D-Net
/
dnet-hadoop
14
5
Fork 11
Code Issues 14 Pull Requests 12 Projects 2 Releases 2 Wiki Activity

Improve performance of transferring the stats-DBs to another cluster and querying the DBs' tables, by ordering Spark to create up to 100 files per table, instead of thousands.

This commit is contained in:
Lampros Smyrnaios 2024-07-03 20:15:33 +03:00
parent 7b7dd32ad5
commit ce0aee21cc
8 changed files with 145 additions and 146 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step13.sql
View File

@ -10,7 +10,7 @@ set mapred.job.queue.name=analytics; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.publication_sources purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.publication_sources purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.publication_sources STORED AS PARQUET 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 SELECT /*+ COALESCE(100) */ p.id, case when d.id is null then 'other' else p.datasource end as datasource
FROM ( FROM (
SELECT substr(p.id, 4) as id, substr(datasource, 4) as datasource SELECT substr(p.id, 4) as id, substr(datasource, 4) as datasource
from ${openaire_db_name}.publication p lateral view explode(p.collectedfrom.key) c as datasource) p from ${openaire_db_name}.publication p lateral view explode(p.collectedfrom.key) c as datasource) p
@ -23,7 +23,7 @@ LEFT OUTER JOIN
DROP TABLE IF EXISTS ${stats_db_name}.dataset_sources purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.dataset_sources purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.dataset_sources STORED AS PARQUET 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 SELECT /*+ COALESCE(100) */ p.id, case when d.id is null then 'other' else p.datasource end as datasource
FROM ( FROM (
SELECT substr(p.id, 4) as id, substr(datasource, 4) as datasource SELECT substr(p.id, 4) as id, substr(datasource, 4) as datasource
from ${openaire_db_name}.dataset p lateral view explode(p.collectedfrom.key) c as datasource) p from ${openaire_db_name}.dataset p lateral view explode(p.collectedfrom.key) c as datasource) p
@ -36,7 +36,7 @@ LEFT OUTER JOIN
DROP TABLE IF EXISTS ${stats_db_name}.software_sources purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.software_sources purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.software_sources STORED AS PARQUET 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 SELECT /*+ COALESCE(100) */ p.id, case when d.id is null then 'other' else p.datasource end as datasource
FROM ( FROM (
SELECT substr(p.id, 4) as id, substr(datasource, 4) as datasource SELECT substr(p.id, 4) as id, substr(datasource, 4) as datasource
from ${openaire_db_name}.software p lateral view explode(p.collectedfrom.key) c as datasource) p from ${openaire_db_name}.software p lateral view explode(p.collectedfrom.key) c as datasource) p
@ -49,7 +49,7 @@ LEFT OUTER JOIN
DROP TABLE IF EXISTS ${stats_db_name}.otherresearchproduct_sources purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.otherresearchproduct_sources purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.otherresearchproduct_sources STORED AS PARQUET 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 SELECT /*+ COALESCE(100) */ p.id, case when d.id is null then 'other' else p.datasource end as datasource
FROM ( FROM (
SELECT substr(p.id, 4) as id, substr(datasource, 4) as datasource SELECT substr(p.id, 4) as id, substr(datasource, 4) as datasource
from ${openaire_db_name}.otherresearchproduct p lateral view explode(p.collectedfrom.key) c as datasource) p from ${openaire_db_name}.otherresearchproduct p lateral view explode(p.collectedfrom.key) c as datasource) p
@ -71,7 +71,7 @@ SELECT * FROM ${stats_db_name}.otherresearchproduct_sources; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.result_orcid purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.result_orcid purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.result_orcid STORED AS PARQUET as CREATE TABLE IF NOT EXISTS ${stats_db_name}.result_orcid STORED AS PARQUET as
select distinct res.id, upper(regexp_replace(res.orcid, 'http://orcid.org/' ,'')) as orcid select /*+ COALESCE(100) */ distinct res.id, upper(regexp_replace(res.orcid, 'http://orcid.org/' ,'')) as orcid
from ( from (
SELECT substr(res.id, 4) as id, auth_pid.value as orcid SELECT substr(res.id, 4) as id, auth_pid.value as orcid
FROM ${openaire_db_name}.result res FROM ${openaire_db_name}.result res
@ -83,7 +83,7 @@ from (
DROP TABLE IF EXISTS ${stats_db_name}.result_result purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.result_result purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.result_result stored as parquet as CREATE TABLE IF NOT EXISTS ${stats_db_name}.result_result stored as parquet as
select substr(rel.source, 4) as source, substr(rel.target, 4) as target, relclass, subreltype select /*+ COALESCE(100) */ substr(rel.source, 4) as source, substr(rel.target, 4) as target, relclass, subreltype
from ${openaire_db_name}.relation rel from ${openaire_db_name}.relation rel
join ${openaire_db_name}.result r1 on rel.source=r1.id join ${openaire_db_name}.result r1 on rel.source=r1.id
join ${openaire_db_name}.result r2 on r2.id=rel.target join ${openaire_db_name}.result r2 on r2.id=rel.target
@ -98,7 +98,7 @@ where reltype='resultResult'
DROP TABLE IF EXISTS ${stats_db_name}.result_citations_oc purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.result_citations_oc purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.result_citations_oc stored as parquet as CREATE TABLE IF NOT EXISTS ${stats_db_name}.result_citations_oc stored as parquet as
select substr(target, 4) as id, count(distinct substr(source, 4)) as citations select /*+ COALESCE(100) */ substr(target, 4) as id, count(distinct substr(source, 4)) as citations
from ${openaire_db_name}.relation rel from ${openaire_db_name}.relation rel
join ${openaire_db_name}.result r1 on rel.source=r1.id join ${openaire_db_name}.result r1 on rel.source=r1.id
join ${openaire_db_name}.result r2 on r2.id=rel.target join ${openaire_db_name}.result r2 on r2.id=rel.target
@ -115,7 +115,7 @@ group by substr(target, 4); /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.result_references_oc purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.result_references_oc purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.result_references_oc stored as parquet as CREATE TABLE IF NOT EXISTS ${stats_db_name}.result_references_oc stored as parquet as
select substr(source, 4) as id, count(distinct substr(target, 4)) as references select /*+ COALESCE(100) */ substr(source, 4) as id, count(distinct substr(target, 4)) as references
from ${openaire_db_name}.relation rel from ${openaire_db_name}.relation rel
join ${openaire_db_name}.result r1 on rel.source=r1.id join ${openaire_db_name}.result r1 on rel.source=r1.id
join ${openaire_db_name}.result r2 on r2.id=rel.target join ${openaire_db_name}.result r2 on r2.id=rel.target

15
dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step14.sql
View File

@ -1,4 +1,5 @@
set mapred.job.queue.name=analytics; /*EOS*/ set mapred.job.queue.name=analytics; /*EOS*/
------------------------------------------------------ ------------------------------------------------------
------------------------------------------------------ ------------------------------------------------------
-- Additional relations -- Additional relations
@ -9,28 +10,28 @@ set mapred.job.queue.name=analytics; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.publication_licenses purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.publication_licenses purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.publication_licenses STORED AS PARQUET 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 SELECT /*+ COALESCE(100) */ 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 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; /*EOS*/ where licenses.value is not null and licenses.value != '' and p.datainfo.deletedbyinference=false and p.datainfo.invisible = FALSE; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.dataset_licenses purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.dataset_licenses purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.dataset_licenses STORED AS PARQUET 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 SELECT /*+ COALESCE(100) */ 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 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; /*EOS*/ where licenses.value is not null and licenses.value != '' and p.datainfo.deletedbyinference=false and p.datainfo.invisible = FALSE; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.software_licenses purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.software_licenses purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.software_licenses STORED AS PARQUET 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 SELECT /*+ COALESCE(100) */ 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 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; /*EOS*/ where licenses.value is not null and licenses.value != '' and p.datainfo.deletedbyinference=false and p.datainfo.invisible = FALSE; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.otherresearchproduct_licenses purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.otherresearchproduct_licenses purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.otherresearchproduct_licenses STORED AS PARQUET 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 SELECT /*+ COALESCE(100) */ 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 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; /*EOS*/ where licenses.value is not null and licenses.value != '' and p.datainfo.deletedbyinference=false and p.datainfo.invisible = FALSE; /*EOS*/
@ -46,13 +47,13 @@ SELECT * FROM ${stats_db_name}.otherresearchproduct_licenses; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.organization_pids purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.organization_pids purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.organization_pids STORED AS PARQUET 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 select /*+ COALESCE(100) */ 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; /*EOS*/ from ${openaire_db_name}.organization o lateral view explode(o.pid) pids as ppid; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.organization_sources purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.organization_sources purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.organization_sources STORED AS PARQUET 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 SELECT /*+ COALESCE(100) */ o.id, case when d.id is null then 'other' else o.datasource end as datasource
FROM ( FROM (
SELECT substr(o.id, 4) as id, substr(instances.instance.key, 4) as datasource SELECT substr(o.id, 4) as id, substr(instances.instance.key, 4) as datasource
from ${openaire_db_name}.organization o lateral view explode(o.collectedfrom) instances as instance) o from ${openaire_db_name}.organization o lateral view explode(o.collectedfrom) instances as instance) o
@ -64,6 +65,6 @@ FROM (
DROP TABLE IF EXISTS ${stats_db_name}.result_accessroute purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.result_accessroute purge; /*EOS*/
CREATE TABLE IF NOT EXISTS ${stats_db_name}.result_accessroute STORED AS PARQUET as CREATE TABLE IF NOT EXISTS ${stats_db_name}.result_accessroute STORED AS PARQUET as
select distinct substr(id,4) as id, accessroute from ${openaire_db_name}.result select /*+ COALESCE(100) */ distinct substr(id,4) as id, accessroute from ${openaire_db_name}.result
lateral view explode (instance.accessright.openaccessroute) openaccessroute as accessroute lateral view explode (instance.accessright.openaccessroute) openaccessroute as accessroute
WHERE datainfo.deletedbyinference=false and datainfo.invisible = FALSE; /*EOS*/ WHERE datainfo.deletedbyinference=false and datainfo.invisible = FALSE; /*EOS*/

