How to handle/determine the possible causes of memory allocation errors

[valentinedwv]

Is there any advice on memory allocations for servers?

[server]
PORT               = 7019
ACCESS_TOKEN       = ${data:NAME}_7643543846_6dMISzlPrD7i
MEMORY_FOR_QUERIES = 25G
CACHE_MAX_SIZE     = 20G
TIMEOUT            = 240s

since this is a container, probably need to shrink memory, a little bit.

Occasional errors we see from the qleverui

Error processing query
Tried to allocate 615.9 MB, but only 384 MB were available

logs:

2025-07-24 18:55:58.599 - ERROR: Aborted Operation
2025-07-24 18:55:58.615 - ERROR: Tried to allocate 615.9 MB, but only 384 MB were available

[ktk]

This basically means the particular query you did is not optimized yet in QLever so it needed more memory than you have available. In case this happens on an open data set, it would be great to have a Qleverfile & the query to test.

[valentinedwv]

Basically, we transform Schema.org Dataset their properties into RDF quads, and query.

Qlever

# removed from ui section: UI_CONFIG = eco

[data]
NAME              = facetsearch
SOURCES           = geocodes_examples geocodes_demo_datasets iris magic earthchem r2r opentopography iedadata wifire cchdo ssdbiodp osmc obis wodb emodnet
BASE_URL          = https://oss.geocodes-aws.earthcube.org/decoder/graphs/latest
GET_DATA_CMD      = for source in ${SOURCES}; do curl -sLo $${source}_release.nq -C - ${BASE_URL}/$${source}_release.nq; done
DESCRIPTION       = eco ${NAME} ${BASE_URL}
TEXT_DESCRIPTION  = eco release graph for ${NAME} sources: ${SOURCES}
FORMAT            = nq


[index]
INPUT_FILES     = *_release.nq
CAT_INPUT_FILES = cat ${INPUT_FILES}
#SETTINGS_JSON   = { "add-text-index": true, "ascii-prefixes-only": false, "num-triples-per-batch": 100000 }
SETTINGS_JSON   = { "add-text-index": true, "ascii-prefixes-only": false, "num-triples-per-batch": 100000, "vocabulary-type": "on-disk-compressed-geo-split" }
[server]
PORT               = 7019
ACCESS_TOKEN       = ${data:NAME}_7643543846_6dMISzlPrD7i
MEMORY_FOR_QUERIES = 25G
CACHE_MAX_SIZE     = 20G
TIMEOUT            = 240s

