Improve sidebar filters logic

[pzwsk]

When clicking on the facet filter coastal flood it returns datasets that are not coastal flood. I initially thought it was because it also returned flood hazard type but it looks like the issue is more generic to all hazard types.

https://jkan.riskdatalibrary.org/datasets/?hazard_type=coastal-flood

[matamadio]

This is the expected behaviour enforced by last update.
The selection returns 1 coastal hazard dataset + all exposure datasets.
To see only the hazard layer, you need to first select Hazard as category.
I understand the intentions but I don't fully agree with this - it's not intuitive, it looks like a bug.

[matamadio]

Another counter-intuitive effect depending on the logic mentioned above:

Hazard type filter includes "None (27)", which returns exposure items; or an empty page if any hazard type is selected.

Another usability issue is the lenght of the sidebar in relation to the refresh on filter selection: the page refreshes to the top at any filter change, so you need to scroll down again to find the filter you were selecting.

We need to rethink the filtering logic to make search more intuitive and practical.

[lydiascarf]

I'd like to clarify and summarize the ask here. What I'm getting is that:

• exposure datasets should not come back if a hazard type is selected
  • should the hazard category filter be automatically applied as well whenever a hazard type is selected?
• changing filters should not lead to a full page refresh

Would making these two changes resolve this issue?

[matamadio]

changing filters should not lead to a full page refresh

Defenetely.

should the hazard category filter be automatically applied as well whenever a hazard type is selected?

No. Hazard type can also refer to Vuln and Loss.

exposure datasets should not come back if a hazard type is selected

Possibly, since most exposure dataset are hazard-independent. The filter results interpretation issue occurs when no risk category is selected: then any exposure could be applied to any hazard type.
I do not have an optimal solution in mind, proposing three alternatives:

1. Easy auto-select of pertinent risk data types

Upon selecting an hazard type, all three risk categories hazard, vuln and loss are automatically selected for the results visualisation.

2. Visual distinction with contextual labeling

Keep showing exposure data but make it visually distinct: add a badge/tag to exposure datasets: "Generic - Applies to all hazards"
Use a different visual style: lighter background, subtle border, or icon

So that results are grouped into sections:

• "Direct matches (Hazard type: Earthquake)"
• "Generic exposure data (Applicable to all hazards)"

Example:

• No hazard filter active: Show all data normally
• Hazard filter active:
  • Show filtered hazard/vulnerability/losses prominently
  • Show exposure in a collapsible "Generic datasets" section (collapsed by default)

Results (32):
- Hazard: 8
- Vulnerability: 12  
- Losses: 5
- Exposure: 7 (generic - applies to all hazards)

3. Smart Filtering with Toggle

Add a user-controlled checkbox: "☑ Include generic exposure data" (on by default)

---

These changes needs to be considered in the broader context of improving the search and filters results mapping using AI - see #31

[lydiascarf]

1. Easy auto-select of pertinent risk data types

Upon selecting an hazard type, all three risk categories hazard, vuln and loss are automatically selected for the results visualisation.

Option 1 would be the quickest way to resolve the issue and it feels like the most intuitive solution to me. We can always use styling or a tooltip to further clarify how it works, but I think this change should happen either way, because this way the filter bank will reflect the state accurately. Filtering to datasets with any particular hazard type is equivalent to filtering to hazard, vuln, and loss filters. I do think a tooltip would be helpful to explain why the automatic selection happens, though.

These changes needs to be considered in the broader context of improving the search and filters results mapping using AI.

I would love to talk more about this point tomorrow in our call tomorrow just so I can make sure I'm understanding how the AI features might impact this design.

I'll work on this today.

[lydiascarf]

Just posted a PR with the filter changes.

Fixing the refresh issue will take a bit more of a refactor. Currently these filters are anchor tags following hrefs, so i'll have to develop some JavaScript to update the filters, DOM, and URL manually. I can start this right away, but it doesn't need to hold up the simpler fixes.

I tried something involving fragment ids where you can navigate to a link but scroll to a specific element (e.g. the hazard type filter bank), but this sometimes doesn't work if the javascript hasn't created the targeted element yet.

[lydiascarf]

Because of #52, we are going to remove the tooltip, remove the disabling, and filtering for multiple risk data type.

[lydiascarf]

We want to:

• sort filters list items by count for country, license, project
• remove none for hazard type
• consider schema version