12
dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step15.sql
View File

@ -18,7 +18,7 @@ non_peer_reviewed as (
select distinct substr(r.id, 4) as id, inst.refereed.classname as refereed select distinct 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 from ${openaire_db_name}.publication r lateral view explode(r.instance) instances as inst
where r.datainfo.deletedbyinference=false and r.datainfo.invisible = FALSE and inst.refereed.classname='nonPeerReviewed') where r.datainfo.deletedbyinference=false and r.datainfo.invisible = FALSE and inst.refereed.classname='nonPeerReviewed')
select distinct * select /*+ COALESCE(100) */ distinct *
from ( from (
select peer_reviewed.* from peer_reviewed select peer_reviewed.* from peer_reviewed
union all union all
@ -36,7 +36,7 @@ non_peer_reviewed as (
select distinct substr(r.id, 4) as id, inst.refereed.classname as refereed select distinct 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 from ${openaire_db_name}.dataset r lateral view explode(r.instance) instances as inst
where r.datainfo.deletedbyinference=false and r.datainfo.invisible = FALSE and inst.refereed.classname='nonPeerReviewed') where r.datainfo.deletedbyinference=false and r.datainfo.invisible = FALSE and inst.refereed.classname='nonPeerReviewed')
select distinct * select /*+ COALESCE(100) */ distinct *
from ( from (
select peer_reviewed.* from peer_reviewed select peer_reviewed.* from peer_reviewed
union all union all
@ -54,7 +54,7 @@ non_peer_reviewed as (
select distinct substr(r.id, 4) as id, inst.refereed.classname as refereed select distinct 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 from ${openaire_db_name}.software r lateral view explode(r.instance) instances as inst
where r.datainfo.deletedbyinference=false and r.datainfo.invisible = FALSE and inst.refereed.classname='nonPeerReviewed') where r.datainfo.deletedbyinference=false and r.datainfo.invisible = FALSE and inst.refereed.classname='nonPeerReviewed')
select distinct * select /*+ COALESCE(100) */ distinct *
from ( from (
select peer_reviewed.* from peer_reviewed select peer_reviewed.* from peer_reviewed
union all union all
@ -72,7 +72,7 @@ non_peer_reviewed as (
select distinct substr(r.id, 4) as id, inst.refereed.classname as refereed select distinct 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 from ${openaire_db_name}.otherresearchproduct r lateral view explode(r.instance) instances as inst
where r.datainfo.deletedbyinference=false and r.datainfo.invisible = FALSE and inst.refereed.classname='nonPeerReviewed') where r.datainfo.deletedbyinference=false and r.datainfo.invisible = FALSE and inst.refereed.classname='nonPeerReviewed')
select distinct * select /*+ COALESCE(100) */ distinct *
from ( from (
select peer_reviewed.* from peer_reviewed select peer_reviewed.* from peer_reviewed
union all union all
@ -92,7 +92,7 @@ select * from ${stats_db_name}.otherresearchproduct_refereed; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.indi_impact_measures purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.indi_impact_measures purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_impact_measures STORED AS PARQUET as create table if not exists ${stats_db_name}.indi_impact_measures STORED AS PARQUET as
select substr(id, 4) as id, measures_ids.id impactmetric, cast(measures_ids.unit.value[0] as double) score, select /*+ COALESCE(100) */ substr(id, 4) as id, measures_ids.id impactmetric, cast(measures_ids.unit.value[0] as double) score,
cast(measures_ids.unit.value[0] as decimal(6,3)) score_dec, measures_ids.unit.value[1] impact_class cast(measures_ids.unit.value[0] as decimal(6,3)) score_dec, measures_ids.unit.value[1] impact_class
from ${openaire_db_name}.result lateral view explode(measures) measures as measures_ids from ${openaire_db_name}.result lateral view explode(measures) measures as measures_ids
where measures_ids.id!='views' and measures_ids.id!='downloads'; /*EOS*/ where measures_ids.id!='views' and measures_ids.id!='downloads'; /*EOS*/
@ -100,7 +100,7 @@ where measures_ids.id!='views' and measures_ids.id!='downloads'; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.result_apc_affiliations purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.result_apc_affiliations purge; /*EOS*/
create table if not exists ${stats_db_name}.result_apc_affiliations STORED AS PARQUET as create table if not exists ${stats_db_name}.result_apc_affiliations STORED AS PARQUET as
select distinct substr(rel.target,4) id, substr(rel.source,4) organization, o.legalname.value name, select /*+ COALESCE(100) */ distinct substr(rel.target,4) id, substr(rel.source,4) organization, o.legalname.value name,
cast(rel.properties[0].value as double) apc_amount, cast(rel.properties[0].value as double) apc_amount,
rel.properties[1].value apc_currency rel.properties[1].value apc_currency
from ${openaire_db_name}.relation rel from ${openaire_db_name}.relation rel

