first

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/contexts.sh

@@ -30,18 +30,14 @@ hdfs dfs -copyFromLocal concepts.csv ${TMP}
 hdfs dfs -chmod -R 777 ${TMP}
 
 echo "Creating and populating impala tables"
-impala-shell -q "invalidate metadata"
+hive -e "create table ${TARGET_DB}.context (id string, name string) row format delimited fields terminated by ','"
-impala-shell -d ${TARGET_DB} -q "invalidate metadata"
+hive -e "create table ${TARGET_DB}.category (context string, id string, name string) row format delimited fields terminated by ','"
-impala-shell -q "create table ${TARGET_DB}.context (id string, name string) row format delimited fields terminated by ','"
+hive -e "create table ${TARGET_DB}.concept (category string, id string, name string) row format delimited fields terminated by ','"
-impala-shell -q "create table ${TARGET_DB}.category (context string, id string, name string) row format delimited fields terminated by ','"
+hive -e "load data inpath '${TMP}/contexts.csv' into table ${TARGET_DB}.context"
-impala-shell -q "create table ${TARGET_DB}.concept (category string, id string, name string) row format delimited fields terminated by ','"
+hive -e "load data inpath '${TMP}/categories.csv' into table ${TARGET_DB}.category"
-impala-shell -d ${TARGET_DB} -q "invalidate metadata"
+hive -e "load data inpath '${TMP}/concepts.csv' into table ${TARGET_DB}.concept"
-impala-shell -q "load data inpath '${TMP}/contexts.csv' into table ${TARGET_DB}.context"
-impala-shell -q "load data inpath '${TMP}/categories.csv' into table ${TARGET_DB}.category"
-impala-shell -q "load data inpath '${TMP}/concepts.csv' into table ${TARGET_DB}.concept"
 
 echo "Cleaning up"
-hdfs dfs -rm -f -r -skipTrash ${TMP}
 rm concepts.csv
 rm categories.csv
 rm contexts.csv

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/finalizedb.sh

@@ -10,9 +10,7 @@ export SOURCE=$1
 export SHADOW=$2
 
 echo "Updating shadow database"
-impala-shell -q "invalidate metadata"
+hive --database ${SOURCE} -e "show tables" | grep -v WARN | sed "s/^\(.*\)/analyze table ${SOURCE}.\1 compute statistics;/" | impala-shell -c -f -
-impala-shell -d ${SOURCE} -q "invalidate metadata"
-impala-shell -d ${SOURCE} -q "show tables" --delimited | sed "s/^\(.*\)/compute stats ${SOURCE}.\1;/" | impala-shell -c -f -
 impala-shell -q "create database if not exists ${SHADOW}"
 impala-shell -d ${SHADOW} -q "show tables" --delimited | sed "s/^/drop view if exists ${SHADOW}./" | sed "s/$/;/" | impala-shell -c -f -
 impala-shell -d ${SOURCE} -q "show tables" --delimited | sed "s/\(.*\)/create view ${SHADOW}.\1 as select * from ${SOURCE}.\1;/" | impala-shell -c -f -

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step10.sql

@@ -4,29 +4,28 @@
 ------------------------------------------------------------------------------------------------
 ------------------------------------------------------------------------------------------------
 CREATE OR REPLACE VIEW ${stats_db_name}.fundref AS
-SELECT *
+SELECT * FROM ${external_stats_db_name}.fundref;
-FROM ${external_stats_db_name}.fundref;
 
 CREATE OR REPLACE VIEW ${stats_db_name}.country AS
-SELECT *
+SELECT * FROM ${external_stats_db_name}.country;
-FROM ${external_stats_db_name}.country;
 
 CREATE OR REPLACE VIEW ${stats_db_name}.countrygdp AS
-SELECT *
+SELECT * FROM ${external_stats_db_name}.countrygdp;
-FROM ${external_stats_db_name}.countrygdp;
 
 CREATE OR REPLACE VIEW ${stats_db_name}.roarmap AS
-SELECT *
+SELECT * FROM ${external_stats_db_name}.roarmap;
-FROM ${external_stats_db_name}.roarmap;
 
 CREATE OR REPLACE VIEW ${stats_db_name}.rndexpediture AS
-SELECT *
+SELECT * FROM ${external_stats_db_name}.rndexpediture;
-FROM ${external_stats_db_name}.rndexpediture;
 