[runtime]
SYSTEM = native
IMAGE  = docker.io/adfreiburg/qlever:latest

[ui]
UI_CONFIG = default

query

PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
prefix schema: <http://schema.org/>
prefix sschema: <https://schema.org/>
PREFIX ql: <http://qlever.cs.uni-freiburg.de/builtin-functions/>
SELECT distinct ?subj ?pubname (GROUP_CONCAT(DISTINCT ?placename; SEPARATOR=", ") AS ?placenames)
        (GROUP_CONCAT(DISTINCT ?kwu; SEPARATOR=", ") AS ?kw)
        ?datep (GROUP_CONCAT(DISTINCT ?url; SEPARATOR=", ") AS ?disurl)
		(COUNT(?text) AS ?score) (SAMPLE(?text) AS ?example_text)
        ?name ?description (GROUP_CONCAT(DISTINCT ?resourceType_u; SEPARATOR=", ") as ?resourceType ) ?type ?g (MAX(?lat) as ?maxlat) (Min(?lat) as ?minlat) (MAX(?lon) as ?maxlon) (Min(?lon) as ?minlon)
 ?maxdepth ?minDepth ?temporalCoverage  ?variableMeasured 
 ?bbox
WHERE {
		    ?subj ?o ?item .
    ?text ql:contains-entity ?item .
    ?text ql:contains-word "water"
          graph ?g {
values ?sosType {
               sschema:Dataset
# sschema:DataCatalog
     schema:Dataset
# schema:DataCatalog
            }
              ?subj a ?sosType .
             ?subj schema:name|sschema:name ?name .
                       ?subj schema:description|sschema:description ?description . }
		values (?type ?resourceType_u) {
            (schema:Dataset "data")
            (sschema:Dataset "data")
            (schema:ResearchProject "researchProject") #BCODMO- project
            (sschema:ResearchProject "researchProject")
             (schema:SoftwareApplication "tool")
             (sschema:SoftwareApplication "tool")
             (schema:Person "person") #BCODMO- Person
             (sschema:Person "person")
              (schema:Event "event") #BCODMO- deployment
              (sschema:Event "event")
               (schema:Award "award") #BCODMO- Award
               (sschema:Award "award")
              (schema:DataCatalog "DataCatalog")
              (sschema:DataCatalog "DataCatalog")
            #(UNDEF "other") # assume it's data. At least we should get name.
          } ?subj a ?type .
            optional {?subj sschema:distribution/sschema:url|sschema:subjectOf/sschema:url|schema:distribution/schema:url|schema:subjectOf/schema:url ?url1 .}
            OPTIONAL {?subj schema:datePublished|sschema:datePublished ?datep1 .}
            OPTIONAL {?subj schema:dateCreated|sschema:dateCreated ?datec .}
            OPTIONAL {?subj schema:dateModified|sschema:dateModified ?datem .}
            OPTIONAL {?subj schema:temporalCoverage|sschema:temporalCoverage ?temporalCoverage .}
            OPTIONAL {?subj schema:publisher/schema:name|sschema:publisher/sschema:name|schema:publisher/schema:legalName|sschema:publisher/sschema:legalName ?pub_name .}
            OPTIONAL {?subj schema:spatialCoverage/schema:name|sschema:spatialCoverage/sschema:name|sschema:sdPublisher ?place_name .}
           # OPTIONAL {?subj schema:spatialCoverage/schema:geo/schema:latitude|sschema:spatialCoverage/sschema:geo/sschema:latitude ?lat .}
           # OPTIONAL {?subj sschema:spatialCoverage/sschema:geo ?geo .
			# ?geo a sschema:GeoCoordinates .
			# ?geo sschema:longitude ?lon .
			# ?geo sschema:latitude ?lat .}
			#OPTIONAL {?subj sschema:spatialCoverage/sschema:geo ?geo .
			# ?geo a sschema:GeoShape .
			# ?geo sschema:box ?box .}
            # bind ( COALESCE(?box) As ?bbox)
				 OPTIONAL {?subj sschema:keywords|sschema:keywords ?kw1 .}
# OPTIONAL {
# ?subj sschema:variableMeasured ?vm .
# ?vm a sschema:PropertyValue .
# ?vm sschema:name ?variableMeasured_a .
# }
				 # getting all names of variables. kills query
                 # if there is more than one location, then coallese to get the versions.
             OPTIONAL {?subj sschema:variableMeasured ?vm .
               ?vm a sschema:PropertyValue .
               ?vm sschema:name ?namedepth .
               FILTER (?namedepth IN ("depth", "CmpDep") ) .
                    ?vm sschema:maxValue ?maxdepth_d .
                  ?vm sschema:minValue ?minDepth_d
           }
		   bind (COALESCE (?namedepth) as ?variableMeasured) 
             bind ( COALESCE(?maxdepth_d) As ?maxdepth)
            bind ( COALESCE(?minDepth_d) As ?minDepth)
             # temporal coverage.
			 
             bind ( COALESCE(?kw1, "") As ?kwu)
            bind ( COALESCE(?url1, "") As ?url)
            # pick one date
            bind ( COALESCE(?datec,?datem,?datep1) As ?datep)
            BIND ( IF ( BOUND(?pub_name), ?pub_name, "No Publisher") as ?pubname ) .
            BIND ( IF ( BOUND(?place_name), ?place_name, "No Placenames") as ?placename ) .
            
        }
        GROUP BY ?subj ?pubname ?placename ?datep ?url ?name ?description ?type ?maxdepth ?minDepth ?temporalCoverage ?bbox ?g ?variableMeasured 
        ORDER BY DESC(?score)
LIMIT 10

[ktk]

Tnx, how big is this in total?

[valentinedwv]

Knowledge base index
Number of triples: 8,519,503
Number of subjects: 1,664,518
Number of predicates: 137
Number of objects: 2,680,336

Fulltext index
Number of distinct text records: 949,014
Number of distinct word occurrences: 10,482,035
Number of distinct entity occurrences: 949,014

qlever:/data# ls
Qleverfile facetsearch.index.sop.meta facetsearch.vocabulary.words.external.offsets
cchdo_release.nq facetsearch.index.spo facetsearch.vocabulary.words.internal
earthchem_release.nq facetsearch.index.spo.meta facetsearch.vocabulary.words.internal.ids
emodnet_release.nq facetsearch.internal.index.pos geocodes_demo_datasets_release.nq
facetsearch.index-log.txt facetsearch.internal.index.pos.meta geocodes_examples_release.nq
facetsearch.index.ops facetsearch.internal.index.pso iedadata_release.nq
facetsearch.index.ops.meta facetsearch.internal.index.pso.meta iris_release.nq
facetsearch.index.osp facetsearch.meta-data.json magic_release.nq
facetsearch.index.osp.meta facetsearch.server-log.txt obis_release.nq
facetsearch.index.patterns facetsearch.settings.json opentopography_release.nq
facetsearch.index.pos facetsearch.text.docsDB osmc_release.nq
facetsearch.index.pos.meta facetsearch.text.index r2r_release.nq
facetsearch.index.pso facetsearch.text.vocabulary ssdbiodp_release.nq
facetsearch.index.pso.meta facetsearch.vocabulary.codebooks wifire_release.nq
facetsearch.index.sop facetsearch.vocabulary.words.external wodb_release.nq

du .

