Comparator to flag invalid primary tag combinations

[bkowshik]

Objective

• Fixes: Features on OpenStreetMap with more than one primary tag #94

Per #94 (comment)

Amazing stats @bkowshik, this directly gives the probability of any tag combination on OSM. With these 286 invalid and 330 possibly erroneous tag combination, can we already start flagging them?

• Happy about how we parse a csv file instead of a json, making it easy on the 👀

Next actions

• Would love to get some 👀 @amishas157
• Merge to master

---

cc: @planemad @geohacker

[geohacker]

@bkowshik this uses a csv file for the tag combination scores - how do you plan to make this maintainable? should we use the taginfo API directly?

[bkowshik]

At the moment, we have a script to generate the csv file here:

• https://github.com/mapbox/osm-compare/blob/master/scripts/primary-map-features.js

Options

@geohacker, I could think of the following 3 options:

1. csv file in the repository
   • TagInfo is not queried for each feature
   • Simple to build and easy to understand
   • Needs to be updated manually using the script ^
2. csv file created during deploy
   • TagInfo is not queried for each feature
   • Latest data is queried from TagInfo during deploy
   • Data is as frequent as the number of deploys
   • Makes the deploy a little slower
3. Query TagInfo
   • TagInfo is queried for each feature which is a lot of features!!!
   • We have the latest data and no manual maintenance

Unless you are seeing something here, I am happy with the current setup, (option 1). We will get a better sense of what is needed once we deploy and 👀 the results for a few days.

[planemad]

👍 the current CSV is good till we see some results from this. We can think about scaling with taginfo if this works as expected.

[amishas157]

@bkowshik This looks good to go. And also regarding keeping the tag info data updated, i think manually running scripts is fine. As there should not be much difference in tag-info data on daily basis, if we think of it in terms of percentage. But as we are scaling our comparators based on taginfo, we can put some cron jobs to run these scripts on a monthly basis or so.

[bkowshik]

Published to npm as version: 4.14.0

[geohacker]

🎉