DigitalCommons / covid-mutual-aid-project

The Linked Data Version of the Mutual Aid Wiki
0 stars 0 forks source link

Dialog bugs #1

Closed ColmMassey closed 3 years ago

ColmMassey commented 4 years ago
wu-lee commented 3 years ago

The link for an initiative I checked is https://www.facebook.com/https://www.facebook.com/groups/220944599058149/?ref=share

That is clearly malformed - this link works: https://www.facebook.com/groups/220944599058149/?ref=share

The URL in the data is not malformed, so it is a dialog issue.

wu-lee commented 3 years ago

Create a ticket to create an approriate Mutual Aid qualifier in ESSGlobal

Is this meant to be conditional on it being a data problem? If not can you explain this a bit more?

ColmMassey commented 3 years ago

Create a ticket to create an approriate Mutual Aid qualifier in ESSGlobal

Is this meant to be conditional on it being a data problem? If not can you explain this a bit more?

The sausage factory has defined them all as co-ops. This is not correct. So we need to stop that.

Seperately it would be good to have a Mutual Aid qualifier for initiatives, which would requiresome thinking.Looks like I've already created a ticket to cover this. https://github.com/SolidarityEconomyAssociation/open-data/issues/52

wu-lee commented 3 years ago

Regarding the Facebook links. What seems to happen is:

But in addition to the problem above, it seems like the original mutual-aid data has non-facebook links in the facebook fields (e.g. Hooshmand World Peace Foundation has http://www.globalpeacehouse.com/, and there seem to be plenty of others). So it's not as simple as just stripping the facebook domain off the link, as it is for limesurvey data.

We could:

In cases one and two we could then strip the https://*.facebook.com domain part from the URLs and let everything work as-is, resolving the broken Facebook link problem. In case two, this might lose information in the process.

Case three has some appeal, but we can't simply let these be free links because they aren't in the other data sets (like oxford). To do this we would also need to adjust the limesurvey conversion in the process: effectively it's a change in our standard CSV schema, which implicitly expects facebook field to have facebook paths rather than full URLs in them.

(This expectation would be better made explicit, IMO, which is what I hope the standard schema will do eventually - include validation rules on the contents of fields, so we can spot these problems earlier rather than later.)

Any thoughts on this @ColmMassey ?

ColmMassey commented 3 years ago

See https://github.com/SolidarityEconomyAssociation/open-data/issues/35

wu-lee commented 3 years ago

Just to clarify, I think the second item (about the organisational strucutre of Mutual-Aid), must have been ticked because there's a ticket for that. (https://github.com/SolidarityEconomyAssociation/open-data/issues/52)

To expand on what's going on: it's a data issue. The converter currently hard-wires it to "Co-operative".

We can hard-wire it to a new qualifier term, but this requires referencing a newer version of ESSGLOBAL than V2a, and we currently can't do that and have the dataset rebuilt automatically. See https://github.com/SolidarityEconomyAssociation/covid-mutual-aid-project/issues/2#issuecomment-810384568 - the reason is that we can't currently mix and match schema versions for different open-data sets on a given sausage-machine build server, and the two we have are on V2a. So I think that needs to be addressed in order to be able to use a new qualifier.

ColmMassey commented 3 years ago

For the moment, let's just drop creating the triple which calls this a co-op

essglobal:organisationalStructure https://w3id.solidarityeconomy.coop/essglobal/V2a/standard/organisational-structure/OS115 .

We can add the Mutual Aid qualifier when we are ready to next upgrade ESSGlobal again.

wu-lee commented 3 years ago

Ok. I discovered this is effectively what happens if I change the conversion script to write "Mutual-Aid" instead of "Co-operative": it doesn't match any term label, and there is no triple created. This then fouls up the SPARQL query which mandates an organisational structure field - however making that optional is a trival change. I'm not sure what then happens in the dialog if this field is absent, will have to check.

[edit] It isn't great that this happens silently - I think the conversion should fail, or at least warn, if there are invalid values.

wu-lee commented 3 years ago

Fixed on dev-0 I think.

Just as an aside: the data seems to be very poor quality. Website and Facebook URLs are all over the place, including broken ones, What's App links, phone numbers, URL shortners to who-knows where, links to non-existant sites, links to irrelevant sites, and text written for human consumption.

ColmMassey commented 3 years ago

This can be pushed to prod now.

wu-lee commented 3 years ago

Done - sausage-factory build also prodded to rebuild the production data, with cleaned-up facebook links.