 CREATE OR REPLACE VIEW ${stats_db_name}.licenses_normalized AS
-SELECT *
+SELECT * FROM ${external_stats_db_name}.licenses_normalized;
-FROM ${external_stats_db_name}.licenses_normalized;
 
+create view ${stats_db_name}.rndexpenditure as
+select * from stats_ext.rndexpediture;
+
+create view ${stats_db_name}.issn_gold_oa_dataset as
+select * from stats_ext.issn_gold_oa_dataset;
 
 ------------------------------------------------------------------------------------------------
 ------------------------------------------------------------------------------------------------

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step11.sql

@@ -102,7 +102,7 @@ WHERE otherresearchproduct_tmp.id IN (SELECT distinct r.id
                                         AND pr.id = p.id
                                         AND to_date(r.date) - to_date(p.enddate) > 0);
 
-CREATE OR REPLACE VIEW ${stats_db_name}.project_results_publication AS
+CREATE OR REPLACE VIEW ${stats_db_name}.project_results_publication STORED AS PARQUET AS
 SELECT result_projects.id          AS result,
        result_projects.project     AS project_results,
        result.date                 as resultdate,

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step13.sql

@@ -5,7 +5,7 @@
 -- Sources related tables/views
 ------------------------------------------------------
 ------------------------------------------------------
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.publication_sources as 
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.publication_sources STORED AS PARQUET as
 SELECT p.id, case when d.id is null then 'other' else p.datasource end as datasource 
 FROM (
     SELECT  substr(p.id, 4) as id, substr(datasource, 4) as datasource 
@@ -16,7 +16,7 @@ LEFT OUTER JOIN
     from ${openaire_db_name}.datasource d 
     WHERE d.datainfo.deletedbyinference=false) d on p.datasource = d.id;
 
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.dataset_sources as 
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.dataset_sources STORED AS PARQUET as
 SELECT p.id, case when d.id is null then 'other' else p.datasource end as datasource 
 FROM (
     SELECT  substr(p.id, 4) as id, substr(datasource, 4) as datasource 
@@ -27,7 +27,7 @@ LEFT OUTER JOIN
     from ${openaire_db_name}.datasource d 
     WHERE d.datainfo.deletedbyinference=false) d on p.datasource = d.id;
     
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.software_sources as 
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.software_sources STORED AS PARQUET as
 SELECT p.id, case when d.id is null then 'other' else p.datasource end as datasource 
 FROM (
     SELECT  substr(p.id, 4) as id, substr(datasource, 4) as datasource 
@@ -38,7 +38,7 @@ LEFT OUTER JOIN
     from ${openaire_db_name}.datasource d 
     WHERE d.datainfo.deletedbyinference=false) d on p.datasource = d.id;
     
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.otherresearchproduct_sources as 
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.otherresearchproduct_sources STORED AS PARQUET as
 SELECT p.id, case when d.id is null then 'other' else p.datasource end as datasource 
 FROM (
     SELECT  substr(p.id, 4) as id, substr(datasource, 4) as datasource 
@@ -59,7 +59,7 @@ UNION ALL
 SELECT * FROM ${stats_db_name}.otherresearchproduct_sources;
 
 
-create table ${stats_db_name}.result_orcid as
+create table ${stats_db_name}.result_orcid STORED AS PARQUET as
 select distinct res.id, regexp_replace(res.orcid, 'http://orcid.org/' ,'') as orcid
 from (
     SELECT substr(res.id, 4) as id, auth_pid.value as orcid

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step14.sql

@@ -5,27 +5,27 @@
 -- Licences related tables/views
 ------------------------------------------------------
 ------------------------------------------------------
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.publication_licenses AS
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.publication_licenses STORED AS PARQUET AS
 SELECT substr(p.id, 4) as id, licenses.value as type 
 from ${openaire_db_name}.publication p LATERAL VIEW explode(p.instance.license) instances as licenses
 where licenses.value is not null and licenses.value != '' and p.datainfo.deletedbyinference=false;
 
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.dataset_licenses AS
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.dataset_licenses STORED AS PARQUET AS
 SELECT substr(p.id, 4) as id, licenses.value as type 
 from ${openaire_db_name}.dataset p LATERAL VIEW explode(p.instance.license) instances as licenses
 where licenses.value is not null and licenses.value != '' and p.datainfo.deletedbyinference=false;
 
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.software_licenses AS
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.software_licenses STORED AS PARQUET AS
 SELECT substr(p.id, 4) as id, licenses.value as type 
 from ${openaire_db_name}.software p LATERAL VIEW explode(p.instance.license) instances as licenses
 where licenses.value is not null and licenses.value != '' and p.datainfo.deletedbyinference=false;
 