10
dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step15_5.sql
View File

@ -6,7 +6,7 @@ set mapred.job.queue.name=analytics; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.result_projectcount purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.result_projectcount purge; /*EOS*/
create table if not exists ${stats_db_name}.result_projectcount STORED AS PARQUET as create table if not exists ${stats_db_name}.result_projectcount STORED AS PARQUET as
select r.id, count(distinct p.id) as count select /*+ COALESCE(100) */ r.id, count(distinct p.id) as count
from ${stats_db_name}.result r 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}.result_projects rp on rp.id=r.id
left outer join ${stats_db_name}.project p on p.id=rp.project left outer join ${stats_db_name}.project p on p.id=rp.project
@ -15,7 +15,7 @@ group by r.id; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.result_fundercount purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.result_fundercount purge; /*EOS*/
create table if not exists ${stats_db_name}.result_fundercount STORED AS PARQUET as create table if not exists ${stats_db_name}.result_fundercount STORED AS PARQUET as
select r.id, count(distinct p.funder) as count select /*+ COALESCE(100) */ r.id, count(distinct p.funder) as count
from ${stats_db_name}.result r 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}.result_projects rp on rp.id=r.id
left outer join ${stats_db_name}.project p on p.id=rp.project left outer join ${stats_db_name}.project p on p.id=rp.project
@ -30,7 +30,7 @@ with rcount as (
left outer join ${stats_db_name}.result_projects rp on rp.project=p.id left outer join ${stats_db_name}.result_projects rp on rp.project=p.id
left outer join ${stats_db_name}.result r on r.id=rp.id left outer join ${stats_db_name}.result r on r.id=rp.id
group by r.type, p.id ) group by r.type, p.id )
select rcount.pid, sum(case when rcount.type='publication' then rcount.count else 0 end) as publications, select /*+ COALESCE(100) */ rcount.pid, sum(case when rcount.type='publication' then rcount.count else 0 end) as publications,
sum(case when rcount.type='dataset' then rcount.count else 0 end) as datasets, sum(case when rcount.type='dataset' then rcount.count else 0 end) as datasets,
sum(case when rcount.type='software' then rcount.count else 0 end) as software, sum(case when rcount.type='software' then rcount.count else 0 end) as software,
sum(case when rcount.type='other' then rcount.count else 0 end) as other sum(case when rcount.type='other' then rcount.count else 0 end) as other
@ -48,7 +48,7 @@ create or replace view ${stats_db_name}.graduatedoctorates as select * from stat
DROP TABLE IF EXISTS ${stats_db_name}.result_instance purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.result_instance purge; /*EOS*/
create table if not exists ${stats_db_name}.result_instance stored as parquet as create table if not exists ${stats_db_name}.result_instance stored as parquet as
select distinct r.* select /*+ COALESCE(100) */ distinct r.*
from ( from (
select substr(r.id, 4) as id, inst.accessright.classname as accessright, inst.accessright.openaccessroute as accessright_uw, substr(inst.collectedfrom.key, 4) as collectedfrom, select substr(r.id, 4) as id, inst.accessright.classname as accessright, inst.accessright.openaccessroute as accessright_uw, substr(inst.collectedfrom.key, 4) as collectedfrom,
substr(inst.hostedby.key, 4) as hostedby, inst.dateofacceptance.value as dateofacceptance, inst.license.value as license, p.qualifier.classname as pidtype, p.value as pid substr(inst.hostedby.key, 4) as hostedby, inst.dateofacceptance.value as dateofacceptance, inst.license.value as license, p.qualifier.classname as pidtype, p.value as pid
@ -58,7 +58,7 @@ join ${stats_db_name}.result res on res.id=r.id; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.result_apc purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.result_apc purge; /*EOS*/
create table if not exists ${stats_db_name}.result_apc STORED AS PARQUET as create table if not exists ${stats_db_name}.result_apc STORED AS PARQUET as
select distinct r.id, r.amount, r.currency select /*+ COALESCE(100) */ distinct r.id, r.amount, r.currency
from ( from (
select substr(r.id, 4) as id, cast(inst.processingchargeamount.value as float) as amount, inst.processingchargecurrency.value as currency select substr(r.id, 4) as id, cast(inst.processingchargeamount.value as float) as amount, inst.processingchargecurrency.value as currency
from ${openaire_db_name}.result r lateral view explode(r.instance) instances as inst) r from ${openaire_db_name}.result r lateral view explode(r.instance) instances as inst) r

116
dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step16-createIndicatorsTables.sql
View File