2071652 .

qlever:/data# ls -l *.nq
-rw-r--r-- 1 root root 106846463 Jul 24 14:31 cchdo_release.nq
-rw-r--r-- 1 root root 62390835 Jul 24 14:30 earthchem_release.nq
-rw-r--r-- 1 root root 87295614 Jul 24 14:31 emodnet_release.nq
-rw-r--r-- 1 root root 326876 Jul 24 14:30 geocodes_demo_datasets_release.nq
-rw-r--r-- 1 root root 415943 Jul 24 14:30 geocodes_examples_release.nq
-rw-r--r-- 1 root root 382924549 Jul 24 14:30 iedadata_release.nq
-rw-r--r-- 1 root root 174340 Jul 24 14:30 iris_release.nq
-rw-r--r-- 1 root root 138960223 Jul 24 14:30 magic_release.nq
-rw-r--r-- 1 root root 104004191 Jul 24 14:31 obis_release.nq
-rw-r--r-- 1 root root 16943723 Jul 24 14:30 opentopography_release.nq
-rw-r--r-- 1 root root 5346795 Jul 24 14:31 osmc_release.nq
-rw-r--r-- 1 root root 570977502 Jul 24 14:30 r2r_release.nq
-rw-r--r-- 1 root root 176212178 Jul 24 14:31 ssdbiodp_release.nq
-rw-r--r-- 1 root root 50316214 Jul 24 14:30 wifire_release.nq
-rw-r--r-- 1 root root 5650172 Jul 24 14:31 wodb_release.nq

[ktk]

Thanks I can load the data and run the query. But:

• The query above returns very fast with no results, but a bunch of warnings for undefined variables
• If I uncomment the parts in the query, it also returns very fast but also no results.

Latest qlever docker image, latest qlever cli

[valentinedwv]

Sorry about the warnings. Migrating from one system, and trying to see where the bottlenecks are. Nice that we don't just dies on such issues.

This one returns data
Here is our instance with the query: https://qlever-iu.geocodes-aws-dev.earthcube.org/facetsearch/qRoEvJ

If you uncomment the
# OPTIONAL { ?subj sschema:variableMeasured/sschema:name ?variableMeasured_a .}
things blow up (which is not surprising, some 'datasets' can have dozens of 'variables')

PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
prefix schema: <http://schema.org/>
prefix sschema: <https://schema.org/>
PREFIX ql: <http://qlever.cs.uni-freiburg.de/builtin-functions/>
SELECT distinct ?subj ?pubname (GROUP_CONCAT(DISTINCT ?placename; SEPARATOR=", ") AS ?placenames)
    (GROUP_CONCAT(DISTINCT ?kwu; SEPARATOR=", ") AS ?kw)
    ?datep  (GROUP_CONCAT(DISTINCT ?url; SEPARATOR=", ") AS ?disurl)
    (COUNT(?text) AS ?score) (SAMPLE(?text) AS ?example_text)
    ?name ?description (GROUP_CONCAT(DISTINCT ?resourceType_u; SEPARATOR=", ") as ?resourceType )  ?type ?g   (MAX(?lat) as ?maxlat) (Min(?lat) as ?minlat) (MAX(?lon) as ?maxlon) (Min(?lon) as ?minlon)
    ?maxDepth ?minDepth ?temporalCoverage (GROUP_CONCAT(DISTINCT ?variableMeasured_a; SEPARATOR=", ") as ?variableMeasured  )
    ?bbox
WHERE {
    ?subj ?o ?item .

