Merge pull request '[stats wf] various fixes, organization ids for inst. dashboard' (#205) from antonis.lempesis/dnet-hadoop:beta into beta

Reviewed-on: D-Net/dnet-hadoop#205

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

@@ -49,8 +49,5 @@ select * from openaire_prod_usage_stats.views_stats;
 -- Creation date of the database
 ------------------------------------------------------------------------------------------------
 ------------------------------------------------------------------------------------------------
-create table ${stats_db_name}.creation_date as
+create table ${stats_db_name}.creation_date STORED AS PARQUET as
 select date_format(current_date(), 'dd-MM-yyyy') as date;
---
--- ANALYZE TABLE ${stats_db_name}.creation_date COMPUTE STATISTICS;
--- ANALYZE TABLE ${stats_db_name}.creation_date COMPUTE STATISTICS FOR COLUMNS;

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

@@ -113,7 +113,4 @@ FROM ${stats_db_name}.result_projects,
      ${stats_db_name}.project
 WHERE result_projects.id = result.id
   AND result.type = 'publication'
-  AND project.id = result_projects.project;
+  AND project.id = result_projects.project;
-
--- ANALYZE TABLE ${stats_db_name}.project COMPUTE STATISTICS;
--- ANALYZE TABLE ${stats_db_name}.project COMPUTE STATISTICS FOR COLUMNS;

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 and d.datainfo.invisible = 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 and d.datainfo.invisible = 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 and d.datainfo.invisible = 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

@@ -5,22 +5,22 @@
 -- 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 and p.datainfo.invisible = 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 and p.datainfo.invisible = 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 and p.datainfo.invisible = 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 and p.datainfo.invisible = FALSE;
@@ -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 and d.datainfo.invisible = FALSE) d on o.datasource = d.id;
+        WHERE d.datainfo.deletedbyinference=false and d.datainfo.invisible = 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

@@ -6,22 +6,22 @@
 ------------------------------------------------------
 ------------------------------------------------------
 
-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 and r.datainfo.invisible = 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 and r.datainfo.invisible = 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 and r.datainfo.invisible = 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 and r.datainfo.invisible = FALSE;

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

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step16-createIndicatorsTables.sql

@@ -265,19 +265,19 @@ left outer join (
 
 create table indi_org_openess stored as parquet as
 WITH datasets_oa as (
-    SELECT ro.organization, count(dg.id) no_oadatasets FROM indi_datasets_gold_oa_new dg
+    SELECT ro.organization, count(dg.id) no_oadatasets FROM indi_datasets_gold_oa dg
     join openaire_prod_stats.result_organization ro on dg.id=ro.id
     join openaire_prod_stats.dataset ds on dg.id=ds.id
     WHERE dg.is_gold=1
     group by ro.organization),
 software_oa as (
-    SELECT ro.organization, count(dg.id) no_oasoftware FROM indi_software_gold_oa_new dg
+    SELECT ro.organization, count(dg.id) no_oasoftware FROM indi_software_gold_oa dg
     join openaire_prod_stats.result_organization ro on dg.id=ro.id
     join openaire_prod_stats.software ds on dg.id=ds.id
     WHERE dg.is_gold=1
     group by ro.organization),
 pubs_oa as (
-    SELECT ro.organization, count(dg.id) no_oapubs FROM indi_pub_gold_oa_new dg
+    SELECT ro.organization, count(dg.id) no_oapubs FROM indi_pub_gold_oa dg
     join openaire_prod_stats.result_organization ro on dg.id=ro.id
     join openaire_prod_stats.publication ds on dg.id=ds.id
     where dg.is_gold=1

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.is_gold=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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 and d.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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/step20-createMonitorDB.sql

