Add validation for entered coordinate values in Geospatial metadata

[stevenwinship]

What this PR does / why we need it: Prevents the user input of invalid geospatial bounding box coordinates

Which issue(s) this PR closes:

• Closes Add validation for entered coordinate values in Geospatial metadata (bounding boxes, etc) #9547

Special notes for your reviewer:

Suggestions on how to test this: try adding non numbers/null values and values outside the norm for coordinates

Does this PR introduce a user interface change? If mockups are available, please link/include them here: error messages were added

Is there a release notes update needed for this change?: No

Additional documentation:

[coveralls]

coverage: 20.084% (+0.08%) from 20.007%
when pulling 2d6efc5 on 9547-validate-geospatial-metadata
into 410eb45 on develop.

[landreev]

@stevenwinship
OK, it is definitely doing the right thing now when all 4 values are entered - great!
It may not be doing what we want when some of the 4 boxes are left empty, however. 
As in, this is accepted:

... but this is not:

It is because of this line:
final Float returnVal = value != null ? Float.parseFloat(value) : min; // defaults to min if value is missing

it should not default to min. The idea was to default to the other value; i.e., if only X1 is entered, we wanted to assume that X2=X1 (what's being done in the current indexing code, in other words). So, the method verifyBoundingBoxCoordinatesWithinRange(...) would need to allow, and return null, then the top-level method would substitute this null with the value of the opposite end of the box - if entered...

Thinking about it though, it may have been a misguided idea, to continue allowing users to leave some of the values empty. It may be making the whole thing even more confusing than it already is.

So, I suggest that we handle it as follows:

1. We require that all 4 values are entered; however,
2. Let's explicitly allow the case where X1=X2 and/or Y1=Y2.

in other words, the following should be allowed:

i.e., <= instead of <.

Just to be a boring nerd, the rationale is that the above (which is roughly the spot of our office at IQSS) is a fairly valid "box". Because 42.3755 is not really a fixed point, but rather a longitude coordinate defined to +-0.0001 of a degree of precision - and that makes a fairly sizeable, non-empty chunk of surface.

[landreev]

I checked in a quick change to the geospatial block tsv and .properties. Added extended labels for the bbox fields on the edit form as discussed.
Also added this bit of formatting for the display form.
To show this:

instead of this:

Also added a 2 sentence release note telling them to update the block.

[landreev]

Approving.

[github-actions]

📦 Pushed preview images as

ghcr.io/gdcc/dataverse:9547-validate-geospatial-metadata

ghcr.io/gdcc/configbaker:9547-validate-geospatial-metadata

🚢 See on GHCR. Use by referencing with full name as printed above, mind the registry name.

[landreev]

@stevenwinship and everybody
I retested extra carefully the scenario where there are some preexisting datasets with invalid bounding box values, and you try to reindex everything. Confirmed that these datasets are now getting indexed properly (minus the bad bboxes that are dropped), instead of failing quietly.

We reindexed our prod. catalogue from scratch after the last deployment (6.0), with a new version of Solr installed (older versions were less picky about these bboxes being correctly defined). We discovered in the process that we had a few thousand automatically-generated datasets that had the North/South values swapped, and they all just quietly disappeared from the search catalog.

I'm going to declare this PR merged and done for real now.

[jggautier]

"Prod. catalogue" is Harvard Dataverse right, and the thousand automatically-generated datasets are datasets published in Harvard Dataverse? If so, what does it mean that datasets disappeared from the catalog? Sounds like they're not searchable anymore 😬

[landreev]

@jggautier There was a very long slack thread about the whole disaster, I'll look it up.

"Prod. catalogue" is Harvard Dataverse right, and the thousand automatically-generated datasets are datasets published in Harvard Dataverse? If so, what does it mean that datasets disappeared from the catalog? Sounds like they're not searchable anymore 😬

[jggautier]

Thanks. And sorry; I try to keep up with the conversation. Sometimes things go over my head