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.otherresearchproduct_licenses AS
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.otherresearchproduct_licenses STORED AS PARQUET AS
 SELECT substr(p.id, 4) as id, licenses.value as type 
 from ${openaire_db_name}.otherresearchproduct p LATERAL VIEW explode(p.instance.license) instances as licenses
 where licenses.value is not null and licenses.value != '' and p.datainfo.deletedbyinference=false;
 
-CREATE VIEW IF NOT EXISTS ${stats_db_name}.result_licenses AS
+CREATE VIEW IF NOT EXISTS ${stats_db_name}.result_licenses STORED AS PARQUET AS
 SELECT * FROM ${stats_db_name}.publication_licenses
 UNION ALL
 SELECT * FROM ${stats_db_name}.dataset_licenses
@@ -34,11 +34,11 @@ SELECT * FROM ${stats_db_name}.software_licenses
 UNION ALL
 SELECT * FROM ${stats_db_name}.otherresearchproduct_licenses;
 
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.organization_pids AS 
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.organization_pids STORED AS PARQUET AS
 select substr(o.id, 4) as id, ppid.qualifier.classname as type, ppid.value as pid 
 from ${openaire_db_name}.organization o lateral view explode(o.pid) pids as ppid;
 
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.organization_sources as 
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.organization_sources STORED AS PARQUET as
 SELECT o.id, case when d.id is null then 'other' else o.datasource end as datasource 
 FROM (
     SELECT  substr(o.id, 4) as id, substr(instances.instance.key, 4) as datasource 
@@ -46,17 +46,4 @@ FROM (
     LEFT OUTER JOIN (
         SELECT substr(d.id, 4) id 
         from ${openaire_db_name}.datasource d 
-        WHERE d.datainfo.deletedbyinference=false) d on o.datasource = d.id;
+        WHERE d.datainfo.deletedbyinference=false) d on o.datasource = d.id;
-
--- ANALYZE TABLE ${stats_db_name}.publication_licenses COMPUTE STATISTICS;
--- ANALYZE TABLE ${stats_db_name}.publication_licenses COMPUTE STATISTICS FOR COLUMNS;
--- ANALYZE TABLE ${stats_db_name}.dataset_licenses COMPUTE STATISTICS;
--- ANALYZE TABLE ${stats_db_name}.dataset_licenses COMPUTE STATISTICS FOR COLUMNS;
--- ANALYZE TABLE ${stats_db_name}.software_licenses COMPUTE STATISTICS;
--- ANALYZE TABLE ${stats_db_name}.software_licenses COMPUTE STATISTICS FOR COLUMNS;
--- ANALYZE TABLE ${stats_db_name}.otherresearchproduct_licenses COMPUTE STATISTICS;
--- ANALYZE TABLE ${stats_db_name}.otherresearchproduct_licenses COMPUTE STATISTICS FOR COLUMNS;
--- ANALYZE TABLE ${stats_db_name}.organization_pids COMPUTE STATISTICS;
--- ANALYZE TABLE ${stats_db_name}.organization_pids COMPUTE STATISTICS FOR COLUMNS;
--- ANALYZE TABLE ${stats_db_name}.organization_sources COMPUTE STATISTICS;
--- ANALYZE TABLE ${stats_db_name}.organization_sources COMPUTE STATISTICS FOR COLUMNS;

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step15.sql

@@ -6,27 +6,27 @@
 ------------------------------------------------------
 ------------------------------------------------------
 
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.publication_refereed as
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.publication_refereed STORED AS PARQUET as
 select substr(r.id, 4) as id, inst.refereed.classname as refereed
 from ${openaire_db_name}.publication r lateral view explode(r.instance) instances as inst
 where r.datainfo.deletedbyinference=false;
 
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.dataset_refereed as
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.dataset_refereed STORED AS PARQUET as
 select substr(r.id, 4) as id, inst.refereed.classname as refereed
 from ${openaire_db_name}.dataset r lateral view explode(r.instance) instances as inst
 where r.datainfo.deletedbyinference=false;
 
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.software_refereed as
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.software_refereed STORED AS PARQUET as
 select substr(r.id, 4) as id, inst.refereed.classname as refereed
 from ${openaire_db_name}.software r lateral view explode(r.instance) instances as inst
 where r.datainfo.deletedbyinference=false;
 