@@ -22,7 +22,16 @@ create table TARGET.result stored as parquet as
             'openorgs____::b84450f9864182c67b8611b5593f4250',
             'openorgs____::d41cf6bd4ab1b1362a44397e0b95c975',
             'openorgs____::eadc8da90a546e98c03f896661a2e4d4',
-            'openorgs____::d2a09b9d5eabb10c95f9470e172d05d2') )) foo;
+            'openorgs____::d2a09b9d5eabb10c95f9470e172d05d2',
+            'openorgs____::d169c7407dd417152596908d48c11460',
+            'openorgs____::1ec924b1759bb16d0a02f2dad8689b21',
+            'openorgs____::2fb1e47b4612688d9de9169d579939a7',
+            'openorgs____::759d59f05d77188faee99b7493b46805',
+            'openorgs____::cad284878801b9465fa51a95b1d779db',
+            'openorgs____::eadc8da90a546e98c03f896661a2e4d4',
+            'openorgs____::c0286313e36479eff8676dba9b724b40'
+            -- ,'openorgs____::c80a8243a5e5c620d7931c88d93bf17a' -- Paris Diderot
+            ) )) foo;
 compute stats TARGET.result;
 
 create table TARGET.result_citations stored as parquet as select * from SOURCE.result_citations orig where exists (select 1 from TARGET.result r where r.id=orig.id);
@@ -52,7 +61,7 @@ compute stats TARGET.result_languages;
 create table TARGET.result_licenses stored as parquet as select * from SOURCE.result_licenses orig where exists (select 1 from TARGET.result r where r.id=orig.id);
 compute stats TARGET.result_licenses;
 
-create table TARGET.licenses_normalized as select * from SOURCE.licenses_normalized;
+create table TARGET.licenses_normalized STORED AS PARQUET as select * from SOURCE.licenses_normalized;
 
 create table TARGET.result_oids stored as parquet as select * from SOURCE.result_oids orig where exists (select 1 from TARGET.result r where r.id=orig.id);
 compute stats TARGET.result_oids;
@@ -83,7 +92,7 @@ compute stats TARGET.result_topics;
 
 create view TARGET.foo1 as select * from SOURCE.result_result rr where rr.source in (select id from TARGET.result);
 create view TARGET.foo2 as select * from SOURCE.result_result rr where rr.target in (select id from TARGET.result);
-create table TARGET.result_result as select distinct * from (select * from TARGET.foo1 union all select * from TARGET.foo2) foufou;
+create table TARGET.result_result STORED AS PARQUET as select distinct * from (select * from TARGET.foo1 union all select * from TARGET.foo2) foufou;
 drop view TARGET.foo1;
 drop view TARGET.foo2;
 compute stats TARGET.result_result;
@@ -157,13 +166,13 @@ create view TARGET.indi_org_openess as select * from SOURCE.indi_org_openess;
 create table TARGET.indi_pub_hybrid_oa_with_cc stored as parquet as select * from SOURCE.indi_pub_hybrid_oa_with_cc orig where exists (select 1 from TARGET.result r where r.id=orig.id);
 compute stats TARGET.indi_pub_hybrid_oa_with_cc;
 
-create table TARGET.indi_pub_downloads stored as parquet as select * from SOURCE.indi_pub_downloads orig where exists (select 1 from TARGET.result r where r.id=orig.id);
+create table TARGET.indi_pub_downloads stored as parquet as select * from SOURCE.indi_pub_downloads orig where exists (select 1 from TARGET.result r where r.id=orig.result_id);
 compute stats TARGET.indi_pub_downloads;
-create table TARGET.indi_pub_downloads_datasource stored as parquet as select * from SOURCE.indi_pub_downloads_datasource orig where exists (select 1 from TARGET.result r where r.id=orig.id);
+create table TARGET.indi_pub_downloads_datasource stored as parquet as select * from SOURCE.indi_pub_downloads_datasource orig where exists (select 1 from TARGET.result r where r.id=orig.result_id);
 compute stats TARGET.indi_pub_downloads_datasource;
-create table TARGET.indi_pub_downloads_year stored as parquet as select * from SOURCE.indi_pub_downloads_year orig where exists (select 1 from TARGET.result r where r.id=orig.id);
+create table TARGET.indi_pub_downloads_year stored as parquet as select * from SOURCE.indi_pub_downloads_year orig where exists (select 1 from TARGET.result r where r.id=orig.result_id);
 compute stats TARGET.indi_pub_downloads_year;
-create table TARGET.indi_pub_downloads_datasource_year stored as parquet as select * from SOURCE.indi_pub_downloads_datasource_year orig where exists (select 1 from TARGET.result r where r.id=orig.id);
+create table TARGET.indi_pub_downloads_datasource_year stored as parquet as select * from SOURCE.indi_pub_downloads_datasource_year orig where exists (select 1 from TARGET.result r where r.id=orig.result_id);
 compute stats TARGET.indi_pub_downloads_datasource_year;
 