@ -1,7 +1,7 @@
-- Sprint 1 ---- -- Sprint 1 ----
drop table if exists ${stats_db_name}.indi_pub_green_oa purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_green_oa purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_green_oa stored as parquet as create table if not exists ${stats_db_name}.indi_pub_green_oa stored as parquet as
select distinct p.id, coalesce(green_oa, 0) as green_oa select /*+ COALESCE(100) */ distinct p.id, coalesce(green_oa, 0) as green_oa
from ${stats_db_name}.publication p from ${stats_db_name}.publication p
left outer join ( left outer join (
select p.id, 1 as green_oa select p.id, 1 as green_oa
@ -12,7 +12,7 @@ left outer join (
drop table if exists ${stats_db_name}.indi_pub_grey_lit purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_grey_lit purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_grey_lit stored as parquet as create table if not exists ${stats_db_name}.indi_pub_grey_lit stored as parquet as
select distinct p.id, coalesce(grey_lit, 0) as grey_lit select /*+ COALESCE(100) */ distinct p.id, coalesce(grey_lit, 0) as grey_lit
from ${stats_db_name}.publication p from ${stats_db_name}.publication p
left outer join ( left outer join (
select p.id, 1 as grey_lit select p.id, 1 as grey_lit
@ -23,7 +23,7 @@ left outer join (
drop table if exists ${stats_db_name}.indi_pub_doi_from_crossref purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_doi_from_crossref purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_doi_from_crossref stored as parquet as create table if not exists ${stats_db_name}.indi_pub_doi_from_crossref stored as parquet as
select distinct p.id, coalesce(doi_from_crossref, 0) as doi_from_crossref select /*+ COALESCE(100) */ distinct p.id, coalesce(doi_from_crossref, 0) as doi_from_crossref
from ${stats_db_name}.publication p from ${stats_db_name}.publication p
left outer join ( left outer join (
select ri.id, 1 as doi_from_crossref from ${stats_db_name}.result_instance ri select ri.id, 1 as doi_from_crossref from ${stats_db_name}.result_instance ri
@ -33,7 +33,7 @@ left outer join (
-- Sprint 2 ---- -- Sprint 2 ----
drop table if exists ${stats_db_name}.indi_result_has_cc_licence purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_result_has_cc_licence purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_result_has_cc_licence stored as parquet as create table if not exists ${stats_db_name}.indi_result_has_cc_licence stored as parquet as
select distinct r.id, (case when lic='' or lic is null then 0 else 1 end) as has_cc_license select /*+ COALESCE(100) */ distinct r.id, (case when lic='' or lic is null then 0 else 1 end) as has_cc_license
from ${stats_db_name}.result r from ${stats_db_name}.result r
left outer join ( left outer join (
select r.id, license.type as lic from ${stats_db_name}.result r select r.id, license.type as lic from ${stats_db_name}.result r
@ -42,7 +42,7 @@ left outer join (
drop table if exists ${stats_db_name}.indi_result_has_cc_licence_url purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_result_has_cc_licence_url purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_result_has_cc_licence_url stored as parquet as create table if not exists ${stats_db_name}.indi_result_has_cc_licence_url stored as parquet as
select distinct r.id, case when lic_host='' or lic_host is null then 0 else 1 end as has_cc_license_url select /*+ COALESCE(100) */ distinct r.id, case when lic_host='' or lic_host is null then 0 else 1 end as has_cc_license_url
from ${stats_db_name}.result r from ${stats_db_name}.result r
left outer join ( left outer join (
select r.id, lower(parse_url(license.type, "HOST")) as lic_host select r.id, lower(parse_url(license.type, "HOST")) as lic_host
@ -52,12 +52,12 @@ left outer join (
drop table if exists ${stats_db_name}.indi_pub_has_abstract purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_has_abstract purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_has_abstract stored as parquet as create table if not exists ${stats_db_name}.indi_pub_has_abstract stored as parquet as
select distinct publication.id, cast(coalesce(abstract, true) as int) has_abstract select /*+ COALESCE(100) */ distinct publication.id, cast(coalesce(abstract, true) as int) has_abstract
from ${stats_db_name}.publication; /*EOS*/ from ${stats_db_name}.publication; /*EOS*/
drop table if exists ${stats_db_name}.indi_result_with_orcid purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_result_with_orcid purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_result_with_orcid stored as parquet as create table if not exists ${stats_db_name}.indi_result_with_orcid stored as parquet as
select distinct r.id, coalesce(has_orcid, 0) as has_orcid select /*+ COALESCE(100) */ distinct r.id, coalesce(has_orcid, 0) as has_orcid
from ${stats_db_name}.result r from ${stats_db_name}.result r
left outer join ( left outer join (
select id, 1 as has_orcid from ${stats_db_name}.result_orcid) tmp on r.id= tmp.id; /*EOS*/ select id, 1 as has_orcid from ${stats_db_name}.result_orcid) tmp on r.id= tmp.id; /*EOS*/
@ -66,7 +66,7 @@ left outer join (
---- Sprint 3 ---- ---- Sprint 3 ----
drop table if exists ${stats_db_name}.indi_funded_result_with_fundref purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_funded_result_with_fundref purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_funded_result_with_fundref stored as parquet as create table if not exists ${stats_db_name}.indi_funded_result_with_fundref stored as parquet as
select distinct r.result as id, coalesce(fundref, 0) as fundref select /*+ COALESCE(100) */ distinct r.result as id, coalesce(fundref, 0) as fundref
from ${stats_db_name}.project_results r from ${stats_db_name}.project_results r
left outer join ( left outer join (
select distinct result, 1 as fundref from ${stats_db_name}.project_results where provenance='Harvested') tmp on r.result= tmp.result; /*EOS*/ select distinct result, 1 as fundref from ${stats_db_name}.project_results where provenance='Harvested') tmp on r.result= tmp.result; /*EOS*/
@ -77,7 +77,7 @@ create table if not exists ${stats_db_name}.indi_result_org_collab stored as par
SELECT ro.organization organization, ro.id, o.name SELECT ro.organization organization, ro.id, o.name
from ${stats_db_name}.result_organization ro from ${stats_db_name}.result_organization ro
join ${stats_db_name}.organization o on o.id=ro.organization where o.name is not null) join ${stats_db_name}.organization o on o.id=ro.organization where o.name is not null)
select o1.organization org1, o1.name org1name1, o2.organization org2, o2.name org2name2, count(o1.id) as collaborations select /*+ COALESCE(100) */ o1.organization org1, o1.name org1name1, o2.organization org2, o2.name org2name2, count(o1.id) as collaborations
from tmp as o1 from tmp as o1
join tmp as o2 where o1.id=o2.id and o1.organization!=o2.organization and o1.name!=o2.name join tmp as o2 where o1.id=o2.id and o1.organization!=o2.organization and o1.name!=o2.name
group by o1.organization, o2.organization, o1.name, o2.name; /*EOS*/ group by o1.organization, o2.organization, o1.name, o2.name; /*EOS*/
@ -89,7 +89,7 @@ create table if not exists ${stats_db_name}.indi_result_org_country_collab store
from ${stats_db_name}.result_organization ro from ${stats_db_name}.result_organization ro
join ${stats_db_name}.organization o on o.id=ro.organization join ${stats_db_name}.organization o on o.id=ro.organization
where country <> 'UNKNOWN' and o.name is not null) where country <> 'UNKNOWN' and o.name is not null)
select o1.organization org1,o1.name org1name1, o2.country country2, count(o1.id) as collaborations select /*+ COALESCE(100) */ o1.organization org1,o1.name org1name1, o2.country country2, count(o1.id) as collaborations
from tmp as o1 join tmp as o2 on o1.id=o2.id from tmp as o1 join tmp as o2 on o1.id=o2.id
where o1.id=o2.id and o1.country!=o2.country where o1.id=o2.id and o1.country!=o2.country
group by o1.organization, o1.id, o1.name, o2.country; /*EOS*/ group by o1.organization, o1.id, o1.name, o2.country; /*EOS*/
@ -100,7 +100,7 @@ create table if not exists ${stats_db_name}.indi_project_collab_org stored as pa
select o.id organization, o.name, ro.project as project select o.id organization, o.name, ro.project as project
from ${stats_db_name}.organization o from ${stats_db_name}.organization o
join ${stats_db_name}.organization_projects ro on o.id=ro.id where o.name is not null) join ${stats_db_name}.organization_projects ro on o.id=ro.id where o.name is not null)
select o1.organization org1,o1.name orgname1, o2.organization org2, o2.name orgname2, count(distinct o1.project) as collaborations select /*+ COALESCE(100) */ o1.organization org1,o1.name orgname1, o2.organization org2, o2.name orgname2, count(distinct o1.project) as collaborations
from tmp as o1 from tmp as o1
join tmp as o2 on o1.project=o2.project join tmp as o2 on o1.project=o2.project
where o1.organization<>o2.organization and o1.name<>o2.name where o1.organization<>o2.organization and o1.name<>o2.name
@ -112,7 +112,7 @@ create table if not exists ${stats_db_name}.indi_project_collab_org_country stor
select o.id organization, o.name, o.country , ro.project as project select o.id organization, o.name, o.country , ro.project as project
from ${stats_db_name}.organization o from ${stats_db_name}.organization o
join ${stats_db_name}.organization_projects ro on o.id=ro.id and o.country <> 'UNKNOWN' and o.name is not null) join ${stats_db_name}.organization_projects ro on o.id=ro.id and o.country <> 'UNKNOWN' and o.name is not null)
select o1.organization org1,o1.name org1name, o2.country country2, count(distinct o1.project) as collaborations select /*+ COALESCE(100) */ o1.organization org1,o1.name org1name, o2.country country2, count(distinct o1.project) as collaborations
from tmp as o1 from tmp as o1
join tmp as o2 on o1.project=o2.project join tmp as o2 on o1.project=o2.project
where o1.organization<>o2.organization and o1.country<>o2.country where o1.organization<>o2.organization and o1.country<>o2.country
@ -124,7 +124,7 @@ create table if not exists ${stats_db_name}.indi_funder_country_collab stored as
join ${stats_db_name}.organization o on o.id=op.id join ${stats_db_name}.organization o on o.id=op.id
join ${stats_db_name}.project p on p.id=op.project join ${stats_db_name}.project p on p.id=op.project
where country <> 'UNKNOWN') where country <> 'UNKNOWN')
select f1.funder, f1.country as country1, f2.country as country2, count(distinct f1.project) as collaborations select /*+ COALESCE(100) */ f1.funder, f1.country as country1, f2.country as country2, count(distinct f1.project) as collaborations
from tmp as f1 from tmp as f1
join tmp as f2 on f1.project=f2.project join tmp as f2 on f1.project=f2.project
where f1.country<>f2.country where f1.country<>f2.country
@ -136,7 +136,7 @@ create table if not exists ${stats_db_name}.indi_result_country_collab stored as
select distinct country, ro.id as result from ${stats_db_name}.organization o select distinct country, ro.id as result from ${stats_db_name}.organization o
join ${stats_db_name}.result_organization ro on o.id=ro.organization join ${stats_db_name}.result_organization ro on o.id=ro.organization
where country <> 'UNKNOWN' and o.name is not null) where country <> 'UNKNOWN' and o.name is not null)
select o1.country country1, o2.country country2, count(o1.result) as collaborations select /*+ COALESCE(100) */ o1.country country1, o2.country country2, count(o1.result) as collaborations
from tmp as o1 from tmp as o1
join tmp as o2 on o1.result=o2.result join tmp as o2 on o1.result=o2.result
where o1.country<>o2.country where o1.country<>o2.country
@ -146,7 +146,7 @@ create table if not exists ${stats_db_name}.indi_result_country_collab stored as
---- Sprint 4 ---- ---- Sprint 4 ----
drop table if exists ${stats_db_name}.indi_pub_diamond purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_diamond purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_diamond stored as parquet as create table if not exists ${stats_db_name}.indi_pub_diamond stored as parquet as
select distinct pd.id, coalesce(in_diamond_journal, 0) as in_diamond_journal select /*+ COALESCE(100) */ distinct pd.id, coalesce(in_diamond_journal, 0) as in_diamond_journal
from ${stats_db_name}.publication_datasources pd from ${stats_db_name}.publication_datasources pd
left outer join ( left outer join (
select pd.id, 1 as in_diamond_journal select pd.id, 1 as in_diamond_journal
@ -157,7 +157,7 @@ create table if not exists ${stats_db_name}.indi_pub_diamond stored as parquet a
drop table if exists ${stats_db_name}.indi_pub_in_transformative purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_in_transformative purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_in_transformative stored as parquet as create table if not exists ${stats_db_name}.indi_pub_in_transformative stored as parquet as
select distinct pd.id, coalesce(is_transformative, 0) as is_transformative select /*+ COALESCE(100) */ distinct pd.id, coalesce(is_transformative, 0) as is_transformative
from ${stats_db_name}.publication pd from ${stats_db_name}.publication pd
left outer join ( left outer join (
select pd.id, 1 as is_transformative select pd.id, 1 as is_transformative
@ -168,7 +168,7 @@ create table if not exists ${stats_db_name}.indi_pub_in_transformative stored as
drop table if exists ${stats_db_name}.indi_pub_closed_other_open purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_closed_other_open purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_closed_other_open stored as parquet as create table if not exists ${stats_db_name}.indi_pub_closed_other_open stored as parquet as
select distinct ri.id, coalesce(pub_closed_other_open, 0) as pub_closed_other_open select /*+ COALESCE(100) */ distinct ri.id, coalesce(pub_closed_other_open, 0) as pub_closed_other_open
from ${stats_db_name}.result_instance ri from ${stats_db_name}.result_instance ri
left outer join ( left outer join (
select ri.id, 1 as pub_closed_other_open select ri.id, 1 as pub_closed_other_open
@ -182,14 +182,14 @@ create table if not exists ${stats_db_name}.indi_pub_closed_other_open stored as