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.otherresearchproduct_refereed as
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.otherresearchproduct_refereed STORED AS PARQUET as
 select substr(r.id, 4) as id, inst.refereed.classname as refereed
 from ${openaire_db_name}.otherresearchproduct r lateral view explode(r.instance) instances as inst
 where r.datainfo.deletedbyinference=false;
 
-CREATE VIEW IF NOT EXISTS ${stats_db_name}.result_refereed as
+CREATE VIEW IF NOT EXISTS ${stats_db_name}.result_refereed STORED AS PARQUET as
 select * from ${stats_db_name}.publication_refereed
 union all
 select * from ${stats_db_name}.dataset_refereed

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step15_5.sql

@@ -1,21 +1,21 @@
 -------------------------------------------
 --- Extra tables, mostly used by indicators
 
-create table ${stats_db_name}.result_projectcount as
+create table ${stats_db_name}.result_projectcount STORED AS PARQUET as
 select r.id, count(distinct p.id) as count
 from ${stats_db_name}.result r
 left outer join ${stats_db_name}.result_projects rp on rp.id=r.id
 left outer join ${stats_db_name}.project p on p.id=rp.project
 group by r.id;
 
-create table ${stats_db_name}.result_fundercount as
+create table ${stats_db_name}.result_fundercount STORED AS PARQUET as
 select r.id, count(distinct p.funder) as count
 from ${stats_db_name}.result r
 left outer join ${stats_db_name}.result_projects rp on rp.id=r.id
 left outer join ${stats_db_name}.project p on p.id=rp.project
 group by r.id;
 
-create table ${stats_db_name}.project_resultcount as
+create table ${stats_db_name}.project_resultcount STORED AS PARQUET as
 with rcount as (
     select p.id as pid, count(distinct r.id) as `count`, r.type as type
     from ${stats_db_name}.project p
@@ -29,8 +29,6 @@ select rcount.pid, sum(case when rcount.type='publication' then rcount.count els
 from rcount
 group by rcount.pid;
 
-create view ${stats_db_name}.rndexpenditure as select * from stats_ext.rndexpediture;
-
 create table ${stats_db_name}.result_instance stored as parquet as
 select distinct r.*
 from (
@@ -39,12 +37,10 @@ from (
          from ${openaire_db_name}.result r lateral view explode(r.instance) instances as inst lateral view explode(inst.pid) pids as p) r
 join ${stats_db_name}.result res on res.id=r.id;
 
-create table ${stats_db_name}.result_apc as
+create table ${stats_db_name}.result_apc STORED AS PARQUET as
 select r.id, r.amount, r.currency
 from (
          select substr(r.id, 4) as id, inst.processingchargeamount.value as amount, inst.processingchargecurrency.value as currency
          from ${openaire_db_name}.result r lateral view explode(r.instance) instances as inst) r
 join ${stats_db_name}.result res on res.id=r.id
-where r.amount is not null;
+where r.amount is not null;
-
-create view ${stats_db_name}.issn_gold_oa_dataset as select * from stats_ext.issn_gold_oa_dataset;

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step16_1-definitions.sql

@@ -3,20 +3,20 @@
 ----------------------------------------------------
 
 -- Peer reviewed:
-create table ${stats_db_name}.result_peerreviewed as
+create table ${stats_db_name}.result_peerreviewed stored as parquet as
 select r.id as id, case when doi.doi_from_crossref=1 and grey.grey_lit=0 then true else false end as peer_reviewed
 from ${stats_db_name}.result r
 left outer join ${stats_db_name}.indi_pub_doi_from_crossref doi on doi.id=r.id
 left outer join ${stats_db_name}.indi_pub_grey_lit grey on grey.id=r.id;
 
 -- Green OA:
-create table ${stats_db_name}.result_greenoa as
+create table ${stats_db_name}.result_greenoa stored as parquet as
 select r.id, case when green.green_oa=1 then true else false end as green
 from ${stats_db_name}.result r
 left outer join ${stats_db_name}.indi_pub_green_oa green on green.id=r.id;
 
 -- GOLD OA:
-create table ${stats_db_name}.result_gold as
+create table ${stats_db_name}.result_gold stored as parquet as
 select r.id, case when gold.gold_oa=1 then true else false end as gold
 from ${stats_db_name}.result r
          left outer join ${stats_db_name}.indi_pub_gold_oa gold on gold.id=r.id;