    ?text ql:contains-entity ?item .
    ?text ql:contains-word "water"
    graph ?g {
        values ?sosType {
            sschema:Dataset
            #               sschema:DataCatalog
            schema:Dataset
            #               schema:DataCatalog
        }
        ?subj    a    ?sosType .
        ?subj schema:name|sschema:name ?name .
        ?subj schema:description|sschema:description ?description . }
    values (?type ?resourceType_u) {
        (schema:Dataset "data")
        (sschema:Dataset "data")
        (schema:ResearchProject "researchProject") #BCODMO- project
        (sschema:ResearchProject  "researchProject")
        (schema:SoftwareApplication  "tool")
        (sschema:SoftwareApplication  "tool")
        (schema:Person  "person") #BCODMO- Person
        (sschema:Person  "person")
        (schema:Event  "event") #BCODMO- deployment
        (sschema:Event  "event")
        (schema:Award  "award") #BCODMO- Award
        (sschema:Award  "award")
        (schema:DataCatalog  "DataCatalog")
        (sschema:DataCatalog  "DataCatalog")
        #(UNDEF "other")  # assume it's data. At least we should get  name.
    } ?subj a ?type .
    optional {?subj sschema:distribution/sschema:url|sschema:subjectOf/sschema:url|schema:distribution/schema:url|schema:subjectOf/schema:url ?url1 .}
    OPTIONAL { ?subj schema:datePublished|sschema:datePublished ?datep1 . }
    OPTIONAL { ?subj schema:dateCreated|sschema:dateCreated ?datec . }
    OPTIONAL { ?subj schema:dateModified|sschema:dateModified ?datem . }
    OPTIONAL { ?subj schema:temporalCoverage|sschema:temporalCoverage ?temporalCoverage . }
    OPTIONAL { ?subj schema:publisher/schema:name|sschema:publisher/sschema:name|schema:publisher/schema:legalName|sschema:publisher/sschema:legalName  ?pub_name . }
    OPTIONAL { ?subj schema:spatialCoverage/schema:name|sschema:spatialCoverage/sschema:name|sschema:sdPublisher ?place_name . }
    # OPTIONAL { ?subj schema:spatialCoverage/schema:geo/schema:latitude|sschema:spatialCoverage/sschema:geo/sschema:latitude ?lat . }
    # OPTIONAL { ?subj sschema:spatialCoverage/sschema:geo ?geo .
    #         ?geo a sschema:GeoCoordinates .
    #		 ?geo sschema:longitude ?lon .
    #		 ?geo sschema:llatitude ?lat . }

    #OPTIONAL { ?subj sschema:spatialCoverage/sschema:geo ?geo .
    #         ?geo a sschema:GeoShape .
    #		 ?geo sschema:box ?box . }
    #     bind ( COALESCE(?box) As ?bbox)
    OPTIONAL { ?subj schema:keywords|sschema:keywords ?kw1 . }
 #   OPTIONAL { ?subj sschema:variableMeasured/sschema:name ?variableMeasured_a .}
    # if there is more than one location, then coallese to get the versions.
    OPTIONAL { ?subj sschema:variableMeasured ?vm .
              ?vm a  sschema:PropertyValue  .
              ?vm sschema:name ?namedepth  .
              FILTER (?namedepth IN ("depth", "CmpDep") ) .
              ?vm sschema:maxValue ?maxDepth_d .
              ?vm sschema:minValue ?minDepth_d
 }
    bind ( COALESCE(?maxDepth_d) As ?maxDepth)
    bind ( COALESCE(?minDepth_d) As ?minDepth)

    # temporal coverage.
    bind ( COALESCE(?kw1, "") As ?kwu)
    bind ( COALESCE(?url1) As ?url)
    # pick one date
    bind ( COALESCE(?datec,?datem,?datep1) As ?datep)
    BIND ( IF ( BOUND(?pub_name), ?pub_name, "No Publisher") as ?pubname ) .
    BIND ( IF ( BOUND(?place_name), ?place_name, "No Placenames") as ?placename ) .
}
GROUP BY ?subj ?pubname ?placename  ?datep ?url  ?name ?description ?type ?maxDepth ?minDepth ?temporalCoverage ?bbox ?g
ORDER BY DESC(?score)
LIMIT 100

[ktk]

OK then you get data and all is fine or what am I missing here?

By the way I don't get any data on my endpoint with this query.

[valentinedwv]

If I run the query with the optional variablemeasured, server stops and the rebuilds.
I don't get a result

for your data store, Is the data getting loaded ?

Have a different issue with recent builds complaining about a Unicode error.

[valentinedwv]

Just ran this, and got this in the log (on this container: latest@sha256:e95f3335f5417f43c00303a0d008bb54d9e9bc1c6407a7cb5547d931e98721bf)

2025-08-03 17:00:37.878 - ERROR: Sort operation was canceled, because time estimate exceeded remaining time by a factor of 3 Last operation: Sort (internal order) on ?subj

