[talk-au] Using roads dataset from data.sa.gov.au

Henry Haselgrove haselgrove at gmail.com
Tue Mar 10 01:01:59 UTC 2015

I agree… with a dataset this large prioritisation is important. All your specific suggestions for culling parts from missing.osm sound good. It would be easy to add an option to the scripts to exclude highway={primary, secondary, track}. And the suggestions you made via github look good too.


However, I think that some more significant changes to the script should be done before this data is unleased onto maproulette. The roads that are currently in missing.osm fall (more or less) under three categories:

                -- roads that are completely absent in OSM

                -- roads that appear in OSM, but have an empty name

                -- roads that appear in OSM and have a non-empty name which is different to the datasa name, either because OSM is wrong or datasa is wrong (or because both are right, such as “Mount Magnificent Road” versus “Mt. Magnificent Road”)


I propose to modify the script to automatically exclude as much as possible from the third category. Because, it will be hard for an armchair mapper to decide whether OSM or datasa is wrong in those cases. I could try to do this over the coming week. 


I’m not sure I agree that the Adelaide metro area should be given particular priority over other areas. But I’m probably biased, since I grew up in rural SA!


Probably we should make a posting to the osm “imports” list before too much longer, to let them know what we’re thinking. 


From: Daniel O'Connor [mailto:daniel.oconnor at gmail.com] 
Sent: Monday, 9 March 2015 5:32 AM
To: Henry Haselgrove
Cc: Alex Sims; OSM Australian Talk List
Subject: Re: [talk-au] Using roads dataset from data.sa.gov.au


So, after doing this manually for a bit; it's generally working well.


There are some where spot checking against other sources suggests the dataset is wrong, how do you suggest we indicate these?


I've put in NOTE or FIXME on the relevant way.



The thing that is troubling me is the size of the dataset - a few hours work barely makes a dent.


I've taken to deleting all Primary/Trunk, Secondary and Track ways from the data set; and then cropping stuff down to the metro adelaide area; and it's still very sizable.



I'd be really keen on maproulette at this point - you seem to be able to produce updated files fairly regularly, adding a few bash scripts to turn that into curl friendly statements seems achievable.


Going to start sending a few pull requests your way to get us started on this.



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-au/attachments/20150309/8701c78a/attachment.html>

More information about the Talk-au mailing list