Revisions and Commits
- Mentioned In
- rMEXTb6a40d878ced: Updated mediawiki/extensions Project: mediawiki/extensions/Wikibase…
rEWBA171c0dda642d: Fix double-escaping issue in QuantityDetailsFormatter
rMEXT0f8e8e82e0bf: Updated mediawiki/extensions Project: mediawiki/extensions/Wikibase…
rEWBA1289e1e92494: Use mocks in QuantityDetailsFormatterTest
- Mentioned Here
- T111111: [RFC] Wikibase should specify the significance level of lowerBound and upperBound
This is already done by applying a UnitFormatter with a LabelDescriptionLookup. The "problem" is that this lookup is not a Wikidata.org lookup but one in your local repository. We agreed that this is enough for what we need on Wikidata.org now. Can be improved later.
Reopening this as it is still showing a URI here: http://wikidata.beta.wmflabs.org/w/index.php?title=Q23950&diff=217186&oldid=43001 And since this is using local URIs I would assume those are treated as they'd be on Wikidata itself.
Showing only the label is not good, because it's ambiguous. This is not just a theoretical problem, but rather nasty in practice: There are a lot of units that have the same name but are slightly different - or also refer to something of a similar name.
For example, imagine someone picked "pint" as a unit. They wanted Q11123, but they accidentally picked Q20429. Someone notices this and fixes the mistake. What does the diff of that edit look like? If we just show the label, the edit is invisible, because both items have the same label ("pint"). That's not good at all...
Showing the label in the diff helps, but we need to show it along with the Q-id or URI, or at least the description. Otherwise it's ambiguous and may lead to confusion.
If you're taking that approach, you should probably also include an indicator of the level of significance of the upper/lower bounds.
I think the upper and lower bounds were created from the input to the user interface or API. The upper bound would be the nominal value plus the uncertainty, and the lower bound would be the nominal value minus the uncertainty. According to the data model (https://www.mediawiki.org/wiki/Wikibase/DataModel#Quantities) the quantity is stored in the data base as three numbers, the amount, the upper bound, and the lower bound. In this simple-minded model, the upper and lower bounds would be exact, absolute limits. (But in real life most quantities with a well-studied uncertainty will be points on the normal distribution, often plus or minus two or three standard deviations.)
That's my point - you need to be clear whether the uncertainties are 1/2/3/more/less standard deviations, as these do differ between references and fields of work. (Leaving aside issues of non-Gaussianity/non-normal distributions etc.)
Guys, you are just starting a Wikibase-DataModel discussion completely unrelated to this task, which is about what is shown in diff views. Your comments will not be read here. Please focus and create a new [RFC] task if you want to suggest changes to the underlying data model. Thank you very much.
OK, sorry - I didn't realise that this was a more fundamental issue about the data model! I thought it was just being omitted from the discussion here. I'll raise it as a different ticket after doing a bit of background reading.
Decision: We'll go with the linked version Thiemo proposes. That is most consistent with what we do already, isn't too in-your-face and still gives people the option of seeing which unit is meant exactly. So we will show the label and link it to the item.