Skip to content

Dropdowns Material

Jump to bottom
github-actions[bot] edited this page Sep 25, 2026 · 2 revisions

Material dropdown

Picks what the sample was taken from: groundwater, soil, tissue, and so on. Multi select. Leaving it empty means "any material".

Material dropdown

What it does

It sits directly under the Substance dropdown in the query editor, and appears whenever the entity type is Samples, inside + Add Filters.

Like Substance, the list comes from the data rather than a fixed list, sorted by how many observations use each material, with the count in brackets: GROUNDWATER (609,496). Unfiltered it returns 171 distinct material types across the whole graph. Labels come straight from the graph, so they arrive in whatever case the source used. Maine data shouts in uppercase (GROUNDWATER, WASTE WATER), water quality portal data is title case (Tissue, Water). That inconsistency is upstream, not a rendering bug.

Those 171 are not shown as one long list. They sit under five headings, and each heading is itself a checkbox you can tick to take everything beneath it, and an arrow you can click to fold it away. That matters because the distribution is lopsided: 131 of the 171 are individual fish species, all under Biota.

Group Unfiltered Maine Cumberland County
Water 14 13 12
Biota 131 26 7
Solid Material 15 9 3
Air 1 — —
Other 10 6 3
Total 171 54 25

The list is region aware in the same way Substance is. Pick Cumberland County in Maine and it narrows from 171 entries to 25. Air disappears entirely below the national level: there are no air samples in Maine.

Where the five groups come from

The graph has no notion of "Water samples" as a category of material type. The grouping is computed in the discovery query itself, by MATERIAL_BUCKET_VALUES in regions.ts: each sample is left joined to the four direct subclasses of coso:MaterialSample, each tagged with a priority, and MIN(?prio) picks one.

coso:BiotaSample          1
coso:SolidMaterialSample  2
coso:WaterSample          3
coso:AirSample            4

The tie break exists because the data is not cleanly disjoint — 171 material types produce 178 type-to-bucket links, so seven of them sit under more than one subclass and MIN picks the lowest number. Anything that matches no subclass has no ?bucketPrio at all and the hook drops it into Other.

Two consequences worth holding on to:

  • Because MIN forces exactly one bucket per material type, the groups partition the list. That is what makes a group's count exactly the sum of its children's counts. Verified for Maine: Water shows 556,269, and its thirteen children sum to 556,269.
  • The priority order is a judgement call baked into the query, not something the ontology says. A material type sampled as both biota and solid material is filed under Biota.

Ticking a group changes nothing about the SPARQL. The group headings are a display concept with no URI behind them. Selecting one expands to its children's real material type URIs before anything is stored, so the query the pipeline sends is identical to the one you would get by ticking every child by hand. Internally the headings carry synthetic codes like __group:Water, and those are stripped before the selection is saved.

Two quirks worth knowing about before you trust what you see:

  • "Material" is looser than it sounds. The query keeps anything a sample was recorded as being made of, which in the water quality portal data includes biological taxa. Selecting Cumberland County surfaces Mytilus edulis, the blue mussel, sitting in the list next to GROUNDWATER. That is what the graph says, and the filter only excludes URIs outside the http://w3id.org/ namespace.
  • Material type URIs live in the root namespace, not the -data one. The August 2026 reload split the graph in two directions at once. Instance data — individual samples and sample points — moved to v2/me-egad-data# and v2/us-wqp-data#. The controlled vocabulary went the other way, out of -data and into the plain roots v1/me-egad# and v1/us-wqp#. A material type is vocabulary, so it is always the root form. Source of truth is pfas-kg, datasets/maine/egad/controlledVocab/sample_type.ttl, which declares @prefix me_egad: <http://w3id.org/sawgraph/v1/me-egad#> and defines 47 terms. Checked live on 2026-09-11: v1/me-egad#sampleMaterialType.GW has 46,110 triples and all three other spellings have none.

The SPARQL query

Endpoint is sawgraph with no region, federation once a state is picked. Unlike the Substance query, the two variants are not the same query with a pattern spliced in. They are written separately, and the difference matters.

Without a region

SELECT ?matType (SAMPLE(?_label) AS ?label) (COUNT(DISTINCT ?observation) AS ?num)
       (MIN(?prio) AS ?bucketPrio)
WHERE {
  ?observation rdf:type coso:ContaminantObservation ;
               coso:analyzedSample ?sample .
  ?sample coso:sampleOfMaterialType ?matType .
  OPTIONAL {
    ?sample rdf:type ?bucketClass .
    VALUES (?bucketClass ?prio) {
      (coso:BiotaSample 1)
      (coso:SolidMaterialSample 2)
      (coso:WaterSample 3)
      (coso:AirSample 4)
    }
  }
  OPTIONAL { ?matType rdfs:label ?_label . }
  FILTER(STRSTARTS(STR(?matType), "http://w3id.org/"))
} GROUP BY ?matType
ORDER BY DESC(?num) ?label

Line by line:

  • ?observation rdf:type coso:ContaminantObservation is every recorded measurement.
  • coso:analyzedSample ?sample is the physical sample that measurement was run on. This is the extra hop compared to Substance, which reads the chemical straight off the observation.
  • ?sample coso:sampleOfMaterialType ?matType is what that sample was made of.
  • OPTIONAL { rdfs:label } because not every material type has a label. Ones that do not fall back to the tail of their URI, done in the hook rather than in SPARQL.
  • The FILTER drops anything outside the w3id.org namespace, which is the crude way this query keeps out URIs from source vocabularies that were never meant to be shown. As of 2026-09-11 it excludes nothing at all — every material type in the graph is already under w3id.org — so it is dead weight rather than an active filter.
  • The OPTIONAL bucket block and MIN(?prio) are what produce the five groups, covered above. Leave them out and you get the same 171 rows, ungrouped.