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step2.sql

@@ -40,13 +40,13 @@ SELECT substr(p.id, 4)                                            as id,
 from ${openaire_db_name}.publication p
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.publication_classifications AS
+CREATE TABLE ${stats_db_name}.publication_classifications STORED AS PARQUET AS
 SELECT substr(p.id, 4) as id, instancetype.classname as type
 from ${openaire_db_name}.publication p
          LATERAL VIEW explode(p.instance.instancetype) instances as instancetype
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.publication_concepts AS
+CREATE TABLE ${stats_db_name}.publication_concepts STORED AS PARQUET AS
 SELECT substr(p.id, 4) as id, case
     when contexts.context.id RLIKE '^[^::]+::[^::]+::.+$' then contexts.context.id
     when contexts.context.id RLIKE '^[^::]+::[^::]+$' then concat(contexts.context.id, '::other')
@@ -55,7 +55,7 @@ from ${openaire_db_name}.publication p
          LATERAL VIEW explode(p.context) contexts as context
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.publication_datasources as
+CREATE TABLE ${stats_db_name}.publication_datasources STORED AS PARQUET as
 SELECT p.id, case when d.id is null then 'other' else p.datasource end as datasource
 FROM (
          SELECT substr(p.id, 4) as id, substr(instances.instance.hostedby.key, 4) as datasource
@@ -66,30 +66,30 @@ FROM (
     from ${openaire_db_name}.datasource d
     WHERE d.datainfo.deletedbyinference = false) d on p.datasource = d.id;
 
-CREATE TABLE ${stats_db_name}.publication_languages AS
+CREATE TABLE ${stats_db_name}.publication_languages STORED AS PARQUET AS
 select substr(p.id, 4) as id, p.language.classname as language
 FROM ${openaire_db_name}.publication p
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.publication_oids AS
+CREATE TABLE ${stats_db_name}.publication_oids STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, oids.ids AS oid
 FROM ${openaire_db_name}.publication p
          LATERAL VIEW explode(p.originalid) oids AS ids
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.publication_pids AS
+CREATE TABLE ${stats_db_name}.publication_pids STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, ppid.qualifier.classname AS type, ppid.value as pid
 FROM ${openaire_db_name}.publication p
          LATERAL VIEW explode(p.pid) pids AS ppid
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.publication_topics as
+CREATE TABLE ${stats_db_name}.publication_topics STORED AS PARQUET as
 select substr(p.id, 4) AS id, subjects.subject.qualifier.classname AS TYPE, subjects.subject.value AS topic
 FROM ${openaire_db_name}.publication p
          LATERAL VIEW explode(p.subject) subjects AS subject
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.publication_citations AS
+CREATE TABLE ${stats_db_name}.publication_citations STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, xpath_string(citation.value, "//citation/id[@type='openaire']/@value") AS cites
 FROM ${openaire_db_name}.publication p
          lateral view explode(p.extrainfo) citations AS citation

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step3.sql

@@ -40,20 +40,20 @@ SELECT substr(d.id, 4)                                            AS id,
 FROM ${openaire_db_name}.dataset d
 WHERE d.datainfo.deletedbyinference = FALSE;
 
-CREATE TABLE ${stats_db_name}.dataset_citations AS
+CREATE TABLE ${stats_db_name}.dataset_citations STORED AS PARQUET AS
 SELECT substr(d.id, 4) AS id, xpath_string(citation.value, "//citation/id[@type='openaire']/@value") AS cites
 FROM ${openaire_db_name}.dataset d
          LATERAL VIEW explode(d.extrainfo) citations AS citation
 WHERE xpath_string(citation.value, "//citation/id[@type='openaire']/@value") != ""
   and d.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.dataset_classifications AS
+CREATE TABLE ${stats_db_name}.dataset_classifications STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, instancetype.classname AS type
 FROM ${openaire_db_name}.dataset p
          LATERAL VIEW explode(p.instance.instancetype) instances AS instancetype
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.dataset_concepts AS
+CREATE TABLE ${stats_db_name}.dataset_concepts STORED AS PARQUET AS
 SELECT substr(p.id, 4) as id, case
                                   when contexts.context.id RLIKE '^[^::]+::[^::]+::.+$' then contexts.context.id
                                   when contexts.context.id RLIKE '^[^::]+::[^::]+$' then concat(contexts.context.id, '::other')
@@ -62,7 +62,7 @@ from ${openaire_db_name}.dataset p
          LATERAL VIEW explode(p.context) contexts as context
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.dataset_datasources AS
+CREATE TABLE ${stats_db_name}.dataset_datasources STORED AS PARQUET AS
 SELECT p.id, case when d.id IS NULL THEN 'other' ELSE p.datasource END AS datasource
 FROM (
          SELECT substr(p.id, 4) as id, substr(instances.instance.hostedby.key, 4) AS datasource
@@ -74,24 +74,24 @@ FROM (
     FROM ${openaire_db_name}.datasource d
     WHERE d.datainfo.deletedbyinference = false) d ON p.datasource = d.id;
 
