Facet for filtering "if property exists"

[timconrad]

Hi.

In MaRDI, we have publication items that can have an optional property "has companion code repository" ( https://portal.mardi4nfdi.de/wiki/Property:P1687 ), e.g. this one: https://portal.mardi4nfdi.de/wiki/Item:Q6292719 .

Can we add some kind of facet mechanism, that would allow to filter for papers where this property exists?

For example, I am interested in papers (items) that contain "optimal transport ". So, when I search and go to the "Publication tab"

( -> https://portal.mardi4nfdi.de/w/index.php?title=+Special%3ASearch&search=optimal+transport+&search=optimal+transport+&go=Go&ns0=1&ns122=1&ns4200=1&ns4202=1&ns4206=1&ns4208=1&ns4210=1 )

I would like on the right-hand side ("Filters") a checkbox with a label "has companion code repository". If I check that checkbox, then only publications are shown where this property exists. Or, it could be another filter with the "heading" "has companion code repository", where the choice "only show items with "has companion code repository" or so.

Best,
Tim

[malberts]

This UI feature is tracked in #119

Currently that query is unofficially supported, but requires manually editing the search query:

• has value: haswbfacet:P123
• does not have value: -haswbfacet:P123

Example:
Publications with Author:
haswbfacet:P1460=Q5976449 haswbfacet:P16
https://portal.mardi4nfdi.de/w/index.php?search=haswbfacet%3AP1460%3DQ5976449+haswbfacet%3AP16&title=Special%3ASearch&profile=advanced&fulltext=1&ns0=1&ns122=1&ns4200=1&ns4202=1&ns4206=1&ns4208=1&ns4210=1

Publications without Author:
haswbfacet:P1460=Q5976449 -haswbfacet:P16
https://portal.mardi4nfdi.de/w/index.php?search=haswbfacet%3AP1460%3DQ5976449+-haswbfacet%3AP16&title=Special%3ASearch&profile=advanced&fulltext=1&ns0=1&ns122=1&ns4200=1&ns4202=1&ns4206=1&ns4208=1&ns4210=1

[timconrad]

That's great. How can we get that into the extension UI?

[JeroenDeDauw]

@timconrad @physikerwelt We are almost done implementing this capability. How does the demo video at #260 look?

[timconrad]

Looks promising! Maybe it would me more intuitive to rename "No value" to "Value not set"?

Also, maybe "Any value" should be checked by default?

[JeroenDeDauw]

@physikerwelt this is ready and can be deployed

[physikerwelt]

Works, and can be closed.

[JeroenDeDauw]

@physikerwelt I finally got round to writing a script for the demo video on Monday and will record the video soon. After that it's blog post time.

[timconrad]

Quick question:

I wanted to update the our "facet search" config to include a new property.

If I change the config, the newly added property is not shown yet on the right hand side.

Is that because I did not yet rebuild the Search Index as described here: https://professional.wiki/en/extension/wikibase-faceted-search#BuildSearchIndex ?

That re-build would take 2-3 days on our current instance.

[JeroenDeDauw]

Facets show up immediately, with one caveat: list facets are not show if there are no selectable values to display. We could easily change the behavior to always show those list facets, but you won't be able to use them if there are no indexed values yet.

[timconrad]

Ah - that explains why we didnt see that. So there is no way to circumvent the (very) lengthy reindexing?

[malberts]

Unfortunately not. The available list values come from Elasticsearch, and that is only available once it indexed the values.

[timconrad]

OK. What I actually need is a way to just check whether a property is available. I am not interested in any value.

So, I want to only search in items, that have, say, property P1687 (which is of type URL) set.

Use case: some of our publication items have code repos attached. Now I only want to see publications that contain the keyword "Wasserstein" AND have a code repository attached.

Is that possible?

[malberts]

If you want to have a facet that shows only "Any value" and "No value", and not the actual list, then it is not currently possible. We will need to add new config to allow hiding the list.

However, that would still depend on reindexing. For Elasticsearch to know whether an item has a value, it needs to update all the indexed items first.