Note that this variant never touches sample points at all. It walks observation to sample to material and stops.

With a region

SELECT ?matType (SAMPLE(?_label) AS ?label) (COUNT(DISTINCT ?observation) AS ?num)
       (MIN(?prio) AS ?bucketPrio)
WHERE {
  ?sp rdf:type coso:SamplePoint .
  ?sp spatial:connectedTo ?_region .
  ?_region rdf:type kwg-ont:AdministrativeRegion_3 ;
           kwg-ont:administrativePartOf+ ?_regionRoot .
  VALUES ?_regionRoot { kwgr:administrativeRegion.USA.23005 }
  ?observation rdf:type coso:ContaminantObservation ;
               coso:observedAtSamplePoint ?sp ;
               coso:analyzedSample ?sample .
  ?sample coso:sampleOfMaterialType ?matType .
  OPTIONAL {
    ?sample rdf:type ?bucketClass .
    VALUES (?bucketClass ?prio) {
      (coso:BiotaSample 1)
      (coso:SolidMaterialSample 2)
      (coso:WaterSample 3)
      (coso:AirSample 4)
    }
  }
  OPTIONAL { ?matType rdfs:label ?_label . }
  FILTER(STRSTARTS(STR(?matType), "http://w3id.org/"))
} GROUP BY ?matType
ORDER BY DESC(?num) ?label

The added part is the same region pattern the Substance dropdown uses, plus the coso:observedAtSamplePoint hop that connects an observation to a place. 23005 is Cumberland County, Maine. A whole state selection uses the two digit code instead, and the administrativePartOf+ walk covers everything underneath it either way.

Verified against the live endpoints on 2026-09-11. Unfiltered returns 171 material types, led by GROUNDWATER at 609,496 observations, then Tissue at 96,272 and Water at 95,779. Maine returns 54, Cumberland County 25.

How it is wired up

Cached under ['materialTypes', <region key>] with staleTime: Infinity and retry: 1, so one fetch per region per session. As with Substance, the fallback doubles as placeholderData, and the empty result handling is asymmetric: with no region the fallback is returned, with a region an empty list is.

The missing label fallback happens in the hook, at useDiscoveryQueries.ts:155, by splitting the URI on # or / and taking the tail.

If it looks wrong

One "Other" heading containing six title case materials, no counts. That is FALLBACK_MATERIAL_TYPES, so discovery failed. The giveaway is the shape as much as the content: the fallback entries carry no group, so every one of them defaults to "Other" and you get a single heading instead of five. Real results shout in uppercase, carry counts, and spread across Water, Biota, Solid Material, Air and Other.

A group's count does not match how many results you get. The number beside a heading is observations, summed from its children. What comes back from a run is sample points or samples. They are different units and will not line up.

A material disappeared from the list but results are still filtered by it. Chips are derived from the current option list, so a selection that falls out of that list after a region change stops being displayed while remaining in the query. Pick a Maine-only species, switch to Illinois, and the control reads "Any material type..." while quietly returning nothing. Clear the control and reselect. This predates the tree and applies to the Substance dropdown too.

Samples you can see on the map vanish as soon as you pick any material. Fourteen samples, carrying 177 observations, have no coso:sampleOfMaterialType at all. The filter matches on that predicate, so anything lacking it is excluded by every possible selection. Nothing to fix in the app; the upstream data simply does not say what those samples were made of.

A hardcoded material URI returns an empty map and no error. This is what broke the "PFHpA Groundwater Samples Downstream from Facilities in Cumberland County" card until ef0c40b. It pinned v1/me-egad-data#sampleMaterialType.GW, which matches zero triples. The filter that gets built is VALUES ?matType { <dead-iri> }, which is perfectly valid SPARQL — the endpoint answers 200 with no rows, and the app has no way to tell that apart from "there is genuinely nothing here". A dead URI does not error, it just quietly removes everything. If a selection returns nothing, count the triples on the URI itself before looking anywhere else.

Species names in a materials list. Genuinely in the graph, under http://w3id.org/sawgraph/v1/us-wqp#biologicalTaxon.*. The namespace filter does not exclude them. If they should not be offered, the filter needs to be narrower than http://w3id.org/.

Labels that are URI tails, like sampleMaterialType.WW. That material has no rdfs:label, so the hook fell back to the URI tail. As of 2026-09-11 this does not happen: all 171 material types carry a label, and .WW in particular is WASTE WATER. If you do see a raw tail, a term has been added upstream without a label.

Empty after picking a county. No observations recorded there. Expected.

See also

SAWGraph Explorer

Home

Dropdowns

Samples

In the repo

Clone this wiki locally