-CREATE TABLE ${stats_db_name}.dataset_languages AS
+CREATE TABLE ${stats_db_name}.dataset_languages STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, p.language.classname AS language
 FROM ${openaire_db_name}.dataset p
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.dataset_oids AS
+CREATE TABLE ${stats_db_name}.dataset_oids STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, oids.ids AS oid
 FROM ${openaire_db_name}.dataset p
          LATERAL VIEW explode(p.originalid) oids AS ids
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.dataset_pids AS
+CREATE TABLE ${stats_db_name}.dataset_pids STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, ppid.qualifier.classname AS type, ppid.value AS pid
 FROM ${openaire_db_name}.dataset p
          LATERAL VIEW explode(p.pid) pids AS ppid
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.dataset_topics AS
+CREATE TABLE ${stats_db_name}.dataset_topics STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, subjects.subject.qualifier.classname AS type, subjects.subject.value AS topic
 FROM ${openaire_db_name}.dataset p
          LATERAL VIEW explode(p.subject) subjects AS subject

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step4.sql

@@ -40,20 +40,20 @@ SELECT substr(s.id, 4)                                            as id,
 from ${openaire_db_name}.software s
 where s.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.software_citations AS
+CREATE TABLE ${stats_db_name}.software_citations STORED AS PARQUET AS
 SELECT substr(s.id, 4) as id, xpath_string(citation.value, "//citation/id[@type='openaire']/@value") AS cites
 FROM ${openaire_db_name}.software s
          LATERAL VIEW explode(s.extrainfo) citations as citation
 where xpath_string(citation.value, "//citation/id[@type='openaire']/@value") != ""
   and s.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.software_classifications AS
+CREATE TABLE ${stats_db_name}.software_classifications STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, instancetype.classname AS type
 FROM ${openaire_db_name}.software p
          LATERAL VIEW explode(p.instance.instancetype) instances AS instancetype
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.software_concepts AS
+CREATE TABLE ${stats_db_name}.software_concepts STORED AS PARQUET AS
 SELECT substr(p.id, 4) as id, case
                                   when contexts.context.id RLIKE '^[^::]+::[^::]+::.+$' then contexts.context.id
                                   when contexts.context.id RLIKE '^[^::]+::[^::]+$' then concat(contexts.context.id, '::other')
@@ -62,7 +62,7 @@ FROM ${openaire_db_name}.software p
          LATERAL VIEW explode(p.context) contexts AS context
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.software_datasources AS
+CREATE TABLE ${stats_db_name}.software_datasources STORED AS PARQUET AS
 SELECT p.id, CASE WHEN d.id IS NULL THEN 'other' ELSE p.datasource end as datasource
 FROM (
          SELECT substr(p.id, 4) AS id, substr(instances.instance.hostedby.key, 4) AS datasource
@@ -74,24 +74,24 @@ FROM (
     FROM ${openaire_db_name}.datasource d
     WHERE d.datainfo.deletedbyinference = false) d ON p.datasource = d.id;
 
-CREATE TABLE ${stats_db_name}.software_languages AS
+CREATE TABLE ${stats_db_name}.software_languages STORED AS PARQUET AS
 select substr(p.id, 4) AS id, p.language.classname AS language
 FROM ${openaire_db_name}.software p
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.software_oids AS
+CREATE TABLE ${stats_db_name}.software_oids STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, oids.ids AS oid
 FROM ${openaire_db_name}.software p
          LATERAL VIEW explode(p.originalid) oids AS ids
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.software_pids AS
+CREATE TABLE ${stats_db_name}.software_pids STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, ppid.qualifier.classname AS type, ppid.value AS pid
 FROM ${openaire_db_name}.software p
          LATERAL VIEW explode(p.pid) pids AS ppid
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.software_topics AS
+CREATE TABLE ${stats_db_name}.software_topics STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, subjects.subject.qualifier.classname AS type, subjects.subject.value AS topic
 FROM ${openaire_db_name}.software p
          LATERAL VIEW explode(p.subject) subjects AS subject