---- Sprint 5 ---- ---- Sprint 5 ----
drop table if exists ${stats_db_name}.indi_result_no_of_copies purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_result_no_of_copies purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_result_no_of_copies stored as parquet as create table if not exists ${stats_db_name}.indi_result_no_of_copies stored as parquet as
select id, count(id) as number_of_copies select /*+ COALESCE(100) */ id, count(id) as number_of_copies
from ${stats_db_name}.result_instance from ${stats_db_name}.result_instance
group by id; /*EOS*/ group by id; /*EOS*/
---- Sprint 6 ---- ---- Sprint 6 ----
drop table if exists ${stats_db_name}.indi_pub_downloads purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_downloads purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_downloads stored as parquet as create table if not exists ${stats_db_name}.indi_pub_downloads stored as parquet as
SELECT result_id, sum(downloads) no_downloads SELECT /*+ COALESCE(100) */ result_id, sum(downloads) no_downloads
from openaire_prod_usage_stats.usage_stats from openaire_prod_usage_stats.usage_stats
join ${stats_db_name}.publication on result_id=id join ${stats_db_name}.publication on result_id=id
where downloads>0 where downloads>0
@ -197,7 +197,7 @@ create table if not exists ${stats_db_name}.indi_pub_downloads stored as parquet
drop table if exists ${stats_db_name}.indi_pub_downloads_datasource purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_downloads_datasource purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_downloads_datasource stored as parquet as create table if not exists ${stats_db_name}.indi_pub_downloads_datasource stored as parquet as
SELECT result_id, repository_id, sum(downloads) no_downloads SELECT /*+ COALESCE(100) */ result_id, repository_id, sum(downloads) no_downloads
from openaire_prod_usage_stats.usage_stats from openaire_prod_usage_stats.usage_stats
join ${stats_db_name}.publication on result_id=id join ${stats_db_name}.publication on result_id=id
where downloads>0 where downloads>0
@ -205,14 +205,14 @@ create table if not exists ${stats_db_name}.indi_pub_downloads_datasource stored
drop table if exists ${stats_db_name}.indi_pub_downloads_year purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_downloads_year purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_downloads_year stored as parquet as create table if not exists ${stats_db_name}.indi_pub_downloads_year stored as parquet as
SELECT result_id, cast(substring(us.`date`, 1,4) as int) as `year`, sum(downloads) no_downloads SELECT /*+ COALESCE(100) */ result_id, cast(substring(us.`date`, 1,4) as int) as `year`, sum(downloads) no_downloads
from openaire_prod_usage_stats.usage_stats us from openaire_prod_usage_stats.usage_stats us
join ${stats_db_name}.publication on result_id=id where downloads>0 join ${stats_db_name}.publication on result_id=id where downloads>0
GROUP BY result_id, substring(us.`date`, 1,4); /*EOS*/ GROUP BY result_id, substring(us.`date`, 1,4); /*EOS*/
drop table if exists ${stats_db_name}.indi_pub_downloads_datasource_year purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_downloads_datasource_year purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_downloads_datasource_year stored as parquet as create table if not exists ${stats_db_name}.indi_pub_downloads_datasource_year stored as parquet as
SELECT result_id, cast(substring(us.`date`, 1,4) as int) as `year`, repository_id, sum(downloads) no_downloads SELECT /*+ COALESCE(100) */ result_id, cast(substring(us.`date`, 1,4) as int) as `year`, repository_id, sum(downloads) no_downloads
from openaire_prod_usage_stats.usage_stats us from openaire_prod_usage_stats.usage_stats us
join ${stats_db_name}.publication on result_id=id join ${stats_db_name}.publication on result_id=id
where downloads>0 where downloads>0
@ -241,7 +241,7 @@ create table if not exists ${stats_db_name}.indi_pub_gold_oa stored as parquet a
UNION ALL UNION ALL
select id, issn_online as issn from ${stats_db_name}.datasource d left semi join gold_oa on gold_oa.issn=d.issn_online) foo select id, issn_online as issn from ${stats_db_name}.datasource d left semi join gold_oa on gold_oa.issn=d.issn_online) foo
) )
SELECT DISTINCT pd.id, coalesce(is_gold, 0) as is_gold SELECT /*+ COALESCE(100) */ DISTINCT pd.id, coalesce(is_gold, 0) as is_gold
FROM ${stats_db_name}.publication pd FROM ${stats_db_name}.publication pd
left outer join ( left outer join (
select pd.id, 1 as is_gold select pd.id, 1 as is_gold
@ -272,7 +272,7 @@ create table if not exists ${stats_db_name}.indi_pub_hybrid_oa_with_cc stored as
FROM ${stats_db_name}.datasource FROM ${stats_db_name}.datasource
WHERE issn_online IS NOT NULL ) as issn WHERE issn_online IS NOT NULL ) as issn
WHERE LENGTH(issn) > 7) WHERE LENGTH(issn) > 7)
SELECT DISTINCT pd.id, coalesce(is_hybrid_oa, 0) as is_hybrid_oa SELECT /*+ COALESCE(100) */ DISTINCT pd.id, coalesce(is_hybrid_oa, 0) as is_hybrid_oa
FROM ${stats_db_name}.publication_datasources pd FROM ${stats_db_name}.publication_datasources pd
LEFT OUTER JOIN ( LEFT OUTER JOIN (
SELECT pd.id, 1 as is_hybrid_oa from ${stats_db_name}.publication_datasources pd SELECT pd.id, 1 as is_hybrid_oa from ${stats_db_name}.publication_datasources pd
@ -284,7 +284,7 @@ create table if not exists ${stats_db_name}.indi_pub_hybrid_oa_with_cc stored as
drop table if exists ${stats_db_name}.indi_pub_hybrid purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_hybrid purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_hybrid stored as parquet as create table if not exists ${stats_db_name}.indi_pub_hybrid stored as parquet as
select distinct p.id, coalesce(is_hybrid, 0) is_hybrid select /*+ COALESCE(100) */ distinct p.id, coalesce(is_hybrid, 0) is_hybrid
from ${stats_db_name}.publication p from ${stats_db_name}.publication p
left outer join ( left outer join (
select p.id, 1 as is_hybrid select p.id, 1 as is_hybrid
@ -313,7 +313,7 @@ create table if not exists ${stats_db_name}.indi_org_fairness stored as parquet
where cast(year as int)>2003 where cast(year as int)>2003
group by ro.organization) group by ro.organization)
--return results_fair/all_results --return results_fair/all_results
select allresults.organization, result_fair.no_result_fair/allresults.no_allresults org_fairness select /*+ COALESCE(100) */ allresults.organization, result_fair.no_result_fair/allresults.no_allresults org_fairness
from allresults from allresults
join result_fair on result_fair.organization=allresults.organization; /*EOS*/ join result_fair on result_fair.organization=allresults.organization; /*EOS*/
@ -336,7 +336,7 @@ select ro.organization, count(distinct ro.id) no_allresults from ${stats_db_name
drop table if exists ${stats_db_name}.indi_org_fairness_pub_pr purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_org_fairness_pub_pr purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_org_fairness_pub_pr stored as parquet as create table if not exists ${stats_db_name}.indi_org_fairness_pub_pr stored as parquet as
select ar.organization, rf.no_result_fair/ar.no_allresults org_fairness select /*+ COALESCE(100) */ ar.organization, rf.no_result_fair/ar.no_allresults org_fairness
from allresults ar from allresults ar
join result_fair rf on rf.organization=ar.organization; /*EOS*/ join result_fair rf on rf.organization=ar.organization; /*EOS*/
@ -357,7 +357,7 @@ CREATE TEMPORARY VIEW allresults as select year, ro.organization, count(distinct
drop table if exists ${stats_db_name}.indi_org_fairness_pub_year purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_org_fairness_pub_year purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_org_fairness_pub_year stored as parquet as create table if not exists ${stats_db_name}.indi_org_fairness_pub_year stored as parquet as
select cast(allresults.year as int) year, allresults.organization, result_fair.no_result_fair/allresults.no_allresults org_fairness select /*+ COALESCE(100) */ cast(allresults.year as int) year, allresults.organization, result_fair.no_result_fair/allresults.no_allresults org_fairness
from allresults from allresults
join result_fair on result_fair.organization=allresults.organization and result_fair.year=allresults.year; /*EOS*/ join result_fair on result_fair.organization=allresults.organization and result_fair.year=allresults.year; /*EOS*/
@ -381,7 +381,7 @@ CREATE TEMPORARY VIEW allresults as
drop table if exists ${stats_db_name}.indi_org_fairness_pub purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_org_fairness_pub purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_org_fairness_pub stored as parquet as create table if not exists ${stats_db_name}.indi_org_fairness_pub stored as parquet as
select ar.organization, rf.no_result_fair/ar.no_allresults org_fairness select /*+ COALESCE(100) */ ar.organization, rf.no_result_fair/ar.no_allresults org_fairness
from allresults ar join result_fair rf from allresults ar join result_fair rf
on rf.organization=ar.organization; /*EOS*/ on rf.organization=ar.organization; /*EOS*/
@ -404,7 +404,7 @@ CREATE TEMPORARY VIEW allresults as
drop table if exists ${stats_db_name}.indi_org_fairness_year purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_org_fairness_year purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_org_fairness_year stored as parquet as create table if not exists ${stats_db_name}.indi_org_fairness_year stored as parquet as
select cast(allresults.year as int) year, allresults.organization, result_fair.no_result_fair/allresults.no_allresults org_fairness select /*+ COALESCE(100) */ cast(allresults.year as int) year, allresults.organization, result_fair.no_result_fair/allresults.no_allresults org_fairness
from allresults from allresults
join result_fair on result_fair.organization=allresults.organization and cast(result_fair.year as int)=cast(allresults.year as int); /*EOS*/ join result_fair on result_fair.organization=allresults.organization and cast(result_fair.year as int)=cast(allresults.year as int); /*EOS*/
@ -427,7 +427,7 @@ CREATE TEMPORARY VIEW allresults as
drop table if exists ${stats_db_name}.indi_org_findable_year purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_org_findable_year purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_org_findable_year stored as parquet as create table if not exists ${stats_db_name}.indi_org_findable_year stored as parquet as
select cast(allresults.year as int) year, allresults.organization, result_with_pid.no_result_with_pid/allresults.no_allresults org_findable select /*+ COALESCE(100) */ cast(allresults.year as int) year, allresults.organization, result_with_pid.no_result_with_pid/allresults.no_allresults org_findable
from allresults from allresults
join result_with_pid on result_with_pid.organization=allresults.organization and cast(result_with_pid.year as int)=cast(allresults.year as int); /*EOS*/ join result_with_pid on result_with_pid.organization=allresults.organization and cast(result_with_pid.year as int)=cast(allresults.year as int); /*EOS*/
@ -450,7 +450,7 @@ select ro.organization, count(distinct ro.id) no_allresults from ${stats_db_name
drop table if exists ${stats_db_name}.indi_org_findable purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_org_findable purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_org_findable stored as parquet as create table if not exists ${stats_db_name}.indi_org_findable stored as parquet as
select allresults.organization, result_with_pid.no_result_with_pid/allresults.no_allresults org_findable select /*+ COALESCE(100) */ allresults.organization, result_with_pid.no_result_with_pid/allresults.no_allresults org_findable
from allresults from allresults
join result_with_pid on result_with_pid.organization=allresults.organization; /*EOS*/ join result_with_pid on result_with_pid.organization=allresults.organization; /*EOS*/
@ -516,7 +516,7 @@ select software_oa.organization, software_oa.no_oasoftware/allsoftware.no_allsof
drop table if exists ${stats_db_name}.indi_org_openess purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_org_openess purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_org_openess stored as parquet as create table if not exists ${stats_db_name}.indi_org_openess stored as parquet as
select allpubsshare.organization, select /*+ COALESCE(100) */ allpubsshare.organization,
(p+if(isnull(s),0,s)+if(isnull(d),0,d))/(1+(case when s is null then 0 else 1 end) (p+if(isnull(s),0,s)+if(isnull(d),0,d))/(1+(case when s is null then 0 else 1 end)
+(case when d is null then 0 else 1 end)) +(case when d is null then 0 else 1 end))
org_openess FROM allpubsshare org_openess FROM allpubsshare
@ -593,7 +593,7 @@ select allsoftware.year, software_oa.organization, software_oa.no_oasoftware/all
drop table if exists ${stats_db_name}.indi_org_openess_year purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_org_openess_year purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_org_openess_year stored as parquet as create table if not exists ${stats_db_name}.indi_org_openess_year stored as parquet as
select cast(allpubsshare.year as int) year, allpubsshare.organization, select /*+ COALESCE(100) */ cast(allpubsshare.year as int) year, allpubsshare.organization,
(p+if(isnull(s),0,s)+if(isnull(d),0,d))/(1+(case when s is null then 0 else 1 end) (p+if(isnull(s),0,s)+if(isnull(d),0,d))/(1+(case when s is null then 0 else 1 end)
+(case when d is null then 0 else 1 end)) +(case when d is null then 0 else 1 end))
org_openess FROM allpubsshare org_openess FROM allpubsshare
@ -617,7 +617,7 @@ DROP VIEW allsoftwaresshare; /*EOS*/
drop table if exists ${stats_db_name}.indi_pub_has_preprint purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_has_preprint purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_has_preprint stored as parquet as create table if not exists ${stats_db_name}.indi_pub_has_preprint stored as parquet as
select distinct p.id, coalesce(has_preprint, 0) as has_preprint select /*+ COALESCE(100) */ distinct p.id, coalesce(has_preprint, 0) as has_preprint
from ${stats_db_name}.publication_classifications p from ${stats_db_name}.publication_classifications p
left outer join ( left outer join (
