delivers good results I’ve been able to find.
If you, or anyone else doing address work in Canada, would be interested, I’m happy to generate the cleaned-up GeoJSON files for cities/towns that have open data portals.
