Dropdowns

Dropdowns

Every filter control in the query editor, and where its options come from.

Five of them ask the knowledge graph what to offer. The rest are fixed lists in the code. This page is the index; the five query backed ones each have their own page.

Populated by SPARQL

Substance Which PFAS chemicals to look for. Shown for Samples. Endpoint sawgraph with no region, federation once a state is picked. Falls back to seven hardcoded PFAS substances.

Material What the sample was taken from. Shown for Samples. Same endpoint switch as Substance. Grouped into Water, Biota, Solid Material, Air and Other, and those headings are themselves tickable and collapsible. Falls back to six material types.

Industry NAICS codes, as a searchable tree. Shown for Facilities. Endpoint fiokg. Falls back to eight codes covering the prebuilt questions.

Industry counts The facility counts shown on each industry tree node. Endpoint federation. No fallback, and nothing at all until a state is selected.

County Narrows to counties inside the selected state. Endpoint spatialkg. No fallback, on purpose.

All five follow the same shape. A hook in useDiscoveryQueries.ts calls a query builder in templates/regions.ts and runs it through executeSparql(). Results are cached by React Query with staleTime: Infinity, so each one fetches at most once per region per session.

Fixed lists in the code

These never touch an endpoint. Listed here so the coverage above is unambiguous.

Type, the entity a block is about. Samples, Facilities, Surface Water Bodies, Wells. EntityTypeSelector.tsx

Relationship. Near, Downstream of, Upstream from. RelationshipSelector.tsx

Distance, shown only when the relationship is Near. Zero to four miles, which map to S2 cell expansion hops rather than real distances. RelationshipSelector.tsx

State. All fifty are listed, but only the ones in AVAILABLE_STATE_FIPS are selectable. The rest render disabled, which is deliberate: it shows what the graph does not cover yet instead of pretending those states do not exist. RegionSelector.tsx, constants/regions.ts

Water Type. Lake or Pond, Swamp or Marsh, Reservoir. WaterBodyFilters.tsx

Well Source. Illinois Wells and Maine Wells. Each is tied to one state, and picking a source that contradicts the selected state raises a warning. WellFilters.tsx

Basemap, on the map itself. Five tile layers. basemaps.ts

Export. One entry, "As JPG". ExportDropdown.tsx

The two select components

Everything on this page is one of two controls.

FlatSelect is the flat one: a searchable list rendered through a portal, single or multi select, with support for disabled options and a loading state. It can show group headings, but they are plain text — you cannot tick or collapse them.

Multi select lists are headed by a Select all row, which carries the number of options it would tick, like Select all (79). That number counts what the checkbox actually toggles, so it excludes disabled options and narrows to the matches while a search term is typed. Counts are comma formatted throughout, both on the Select all row and on individual options.

HierarchicalSelect is the tree: every node carries a checkbox and, if it has children, a fold arrow. Ticking a parent takes everything under it, partial selections show an indeterminate box, and counts roll up from the leaves.

Which one a dropdown gets comes down to whether its options have a hierarchy worth navigating. Industry uses the tree because NAICS codes nest and a flat list of 5,596 is unusable. Material uses it because 171 options are lopsided — 131 of them are fish species — and being able to say "all water samples" in one click, or fold Biota away, is the difference between usable and not. Everything else is a flat list.

The two arrive at their tree differently. Industry has no explicit parent in the data, so the tree is derived from NAICS code prefixes. Material passes an explicit parent, because a material type URI says nothing about which bucket it belongs to. buildTree handles both.

Filters that are not dropdowns

For completeness, the Samples filter panel also has Min and Max concentration number inputs in ng/L, and an "Include non-detects" checkbox. None of them query anything; they are applied inside the pipeline queries.

The non-detects checkbox is ticked by default and moves the answer more than any dropdown on this page, because non-detects are 70 to 77% of the observations. It filters both which sample points appear and which rows appear inside their popups. See Include non-detects. SampleFilters.tsx