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step5.sql

@@ -40,18 +40,18 @@ FROM ${openaire_db_name}.otherresearchproduct o
 WHERE o.datainfo.deletedbyinference = FALSE;
 
 -- Otherresearchproduct_citations
-CREATE TABLE ${stats_db_name}.otherresearchproduct_citations AS
+CREATE TABLE ${stats_db_name}.otherresearchproduct_citations STORED AS PARQUET AS
 SELECT substr(o.id, 4) AS id, xpath_string(citation.value, "//citation/id[@type='openaire']/@value") AS cites
 FROM ${openaire_db_name}.otherresearchproduct o LATERAL VIEW explode(o.extrainfo) citations AS citation
 WHERE xpath_string(citation.value, "//citation/id[@type='openaire']/@value") != ""
   and o.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.otherresearchproduct_classifications AS
+CREATE TABLE ${stats_db_name}.otherresearchproduct_classifications STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, instancetype.classname AS type
 FROM ${openaire_db_name}.otherresearchproduct p LATERAL VIEW explode(p.instance.instancetype) instances AS instancetype
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.otherresearchproduct_concepts AS
+CREATE TABLE ${stats_db_name}.otherresearchproduct_concepts STORED AS PARQUET AS
 SELECT substr(p.id, 4) as id, case
                                   when contexts.context.id RLIKE '^[^::]+::[^::]+::.+$' then contexts.context.id
                                   when contexts.context.id RLIKE '^[^::]+::[^::]+$' then concat(contexts.context.id, '::other')
@@ -59,7 +59,7 @@ SELECT substr(p.id, 4) as id, case
 FROM ${openaire_db_name}.otherresearchproduct p LATERAL VIEW explode(p.context) contexts AS context
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.otherresearchproduct_datasources AS
+CREATE TABLE ${stats_db_name}.otherresearchproduct_datasources STORED AS PARQUET AS
 SELECT p.id, CASE WHEN d.id IS NULL THEN 'other' ELSE p.datasource END AS datasource
 FROM (SELECT substr(p.id, 4) AS id, substr(instances.instance.hostedby.key, 4) AS datasource
       from ${openaire_db_name}.otherresearchproduct p lateral view explode(p.instance) instances as instance
@@ -68,22 +68,22 @@ FROM (SELECT substr(p.id, 4) AS id, substr(instances.instance.hostedby.key, 4) A
                          from ${openaire_db_name}.datasource d
                          WHERE d.datainfo.deletedbyinference = false) d on p.datasource = d.id;
 
-CREATE TABLE ${stats_db_name}.otherresearchproduct_languages AS
+CREATE TABLE ${stats_db_name}.otherresearchproduct_languages STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, p.language.classname AS language
 FROM ${openaire_db_name}.otherresearchproduct p
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.otherresearchproduct_oids AS
+CREATE TABLE ${stats_db_name}.otherresearchproduct_oids STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, oids.ids AS oid
 FROM ${openaire_db_name}.otherresearchproduct p LATERAL VIEW explode(p.originalid) oids AS ids
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.otherresearchproduct_pids AS
+CREATE TABLE ${stats_db_name}.otherresearchproduct_pids STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, ppid.qualifier.classname AS type, ppid.value AS pid
 FROM ${openaire_db_name}.otherresearchproduct p LATERAL VIEW explode(p.pid) pids AS ppid
 where p.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.otherresearchproduct_topics AS
+CREATE TABLE ${stats_db_name}.otherresearchproduct_topics STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, subjects.subject.qualifier.classname AS type, subjects.subject.value AS topic
 FROM ${openaire_db_name}.otherresearchproduct p LATERAL VIEW explode(p.subject) subjects AS subject
 where p.datainfo.deletedbyinference = false;

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step6.sql

@@ -3,7 +3,7 @@
 -- Project table/view and Project related tables/views
 ------------------------------------------------------
 ------------------------------------------------------
-CREATE TABLE ${stats_db_name}.project_oids AS
+CREATE TABLE ${stats_db_name}.project_oids STORED AS PARQUET AS
 SELECT substr(p.id, 4) AS id, oids.ids AS oid
 FROM ${openaire_db_name}.project p LATERAL VIEW explode(p.originalid) oids AS ids;
 CREATE TABLE ${stats_db_name}.project_organizations AS
@@ -12,13 +12,13 @@ from ${openaire_db_name}.relation r
 WHERE r.reltype = 'projectOrganization'
   and r.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.project_results AS
+CREATE TABLE ${stats_db_name}.project_results STORED AS PARQUET AS
 SELECT substr(r.target, 4) AS id, substr(r.source, 4) AS result, r.datainfo.provenanceaction.classname as provenance
 FROM ${openaire_db_name}.relation r
 WHERE r.reltype = 'resultProject'
   and r.datainfo.deletedbyinference = false;
 