select p.id, 1 as has_preprint select p.id, 1 as has_preprint
@ -627,7 +627,7 @@ from ${stats_db_name}.publication_classifications p
drop table if exists ${stats_db_name}.indi_pub_in_subscribed purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_in_subscribed purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_in_subscribed stored as parquet as create table if not exists ${stats_db_name}.indi_pub_in_subscribed stored as parquet as
select distinct p.id, coalesce(is_subscription, 0) as is_subscription select /*+ COALESCE(100) */ distinct p.id, coalesce(is_subscription, 0) as is_subscription
from ${stats_db_name}.publication p from ${stats_db_name}.publication p
left outer join( left outer join(
select p.id, 1 as is_subscription from ${stats_db_name}.publication p select p.id, 1 as is_subscription from ${stats_db_name}.publication p
@ -640,7 +640,7 @@ from ${stats_db_name}.publication p
drop table if exists ${stats_db_name}.indi_result_with_pid purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_result_with_pid purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_result_with_pid stored as parquet as create table if not exists ${stats_db_name}.indi_result_with_pid stored as parquet as
select distinct p.id, coalesce(result_with_pid, 0) as result_with_pid select /*+ COALESCE(100) */ distinct p.id, coalesce(result_with_pid, 0) as result_with_pid
from ${stats_db_name}.result p from ${stats_db_name}.result p
left outer join ( left outer join (
select p.id, 1 as result_with_pid select p.id, 1 as result_with_pid
@ -654,7 +654,7 @@ group by rf.id; /*EOS*/
drop table if exists ${stats_db_name}.indi_pub_interdisciplinarity purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_interdisciplinarity purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_pub_interdisciplinarity stored as parquet as create table if not exists ${stats_db_name}.indi_pub_interdisciplinarity stored as parquet as
select distinct p.id as id, coalesce(is_interdisciplinary, 0) select /*+ COALESCE(100) */ distinct p.id as id, coalesce(is_interdisciplinary, 0)
as is_interdisciplinary as is_interdisciplinary
from pub_fos_totals p from pub_fos_totals p
left outer join ( left outer join (
@ -666,7 +666,7 @@ drop view pub_fos_totals; /*EOS*/
drop table if exists ${stats_db_name}.indi_pub_bronze_oa purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_bronze_oa purge; /*EOS*/
create table ${stats_db_name}.indi_pub_bronze_oa stored as parquet as create table ${stats_db_name}.indi_pub_bronze_oa stored as parquet as
select distinct p.id,coalesce(is_bronze_oa,0) is_bronze_oa select /*+ COALESCE(100) */ distinct p.id,coalesce(is_bronze_oa,0) is_bronze_oa
from ${stats_db_name}.publication p from ${stats_db_name}.publication p
left outer join ( left outer join (
select p.id, 1 as is_bronze_oa select p.id, 1 as is_bronze_oa
@ -689,7 +689,7 @@ where p.end_year is NOT NULL and r.year is not null; /*EOS*/
drop table if exists ${stats_db_name}.indi_is_project_result_after purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_is_project_result_after purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_is_project_result_after stored as parquet as create table if not exists ${stats_db_name}.indi_is_project_result_after stored as parquet as
select pry.project_id, pry.acronym, pry.result_id, select /*+ COALESCE(100) */ pry.project_id, pry.acronym, pry.result_id,
coalesce(is_project_result_after, 0) as is_project_result_after coalesce(is_project_result_after, 0) as is_project_result_after
from project_year_result_year pry from project_year_result_year pry
left outer join (select pry.project_id, pry.acronym, pry.result_id, 1 as is_project_result_after left outer join (select pry.project_id, pry.acronym, pry.result_id, 1 as is_project_result_after
@ -701,7 +701,7 @@ drop view project_year_result_year; /*EOS*/
drop table if exists ${stats_db_name}.indi_is_funder_plan_s purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_is_funder_plan_s purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_is_funder_plan_s stored as parquet as create table if not exists ${stats_db_name}.indi_is_funder_plan_s stored as parquet as
select distinct f.id, f.name, coalesce(is_funder_plan_s, 0) as is_funder_plan_s select /*+ COALESCE(100) */ distinct f.id, f.name, coalesce(is_funder_plan_s, 0) as is_funder_plan_s
from ${stats_db_name}.funder f from ${stats_db_name}.funder f
left outer join (select id, name, 1 as is_funder_plan_s from ${stats_db_name}.funder left outer join (select id, name, 1 as is_funder_plan_s from ${stats_db_name}.funder
join stats_ext.plan_s_short on c_o_alition_s_organisation_funder=name) tmp join stats_ext.plan_s_short on c_o_alition_s_organisation_funder=name) tmp
@ -722,7 +722,7 @@ create table if not exists ${stats_db_name}.indi_funder_fairness stored as parqu
join ${stats_db_name}.project p on p.id=rp.project join ${stats_db_name}.project p on p.id=rp.project
where cast(year as int)>2003 where cast(year as int)>2003
group by p.funder) group by p.funder)
select allresults.funder, result_fair.no_result_fair/allresults.no_allresults funder_fairness select /*+ COALESCE(100) */ allresults.funder, result_fair.no_result_fair/allresults.no_allresults funder_fairness
from allresults from allresults
join result_fair on result_fair.funder=allresults.funder; /*EOS*/ join result_fair on result_fair.funder=allresults.funder; /*EOS*/
@ -745,7 +745,7 @@ allresults as
join ${stats_db_name}.result r on r.id=rc.id join ${stats_db_name}.result r on r.id=rc.id
where cast(year as int)>2003 where cast(year as int)>2003
group by rc.ri_initiative) group by rc.ri_initiative)
select allresults.ri_initiative, result_fair.no_result_fair/allresults.no_allresults ris_fairness select /*+ COALESCE(100) */ allresults.ri_initiative, result_fair.no_result_fair/allresults.no_allresults ris_fairness
from allresults from allresults
join result_fair on result_fair.ri_initiative=allresults.ri_initiative; /*EOS*/ join result_fair on result_fair.ri_initiative=allresults.ri_initiative; /*EOS*/
@ -817,15 +817,13 @@ select software_oa.funder, software_oa.no_oasoftware/allsoftware.no_allsoftware
drop table if exists ${stats_db_name}.indi_funder_openess purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_funder_openess purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_funder_openess stored as parquet as create table if not exists ${stats_db_name}.indi_funder_openess stored as parquet as
select allpubsshare.funder, select /*+ COALESCE(100) */ allpubsshare.funder,
(p+if(isnull(s),0,s)+if(isnull(d),0,d))/(1+(case when s is null then 0 else 1 end) (p+if(isnull(s),0,s)+if(isnull(d),0,d))/(1+(case when s is null then 0 else 1 end)
+(case when d is null then 0 else 1 end)) +(case when d is null then 0 else 1 end)) funder_openess
funder_openess FROM allpubsshare FROM allpubsshare
left outer join (select funder,d from left outer join (select funder,d from alldatasetssshare) tmp1
alldatasetssshare) tmp1
on tmp1.funder=allpubsshare.funder on tmp1.funder=allpubsshare.funder
left outer join (select funder,s from left outer join (select funder,s from allsoftwaresshare) tmp2
allsoftwaresshare) tmp2
on tmp2.funder=allpubsshare.funder; /*EOS*/ on tmp2.funder=allpubsshare.funder; /*EOS*/
DROP VIEW pubs_oa; /*EOS*/ DROP VIEW pubs_oa; /*EOS*/
@ -905,7 +903,7 @@ select software_oa.ri_initiative, software_oa.no_oasoftware/allsoftware.no_allso
drop table if exists ${stats_db_name}.indi_ris_openess purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_ris_openess purge; /*EOS*/
create table if not exists ${stats_db_name}.indi_ris_openess stored as parquet as create table if not exists ${stats_db_name}.indi_ris_openess stored as parquet as
select allpubsshare.ri_initiative, select /*+ COALESCE(100) */ allpubsshare.ri_initiative,
(p+if(isnull(s),0,s)+if(isnull(d),0,d))/(1+(case when s is null then 0 else 1 end) (p+if(isnull(s),0,s)+if(isnull(d),0,d))/(1+(case when s is null then 0 else 1 end)
+(case when d is null then 0 else 1 end)) +(case when d is null then 0 else 1 end))
ris_openess FROM allpubsshare ris_openess FROM allpubsshare
@ -943,7 +941,7 @@ with result_findable as
join ${stats_db_name}.project p on p.id=rp.project join ${stats_db_name}.project p on p.id=rp.project
where cast(year as int)>2003 where cast(year as int)>2003
group by p.funder) group by p.funder)
select allresults.funder, result_findable.no_result_findable/allresults.no_allresults funder_findable select /*+ COALESCE(100) */ allresults.funder, result_findable.no_result_findable/allresults.no_allresults funder_findable
from allresults from allresults
join result_findable on result_findable.funder=allresults.funder; /*EOS*/ join result_findable on result_findable.funder=allresults.funder; /*EOS*/
@ -967,7 +965,7 @@ allresults as
join ${stats_db_name}.result r on r.id=rc.id join ${stats_db_name}.result r on r.id=rc.id
where cast(r.year as int)>2003 where cast(r.year as int)>2003
group by rc.ri_initiative) group by rc.ri_initiative)
select allresults.ri_initiative, result_findable.no_result_findable/allresults.no_allresults ris_findable select /*+ COALESCE(100) */ allresults.ri_initiative, result_findable.no_result_findable/allresults.no_allresults ris_findable
from allresults from allresults
join result_findable on result_findable.ri_initiative=allresults.ri_initiative; /*EOS*/ join result_findable on result_findable.ri_initiative=allresults.ri_initiative; /*EOS*/
@ -988,7 +986,7 @@ union all
select op.name from stats_ext.insitutions_for_publicly_funded pf select op.name from stats_ext.insitutions_for_publicly_funded pf
join org_names_pids op on (op.name=pf.name or op.pid=pf.ror) join org_names_pids op on (op.name=pf.name or op.pid=pf.ror)
and pf.publicly_funded='yes') foo) and pf.publicly_funded='yes') foo)
select distinct p.id, coalesce(publicly_funded, 0) as publicly_funded select /*+ COALESCE(100) */ distinct p.id, coalesce(publicly_funded, 0) as publicly_funded
from ${stats_db_name}.publication p from ${stats_db_name}.publication p
left outer join ( left outer join (
select distinct ro.id, 1 as publicly_funded from ${stats_db_name}.result_organization ro select distinct ro.id, 1 as publicly_funded from ${stats_db_name}.result_organization ro
@ -997,7 +995,7 @@ join publicly_funded_orgs pfo on o.name=pfo.name) tmp on p.id=tmp.id; /*EOS*/
drop table if exists ${stats_db_name}.indi_pub_green_with_license purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_pub_green_with_license purge; /*EOS*/
create table ${stats_db_name}.indi_pub_green_with_license stored as parquet as create table ${stats_db_name}.indi_pub_green_with_license stored as parquet as
select distinct p.id, coalesce(green_with_license, 0) as green_with_license select /*+ COALESCE(100) */ distinct p.id, coalesce(green_with_license, 0) as green_with_license
from ${stats_db_name}.publication p from ${stats_db_name}.publication p
left outer join ( left outer join (
select distinct p.id, 1 as green_with_license from ${stats_db_name}.publication p select distinct p.id, 1 as green_with_license from ${stats_db_name}.publication p
@ -1008,7 +1006,7 @@ left outer join (
drop table if exists ${stats_db_name}.result_country purge; /*EOS*/ drop table if exists ${stats_db_name}.result_country purge; /*EOS*/
create table ${stats_db_name}.result_country stored as parquet as create table ${stats_db_name}.result_country stored as parquet as
select distinct id, country select /*+ COALESCE(100) */ distinct id, country
from ( from (
select ro.id, o.country select ro.id, o.country
from ${stats_db_name}.result_organization ro from ${stats_db_name}.result_organization ro
@ -1023,7 +1021,7 @@ where rc.country is not null; /*EOS*/
drop table if exists ${stats_db_name}.indi_result_oa_with_license purge; /*EOS*/ drop table if exists ${stats_db_name}.indi_result_oa_with_license purge; /*EOS*/
create table ${stats_db_name}.indi_result_oa_with_license stored as parquet as create table ${stats_db_name}.indi_result_oa_with_license stored as parquet as
select distinct r.id, coalesce(oa_with_license,0) as oa_with_license select /*+ COALESCE(100) */ distinct r.id, coalesce(oa_with_license,0) as oa_with_license
from ${stats_db_name}.result r from ${stats_db_name}.result r
left outer join (select distinct r.id, 1 as oa_with_license from ${stats_db_name}.result r left outer join (select distinct r.id, 1 as oa_with_license from ${stats_db_name}.result r
join ${stats_db_name}.result_licenses rl on rl.id=r.id where r.bestlicence='Open Access') tmp on r.id=tmp.id; /*EOS*/ join ${stats_db_name}.result_licenses rl on rl.id=r.id where r.bestlicence='Open Access') tmp on r.id=tmp.id; /*EOS*/
@ -1033,7 +1031,7 @@ create table ${stats_db_name}.indi_result_oa_without_license stored as parquet a
with without_license as with without_license as
(select distinct id from ${stats_db_name}.indi_result_oa_with_license (select distinct id from ${stats_db_name}.indi_result_oa_with_license
where oa_with_license=0) where oa_with_license=0)
select distinct r.id, coalesce(oa_without_license,0) as oa_without_license select /*+ COALESCE(100) */ distinct r.id, coalesce(oa_without_license,0) as oa_without_license
from ${stats_db_name}.result r from ${stats_db_name}.result r
left outer join (select distinct r.id, 1 as oa_without_license left outer join (select distinct r.id, 1 as oa_without_license
from ${stats_db_name}.result r from ${stats_db_name}.result r
@ -1044,7 +1042,7 @@ drop table if exists ${stats_db_name}.indi_result_under_transformative purge; /*
create table ${stats_db_name}.indi_result_under_transformative stored as parquet as create table ${stats_db_name}.indi_result_under_transformative stored as parquet as
with transformative_dois as ( with transformative_dois as (
select distinct doi from stats_ext.transformative_facts) select distinct doi from stats_ext.transformative_facts)
select distinct r.id, coalesce(under_transformative,0) as under_transformative select /*+ COALESCE(100) */ distinct r.id, coalesce(under_transformative,0) as under_transformative
from ${stats_db_name}.result r from ${stats_db_name}.result r
left outer join ( left outer join (
select distinct rp.id, 1 as under_transformative select distinct rp.id, 1 as under_transformative

