Vicmap Address Import Progress

Breaking out from the existing thread at Should we import vicmap address data for victoria? - #19 by aharvey

The import has progressed and so far I’ve completed the uploads for Stage 3, adding new addresses from Vicmap for blocks which did not previously contain any addresses.

The next stage is to upload new addresses from Vicmap for blocks which did previously contain addresses, taking into account conflation to try and avoid adding duplicate data.

As a result of this import, there are situations where we imported multiple address nodes at the same location. While there are sometimes valid reasons for this, a lot of the time it’s limitations of Vicmap data and there is room for improvement.

Therefore I’ve created a MapRoulette challenge at https://mpr.lt/c/50541 which identifies all of these locations for manual review.

2 Likes

Stage 5 completed Import/Catalogue/Vicmap Address - OpenStreetMap Wiki

Now the bulk of new addresses have been added.

Great work Andrew - a valuable addition to OSM!

1 Like

@mueschel raised a point about the addr:flats keys when the value overflowed 255 characters (the maximum allowed tag value). In the import we had used addr:flats, addr:flats2, addr:flats3

At the time of preparing the import I wasn’t aware of the best practice for this and had raised the point at Talk:Key:addr:flats - OpenStreetMap Wiki.

Instead @mueschel suggested to use numbered suffix in key scheme per Multiple values - OpenStreetMap Wiki.

Purely relying on this scheme instead of a semi-colon separator Multiple values - OpenStreetMap Wiki would result in too many tags, so combining both schemes would be the only viable option in my view.

However according to overpass turbo there are 49 nodes where we made use of addr:flats2. and I think it would be worth manually reviewing them all to ensure they are indeed correct. They would probably require a ground survey to confirm. Should we open map notes for all of these? I don’t want to spam notes though, especially if it would be hard for people to solve them.

I also think it’s worth a wider discussion to determine if there are better ways of tagging these.

eg. addr:flats:floor:1=list of units on floor 1, addr:flats:floor:2=list of units on floor 2 etc.

I think they will definitely need local review for the best outcome.

I also noticed on this node that the split of the numbers triggered between two units
addr:flats=…5615-5618;5710**-**
addr:flats2=5713;5715-5720

I will check a few out and can fix these but I think notes are a good idea to highlight the issue. certainly some strange numbering in some locations which could also be vicmap errors.

That was actually by design, when the value overflowed 255 characters, it’s split exactly there, I didn’t attempt to split only at the semi-colon. This is something we could do an automated edit along with using the suffixed colon.

If we feel the values are too unwieldy I’m also happy to delete them and then leave them to be surveyed.

I have fixed a lot of them

Yeah I saw, but I wouldn’t call them broken to begin with :stuck_out_tongue:

The more I think about this the more I feel it would be better to split the unit numbers per level with tags like

addr:flats:level:8=801-803;805-809
addr:flats:level:9=901-903;905-909

and/or

addr:flats:level:ref:G=1-3;5-9

Per the existing tags level=* and level:ref=*.

This will keep the tag values shorter and make it easier for us mappers too, while also adding extra data about the level each unit number is found on.

The only limitation is sometimes it would be obvious from the numbering scheme or the letter boxes alone.

The downside is this requires one to enter the building to map, which in many instances you wouldn’t be able to access. Quite different to just surveying the letterboxes from outside.

EDIT: proposed at Talk:Key:addr:flats - OpenStreetMap Wiki*

That would be much clearer but I always get confused between floor and level in these circumstances. Maybe consider adding that into the note?

overpass turbo - latest for the 49 sites

That’s not what I meant. My comment was just about using the common scheme addr:flats:3 instead of the uncommon addr:flats3. The semicolon-separated lists are fine.

That would be a totally new and up to now unsupported scheme for addresses.
It might be better to have individual nodes for each level.

My understanding of the two schemes at Multiple values - OpenStreetMap Wiki

is if you have a set of multiple values a, b, c, d the suggested schemes are tag=a;b;c;d or tag=a + tag:1=b + tag:2=c, + tag:3=d.

What I’m saying is we would need to combine these two to use

tag=a;b + tag:1=c;d.

I’m fine with this, but it doesn’t really solve the problem of these values being hard to review and edit.

eg, if you need to tweak a value, then you suddenly find tag:1 is too long, you need to change tag:2, tag:3 etc to move everything back to make space.

It’s just too unwieldy.

Does anything support addr:flats:2 etc. at the moment anyway?

I checked the Vicmap source data, and for each unit it also includes the floor number!

So from this we can instead tag addr:flats:level:N=X as the list/range of flats X on each level N.

I don’t have time to work on this myself at the moment, so I’m happy with any option.

  1. Leave what’s there there, and anyone can change as needed
  2. I go through and delete some of these imported addr:flats tags where they are just too long to work with
  3. Someone else volunteers to try and make use of the vicmap data to clean up these tag values.

I’d just leave it for now. It looks like Nominatim doesn’t currently handle addr:unit (and its friends):

Which makes this all a bit theoretical. If they ever figure out how to consume these then we can bulk edit them to match whatever scheme they come up with.

1 Like