-create table ${stats_db_name}.project_classification as
+create table ${stats_db_name}.project_classification STORED AS PARQUET as
 select substr(p.id, 4) as id, class.h2020programme.code, class.level1, class.level2, class.level3
 from ${openaire_db_name}.project p
     lateral view explode(p.h2020classification) classifs as class
@@ -74,7 +74,7 @@ SELECT substr(p.id, 4)                                                 AS id,
 FROM ${openaire_db_name}.project p
 WHERE p.datainfo.deletedbyinference = false;
 
-create table ${stats_db_name}.funder as
+create table ${stats_db_name}.funder STORED AS PARQUET as
 select distinct xpath_string(fund, '//funder/id')        as id,
                 xpath_string(fund, '//funder/name')      as name,
                 xpath_string(fund, '//funder/shortname') as shortname

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step7.sql

@@ -123,13 +123,13 @@ UNION ALL
 SELECT *
 FROM ${stats_db_name}.otherresearchproduct_topics;
 
-CREATE TABLE ${stats_db_name}.result_organization AS
+CREATE TABLE ${stats_db_name}.result_organization STORED AS PARQUET AS
 SELECT substr(r.target, 4) AS id, substr(r.source, 4) AS organization
 FROM ${openaire_db_name}.relation r
 WHERE r.reltype = 'resultOrganization'
   and r.datainfo.deletedbyinference = false;
 
-CREATE TABLE ${stats_db_name}.result_projects AS
+CREATE TABLE ${stats_db_name}.result_projects STORED AS PARQUET AS
 select pr.result AS id, pr.id AS project, datediff(p.enddate, p.startdate) AS daysfromend, pr.provenance as provenance
 FROM ${stats_db_name}.result r
          JOIN ${stats_db_name}.project_results pr ON r.id = pr.result

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step8.sql

@@ -80,15 +80,15 @@ UPDATE ${stats_db_name}.datasource_tmp
 SET yearofvalidation=null
 WHERE yearofvalidation = '-1';
 
-CREATE TABLE ${stats_db_name}.datasource_languages AS
+CREATE TABLE ${stats_db_name}.datasource_languages STORED AS PARQUET AS
 SELECT substr(d.id, 4) AS id, langs.languages AS language
 FROM ${openaire_db_name}.datasource d LATERAL VIEW explode(d.odlanguages.value) langs AS languages;
 
-CREATE TABLE ${stats_db_name}.datasource_oids AS
+CREATE TABLE ${stats_db_name}.datasource_oids STORED AS PARQUET AS
 SELECT substr(d.id, 4) AS id, oids.ids AS oid
 FROM ${openaire_db_name}.datasource d LATERAL VIEW explode(d.originalid) oids AS ids;
 
-CREATE TABLE ${stats_db_name}.datasource_organizations AS
+CREATE TABLE ${stats_db_name}.datasource_organizations STORED AS PARQUET AS
 SELECT substr(r.target, 4) AS id, substr(r.source, 4) AS organization
 FROM ${openaire_db_name}.relation r
 WHERE r.reltype = 'datasourceOrganization'
@@ -96,11 +96,11 @@ WHERE r.reltype = 'datasourceOrganization'
 
 -- datasource sources:
 -- where the datasource info have been collected from.
-create table if not exists ${stats_db_name}.datasource_sources AS
+create table if not exists ${stats_db_name}.datasource_sources STORED AS PARQUET AS
 select substr(d.id, 4) as id, substr(cf.key, 4) as datasource
 from ${openaire_db_name}.datasource d lateral view explode(d.collectedfrom) cfrom as cf
 where d.datainfo.deletedbyinference = false;
 
-CREATE OR REPLACE VIEW ${stats_db_name}.datasource_results AS
+CREATE OR REPLACE VIEW ${stats_db_name}.datasource_results STORED AS PARQUET AS
 SELECT datasource AS id, id AS result
 FROM ${stats_db_name}.result_datasources;