6
dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step16_1-definitions.sql
View File

@ -8,7 +8,7 @@ set mapred.job.queue.name=analytics; /*EOS*/
drop table if exists ${stats_db_name}.result_peerreviewed purge; /*EOS*/ drop table if exists ${stats_db_name}.result_peerreviewed purge; /*EOS*/
create table IF NOT EXISTS ${stats_db_name}.result_peerreviewed STORED AS PARQUET as create table IF NOT EXISTS ${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 select /*+ COALESCE(100) */ 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 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_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; /*EOS*/ left outer join ${stats_db_name}.indi_pub_grey_lit grey on grey.id=r.id; /*EOS*/
@ -17,7 +17,7 @@ left outer join ${stats_db_name}.indi_pub_grey_lit grey on grey.id=r.id; /*EOS*/
drop table if exists ${stats_db_name}.result_greenoa purge; /*EOS*/ drop table if exists ${stats_db_name}.result_greenoa purge; /*EOS*/
create table IF NOT EXISTS ${stats_db_name}.result_greenoa STORED AS PARQUET as create table IF NOT EXISTS ${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 select /*+ COALESCE(100) */ r.id, case when green.green_oa=1 then true else false end as green
from ${stats_db_name}.result r from ${stats_db_name}.result r
left outer join ${stats_db_name}.indi_pub_green_oa green on green.id=r.id; /*EOS*/ left outer join ${stats_db_name}.indi_pub_green_oa green on green.id=r.id; /*EOS*/
@ -25,6 +25,6 @@ left outer join ${stats_db_name}.indi_pub_green_oa green on green.id=r.id; /*EOS
drop table if exists ${stats_db_name}.result_gold purge; /*EOS*/ drop table if exists ${stats_db_name}.result_gold purge; /*EOS*/
create table IF NOT EXISTS ${stats_db_name}.result_gold STORED AS PARQUET as create table IF NOT EXISTS ${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 select /*+ COALESCE(100) */ r.id, case when gold.is_gold=1 then true else false end as gold
from ${stats_db_name}.result r from ${stats_db_name}.result r
left outer join ${stats_db_name}.indi_pub_gold_oa gold on gold.id=r.id; /*EOS*/ left outer join ${stats_db_name}.indi_pub_gold_oa gold on gold.id=r.id; /*EOS*/