2025-08-03 17:00:30.469 - INFO: Requested media types of the result are: "application/sparql-results+json"
2025-08-03 17:00:30.541 - INFO: Query planning done in 97 ms
2025-08-03 17:00:31.320 - WARN: No timeout check has been performed for at least 93ms, should be at most 50ms. Checked at Join.cpp:322 at stage "Join on ?_QLever_internal_variable_qp_15".
2025-08-03 17:00:31.795 - WARN: No timeout check has been performed for at least 213ms, should be at most 50ms. Checked at Join.cpp:322 at stage "Join on ?_QLever_internal_variable_qp_15".
2025-08-03 17:00:32.662 - WARN: No timeout check has been performed for at least 373ms, should be at most 50ms. Checked at Join.cpp:322 at stage "Join on ?_QLever_internal_variable_qp_15".
2025-08-03 17:00:34.374 - WARN: No timeout check has been performed for at least 830ms, should be at most 50ms. Checked at Join.cpp:322 at stage "Join on ?_QLever_internal_variable_qp_15".
2025-08-03 17:00:37.606 - WARN: No timeout check has been performed for at least 1,648ms, should be at most 50ms. Checked at Join.cpp:322 at stage "Join on ?_QLever_internal_variable_qp_15".
2025-08-03 17:00:37.878 - ERROR: Sort operation was canceled, because time estimate exceeded remaining time by a factor of 3 Last operation: Sort (internal order) on ?subj

[ktk]

Can you give me the exact query that fails on your system. I used your QLeverfile and built the index accordingly.

[valentinedwv]

waiting for #2271 to build a container, or get merged... so I can use the newer version

Running against:
https://qlever-iu.geocodes-aws-dev.earthcube.org/facetsearch/

might need to dig a bit more. The UI that calling this gets a sort error
https://qlever-test.geocodes-aws-dev.earthcube.org/#/search/?q=cruise&searchExactMatch=false&resourceType=all

The qleverUI get a memory allocation error.

exception from qlevertest ... look for facessearch? network call.

exception
: 
"Sort operation was canceled, because time estimate exceeded remaining time by a factor of 3 Last operation: Sort (internal order) on ?subj"
query
: 
"PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>\nprefix schema: <http://schema.org/>\nprefix sschema: <https://schema.org/>\nPREFIX ql: <http://qlever.cs.uni-freiburg.de/builtin-functions/>\nSELECT distinct ?subj ?pubname (GROUP_CONCAT(DISTINCT ?placename; SEPARATOR=\", \") AS ?placenames)\n    (GROUP_CONCAT(DISTINCT ?kwu; SEPARATOR=\", \") AS ?kw)\n    ?datep  (GROUP_CONCAT(DISTINCT ?url; SEPARATOR=\", \") AS ?disurl)\n    (COUNT(?text) AS ?score) (SAMPLE(?text) AS ?example_text)\n    ?name ?description (GROUP_CONCAT(DISTINCT ?resourceType_u; SEPARATOR=\", \") as ?resourceType )  ?type ?g   (MAX(?lat) as ?maxlat) (Min(?lat) as ?minlat) (MAX(?lon) as ?maxlon) (Min(?lon) as ?minlon)\n    ?maxDepth ?minDepth ?temporalCoverage (GROUP_CONCAT(DISTINCT ?variableMeasured_a; SEPARATOR=\", \") as ?variableMeasured  )\n    ?bbox\nWHERE {\n    ?subj ?o ?item .\n\n    ?text ql:contains-entity ?item .\n    ?text ql:contains-word \"cruise\"\n    graph ?g {\n        values ?sosType {\n            sschema:Dataset\n            #               sschema:DataCatalog\n            schema:Dataset\n            #               schema:DataCatalog\n        }\n        ?subj    a    ?sosType .\n        ?subj schema:name|sschema:name ?name .