 --denorm

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 and d.datainfo.invisible=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 and d.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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 and d.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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 and s.datainfo.invisible=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 and s.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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 and d.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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 and o.datainfo.invisible=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 and o.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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 and d.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=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 and p.datainfo.invisible=false;

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

@@ -3,24 +3,24 @@
 -- 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
 where p.datainfo.deletedbyinference=false  and p.datainfo.invisible=false;
 
-CREATE TABLE ${stats_db_name}.project_organizations AS
+CREATE TABLE ${stats_db_name}.project_organizations STORED AS PARQUET AS
 SELECT substr(r.source, 4) AS id, substr(r.target, 4) AS organization
 from ${openaire_db_name}.relation r
 WHERE r.reltype = 'projectOrganization'
   and r.datainfo.deletedbyinference = false and r.datainfo.invisible=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 and r.datainfo.invisible=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
@@ -76,7 +76,7 @@ SELECT substr(p.id, 4)                                                 AS id,
 FROM ${openaire_db_name}.project p
 WHERE p.datainfo.deletedbyinference = false and p.datainfo.invisible=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 and r.datainfo.invisible=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

@@ -80,24 +80,24 @@ 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
 where d.datainfo.deletedbyinference=false and d.datainfo.invisible=false;
 
-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
 where d.datainfo.deletedbyinference=false and d.datainfo.invisible=false;
 
-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' and r.datainfo.deletedbyinference = false and r.datainfo.invisible=false;
 
 -- 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 and d.datainfo.invisible=false;

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step9.sql

@@ -3,7 +3,7 @@
 -- Organization table/view and Organization related tables/views
 ----------------------------------------------------------------
 ----------------------------------------------------------------
-CREATE TABLE IF NOT EXISTS ${stats_db_name}.organization AS
+CREATE TABLE IF NOT EXISTS ${stats_db_name}.organization STORED AS PARQUET AS
 SELECT substr(o.id, 4)        as id,
        o.legalname.value      as name,
        o.legalshortname.value as legalshortname,
@@ -11,13 +11,10 @@ SELECT substr(o.id, 4)        as id,
 FROM ${openaire_db_name}.organization o
 WHERE o.datainfo.deletedbyinference = FALSE and o.datainfo.invisible = FALSE;
 
-CREATE OR REPLACE VIEW ${stats_db_name}.organization_datasources AS
+CREATE OR REPLACE VIEW ${stats_db_name}.organization_datasources STORED AS PARQUET AS
 SELECT organization AS id, id AS datasource
 FROM ${stats_db_name}.datasource_organizations;
 
-CREATE OR REPLACE VIEW ${stats_db_name}.organization_projects AS
+CREATE OR REPLACE VIEW ${stats_db_name}.organization_projects STORED AS PARQUET AS
 SELECT id AS project, organization as id
-FROM ${stats_db_name}.project_organizations;
+FROM ${stats_db_name}.project_organizations;
-
--- ANALYZE TABLE ${stats_db_name}.organization COMPUTE STATISTICS;
--- ANALYZE TABLE ${stats_db_name}.organization COMPUTE STATISTICS FOR COLUMNS;