\n        ?subj schema:description|sschema:description ?description . }\n    values (?type ?resourceType_u) {\n        (schema:Dataset \"data\")\n        (sschema:Dataset \"data\")\n        (schema:ResearchProject \"researchProject\") #BCODMO- project\n        (sschema:ResearchProject  \"researchProject\")\n        (schema:SoftwareApplication  \"tool\")\n        (sschema:SoftwareApplication  \"tool\")\n        (schema:Person  \"person\") #BCODMO- Person\n        (sschema:Person  \"person\")\n        (schema:Event  \"event\") #BCODMO- deployment\n        (sschema:Event  \"event\")\n        (schema:Award  \"award\") #BCODMO- Award\n        (sschema:Award  \"award\")\n        (schema:DataCatalog  \"DataCatalog\")\n        (sschema:DataCatalog  \"DataCatalog\")\n        #(UNDEF \"other\")  # assume it's data. At least we should get  name.\n    } ?subj a ?type .\n    optional {?subj sschema:distribution/sschema:url|sschema:subjectOf/sschema:url|schema:distribution/schema:url|schema:subjectOf/schema:url ?url1 .}\n    OPTIONAL {?subj schema:datePublished|sschema:datePublished ?datep1 .}\n    OPTIONAL {?subj schema:dateCreated|sschema:dateCreated ?datec .}\n    OPTIONAL {?subj schema:dateModified|sschema:dateModified ?datem .}\n    OPTIONAL {?subj schema:temporalCoverage|sschema:temporalCoverage ?temporalCoverage .}\n    OPTIONAL {?subj schema:publisher/schema:name|sschema:publisher/sschema:name|schema:publisher/schema:legalName|sschema:publisher/sschema:legalName  ?pub_name .}\n    OPTIONAL {?subj schema:spatialCoverage/schema:name|sschema:spatialCoverage/sschema:name|sschema:sdPublisher ?place_name .}\n    # OPTIONAL {?subj schema:spatialCoverage/schema:geo/schema:latitude|sschema:spatialCoverage/sschema:geo/sschema:latitude ?lat .}\n    # OPTIONAL {?subj sschema:spatialCoverage/sschema:geo ?geo .\n    #         ?geo a sschema:GeoCoordinates .\n    #\t\t ?geo sschema:longitude ?lon .\n    #\t\t ?geo sschema:llatitude ?lat .}\n\n    #OPTIONAL {?subj sschema:spatialCoverage/sschema:geo ?geo .\n    #         ?geo a sschema:GeoShape .\n    #\t\t ?geo sschema:box ?box .}\n    #     bind ( COALESCE(?box) As ?bbox)\n    OPTIONAL {?subj schema:keywords|sschema:keywords ?kw1 .}\n#    OPTIONAL {\n#        ?subj sschema:variableMeasured ?vm .\n#        ?vm a sschema:PropertyValue .\n#        ?vm sschema:name ?variableMeasured_a .\n#    }\n    # if there is more than one location, then coallese to get the versions.\n    OPTIONAL {?subj sschema:variableMeasured ?vm .\n              ?vm a  sschema:PropertyValue  .\n              ?vm sschema:name ?namedepth  .\n              FILTER (?namedepth IN (\"depth\", \"CmpDep\") ) .\n              ?vm sschema:maxValue ?maxDepth_d .\n              ?vm sschema:minValue ?minDepth_d\n    }\n    bind ( COALESCE(?maxDepth_d) As ?maxDepth)\n    bind ( COALESCE(?minDepth_d) As ?minDepth)\n\n    # temporal coverage.\n    bind ( COALESCE(?kw1, \"\") As ?kwu)\n    bind ( COALESCE(?url1) As ?url)\n    # pick one date\n    bind ( COALESCE(?datec,?datem,?datep1) As ?datep)\n    BIND ( IF ( BOUND(?pub_name), ?pub_name, \"No Publisher\") as ?pubname ) .\n    BIND ( IF ( BOUND(?place_name), ?place_name, \"No Placenames\") as ?placename ) .\n}\nGROUP BY ?subj ?pubname ?placename  ?datep ?url  ?name ?description ?type ?maxDepth ?minDepth ?temporalCoverage ?bbox ?g\nORDER BY DESC(?score)\nLIMIT 1000\n\n\n"

PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
prefix schema: <http://schema.org/>
prefix sschema: <https://schema.org/>
PREFIX ql: <http://qlever.cs.uni-freiburg.de/builtin-functions/>
SELECT distinct ?subj ?pubname (GROUP_CONCAT(DISTINCT ?placename; SEPARATOR=", ") AS ?placenames)
    (GROUP_CONCAT(DISTINCT ?kwu; SEPARATOR=", ") AS ?kw)
    ?datep  (GROUP_CONCAT(DISTINCT ?url; SEPARATOR=", ") AS ?disurl)
    (COUNT(?text) AS ?score) (SAMPLE(?text) AS ?example_text)
    ?name ?description (GROUP_CONCAT(DISTINCT ?resourceType_u; SEPARATOR=", ") as ?resourceType )  ?type ?g   (MAX(?lat) as ?maxlat) (Min(?lat) as ?minlat) (MAX(?lon) as ?maxlon) (Min(?lon) as ?minlon)
    ?maxDepth ?minDepth ?temporalCoverage (GROUP_CONCAT(DISTINCT ?variableMeasured_a; SEPARATOR=", ") as ?variableMeasured  )
    ?bbox
WHERE {
    ?subj ?o ?item .

    ?text ql:contains-entity ?item .
    ?text ql:contains-word "cruise"
    graph ?g {
        values ?sosType {
            sschema:Dataset
            #               sschema:DataCatalog
            schema:Dataset
            #               schema:DataCatalog
        }
        ?subj    a    ?sosType .
        ?subj schema:name|sschema:name ?name .
        ?subj schema:description|sschema:description ?description . }
    values (?type ?resourceType_u) {
        (schema:Dataset "data")
        (sschema:Dataset "data")
        (schema:ResearchProject "researchProject") #BCODMO- project
        (sschema:ResearchProject  "researchProject")
        (schema:SoftwareApplication  "tool")
        (sschema:SoftwareApplication  "tool")
        (schema:Person  "person") #BCODMO- Person
        (sschema:Person  "person")
        (schema:Event  "event") #BCODMO- deployment
        (sschema:Event  "event")
        (schema:Award  "award") #BCODMO- Award
        (sschema:Award  "award")
        (schema:DataCatalog  "DataCatalog")
        (sschema:DataCatalog  "DataCatalog")
        #(UNDEF "other")  # assume it's data. At least we should get  name.
    } ?subj a ?type .
    optional {?subj sschema:distribution/sschema:url|sschema:subjectOf/sschema:url|schema:distribution/schema:url|schema:subjectOf/schema:url ?url1 .}
    OPTIONAL {?subj schema:datePublished|sschema:datePublished ?datep1 .}
    OPTIONAL {?subj schema:dateCreated|sschema:dateCreated ?datec .}
    OPTIONAL {?subj schema:dateModified|sschema:dateModified ?datem .}
    OPTIONAL {?subj schema:temporalCoverage|sschema:temporalCoverage ?temporalCoverage .}
    OPTIONAL {?subj schema:publisher/schema:name|sschema:publisher/sschema:name|schema:publisher/schema:legalName|sschema:publisher/sschema:legalName  ?pub_name .}
    OPTIONAL {?subj schema:spatialCoverage/schema:name|sschema:spatialCoverage/sschema:name|sschema:sdPublisher ?place_name .}
    # OPTIONAL {?subj schema:spatialCoverage/schema:geo/schema:latitude|sschema:spatialCoverage/sschema:geo/sschema:latitude ?lat .}
    # OPTIONAL {?subj sschema:spatialCoverage/sschema:geo ?geo .
    #         ?geo a sschema:GeoCoordinates .
    #		 ?geo sschema:longitude ?lon .
    #		 ?geo sschema:llatitude ?lat .}

    #OPTIONAL {?subj sschema:spatialCoverage/sschema:geo ?geo .
    #         ?geo a sschema:GeoShape .
    #		 ?geo sschema:box ?box .}
    #     bind ( COALESCE(?box) As ?bbox)
    OPTIONAL {?subj schema:keywords|sschema:keywords ?kw1 .}
#    OPTIONAL {
#        ?subj sschema:variableMeasured ?vm .
#        ?vm a sschema:PropertyValue .
#        ?vm sschema:name ?variableMeasured_a .
#    }
    # if there is more than one location, then coallese to get the versions.
    OPTIONAL {?subj sschema:variableMeasured ?vm .
              ?vm a  sschema:PropertyValue  .
              ?vm sschema:name ?namedepth  .
              FILTER (?namedepth IN ("depth", "CmpDep") ) .
              ?vm sschema:maxValue ?maxDepth_d .
              ?vm sschema:minValue ?minDepth_d
    }
    bind ( COALESCE(?maxDepth_d) As ?maxDepth)
    bind ( COALESCE(?minDepth_d) As ?minDepth)

    # temporal coverage.
    bind ( COALESCE(?kw1, "") As ?kwu)
    bind ( COALESCE(?url1) As ?url)
    # pick one date
    bind ( COALESCE(?datec,?datem,?datep1) As ?datep)
    BIND ( IF ( BOUND(?pub_name), ?pub_name, "No Publisher") as ?pubname ) .
    BIND ( IF ( BOUND(?place_name), ?place_name, "No Placenames") as ?placename ) .
}
GROUP BY ?subj ?pubname ?placename  ?datep ?url  ?name ?description ?type ?maxDepth ?minDepth ?temporalCoverage ?bbox ?g
ORDER BY DESC(?score)
LIMIT 1000

[valentinedwv]

Think this was because I my wisdom of hearing that there were issues with named graphs, I moved a bunch of the optional items outside of the graph, reverted that.

I also move the type filter to the top, before the text search.

PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
prefix schema: <http://schema.org/>
prefix sschema: <https://schema.org/>
PREFIX ql: <http://qlever.cs.uni-freiburg.de/builtin-functions/>
SELECT distinct ?g  ?subj ?type  ?name ?description ?pubname (GROUP_CONCAT(DISTINCT ?placename; SEPARATOR=", ") AS ?placenames)
    (GROUP_CONCAT(DISTINCT ?kwu; SEPARATOR=", ") AS ?kw)
    ?datep  (GROUP_CONCAT(DISTINCT ?url; SEPARATOR=", ") AS ?disurl)
    (COUNT(?text) AS ?score) (SAMPLE(?text) AS ?example_text)
    (GROUP_CONCAT(DISTINCT ?resourceType_u; SEPARATOR=", ") as ?resourceType )   (MAX(?lat) as ?maxlat) (Min(?lat) as ?minlat) (MAX(?lon) as ?maxlon) (Min(?lon) as ?minlon)
    ?maxDepth ?minDepth ?temporalCoverage (GROUP_CONCAT(DISTINCT ?variableMeasured_a; SEPARATOR=", ") as ?variableMeasured  )
    ?bbox
WHERE {
    values ?sosType {
        sschema:Dataset
        #               sschema:DataCatalog
        schema:Dataset
        #               schema:DataCatalog
    }
    ?subj    a    ?sosType .
    values (?type ?resourceType_u) {
        (schema:Dataset "data")
        (sschema:Dataset "data")
        (schema:ResearchProject "researchProject") #BCODMO- project
        (sschema:ResearchProject  "researchProject")
        (schema:SoftwareApplication  "tool")
        (sschema:SoftwareApplication  "tool")
        (schema:Person  "person") #BCODMO- Person
        (sschema:Person  "person")
        (schema:Event  "event") #BCODMO- deployment
        (sschema:Event  "event")
        (schema:Award  "award") #BCODMO- Award
        (sschema:Award  "award")
        (schema:DataCatalog  "DataCatalog")
        (sschema:DataCatalog  "DataCatalog")
        #(UNDEF "other")  # assume it's data. At least we should get  name.
    } ?subj a ?type .
    ?subj ?o ?item .

    ?text ql:contains-entity ?item .
    ?text ql:contains-word "water"
    graph ?g {

        ?subj schema:name|sschema:name ?name .
        ?subj schema:description|sschema:description ?description .

        OPTIONAL {?subj schema:publisher/schema:legalName|sschema:publisher/sschema:legalName  ?legalName .}
        OPTIONAL {?subj schema:publisher/schema:name|sschema:publisher/sschema:name  ?publisher .}
        bind ( COALESCE(?publisher,?legalName, "No Publisher") As ?pubname)
        OPTIONAL {?subj schema:datePublished|sschema:datePublished ?datep1 .}
        OPTIONAL {?subj schema:dateCreated|sschema:dateCreated ?datec .}
        OPTIONAL {?subj schema:dateModified|sschema:dateModified ?datem .}
        bind ( COALESCE(?datec,?datem,?datep1) As ?datep)
        OPTIONAL {?subj schema:temporalCoverage|sschema:temporalCoverage ?temporalCoverage .}
        .
        OPTIONAL {?subj schema:spatialCoverage/schema:name|sschema:spatialCoverage/sschema:name|sschema:sdPublisher ?place_name .}
        BIND ( IF ( BOUND(?place_name), ?place_name, "No Placenames") as ?placename ) .
        optional {?subj sschema:distribution/sschema:url|sschema:subjectOf/sschema:url|schema:distribution/schema:url|schema:subjectOf/schema:url ?url .}
        OPTIONAL {?subj schema:keywords|sschema:keywords ?kwu .}

        # OPTIONAL {?subj schema:spatialCoverage/schema:geo/schema:latitude|sschema:spatialCoverage/sschema:geo/sschema:latitude ?lat .}
        # OPTIONAL {?subj sschema:spatialCoverage/sschema:geo ?geo .
        #         ?geo a sschema:GeoCoordinates .
        #		 ?geo sschema:longitude ?lon .
        #		 ?geo sschema:llatitude ?lat .}

        #OPTIONAL {?subj sschema:spatialCoverage/sschema:geo ?geo .
        #         ?geo a sschema:GeoShape .
        #		 ?geo sschema:box ?box .}
        #     bind ( COALESCE(?box) As ?bbox)

        #    OPTIONAL {
        #        ?subj sschema:variableMeasured ?vm .
        #        ?vm a sschema:PropertyValue .
        #        ?vm sschema:name ?variableMeasured_a .
        #    }
        # if there is more than one location, then coallese to get the versions.
        OPTIONAL {?subj sschema:variableMeasured ?vmd .
                  ?vmd a  sschema:PropertyValue  .
                  ?vmd sschema:name ?namedepth  .
                  FILTER (?namedepth IN ("depth", "CmpDep") ) .
                  ?vmd sschema:maxValue ?maxDepth_d .
                  ?vmd sschema:minValue ?minDepth_d
        bind ( COALESCE(?maxDepth_d) As ?maxDepth)
        bind ( COALESCE(?minDepth_d) As ?minDepth)
    }

}
GROUP BY ?g ?subj   ?name ?description ?type ?pubname ?placename  ?datep ?maxDepth ?minDepth ?temporalCoverage ?url ?bbox
ORDER BY DESC(?score)
LIMIT 1000