38
dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step21-createObservatoryDB.sql
View File

@ -1,7 +1,7 @@
set mapred.job.queue.name=analytics; /*EOS*/ set mapred.job.queue.name=analytics; /*EOS*/
create table ${observatory_db_name}.result_cc_licence stored as parquet as create table ${observatory_db_name}.result_cc_licence stored as parquet as
select r.id, coalesce(rln.count, 0) > 0 as cc_licence select /*+ COALESCE(100) */ r.id, coalesce(rln.count, 0) > 0 as cc_licence
from ${stats_db_name}.result r from ${stats_db_name}.result r
left outer join ( left outer join (
select rl.id, sum(case when rl.type like 'CC%' then 1 else 0 end) as count select rl.id, sum(case when rl.type like 'CC%' then 1 else 0 end) as count
@ -11,7 +11,7 @@ from ${stats_db_name}.result r
create table ${observatory_db_name}.result_affiliated_country stored as parquet as create table ${observatory_db_name}.result_affiliated_country stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -41,7 +41,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
create table ${observatory_db_name}.result_affiliated_year stored as parquet as create table ${observatory_db_name}.result_affiliated_year stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -71,7 +71,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
create table ${observatory_db_name}.result_affiliated_year_country stored as parquet as create table ${observatory_db_name}.result_affiliated_year_country stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -101,7 +101,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
create table ${observatory_db_name}.result_affiliated_datasource stored as parquet as create table ${observatory_db_name}.result_affiliated_datasource stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -132,7 +132,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, d.name; /*EOS*/ cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, d.name; /*EOS*/
create table ${observatory_db_name}.result_affiliated_datasource_country stored as parquet as create table ${observatory_db_name}.result_affiliated_datasource_country stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -163,7 +163,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, d.name, c.code, c.name; /*EOS*/ cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, d.name, c.code, c.name; /*EOS*/
create table ${observatory_db_name}.result_affiliated_organization stored as parquet as create table ${observatory_db_name}.result_affiliated_organization stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -192,7 +192,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, o.name; /*EOS*/ cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, o.name; /*EOS*/
create table ${observatory_db_name}.result_affiliated_organization_country stored as parquet as create table ${observatory_db_name}.result_affiliated_organization_country stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -221,7 +221,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, o.name, c.code, c.name; /*EOS*/ cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, o.name, c.code, c.name; /*EOS*/
create table ${observatory_db_name}.result_affiliated_funder stored as parquet as create table ${observatory_db_name}.result_affiliated_funder stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -252,7 +252,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, p.funder; /*EOS*/ cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, p.funder; /*EOS*/
create table ${observatory_db_name}.result_affiliated_funder_country stored as parquet as create table ${observatory_db_name}.result_affiliated_funder_country stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -283,7 +283,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, p.funder, c.code, c.name; /*EOS*/ cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, p.funder, c.code, c.name; /*EOS*/
create table ${observatory_db_name}.result_deposited_country stored as parquet as create table ${observatory_db_name}.result_deposited_country stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -314,7 +314,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, c.code, c.name; /*EOS*/ cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, c.code, c.name; /*EOS*/
create table ${observatory_db_name}.result_deposited_year stored as parquet as create table ${observatory_db_name}.result_deposited_year stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -346,7 +346,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
create table ${observatory_db_name}.result_deposited_year_country stored as parquet as create table ${observatory_db_name}.result_deposited_year_country stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -377,7 +377,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, r.year, c.code, c.name; /*EOS*/ cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, r.year, c.code, c.name; /*EOS*/
create table ${observatory_db_name}.result_deposited_datasource stored as parquet as create table ${observatory_db_name}.result_deposited_datasource stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -408,7 +408,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, d.name; /*EOS*/ cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, d.name; /*EOS*/
create table ${observatory_db_name}.result_deposited_datasource_country stored as parquet as create table ${observatory_db_name}.result_deposited_datasource_country stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -439,7 +439,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, d.name, c.code, c.name; /*EOS*/ cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, d.name, c.code, c.name; /*EOS*/
create table ${observatory_db_name}.result_deposited_organization stored as parquet as create table ${observatory_db_name}.result_deposited_organization stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -470,7 +470,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, o.name; /*EOS*/ cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, o.name; /*EOS*/
create table ${observatory_db_name}.result_deposited_organization_country stored as parquet as create table ${observatory_db_name}.result_deposited_organization_country stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -501,7 +501,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, o.name, c.code, c.name; /*EOS*/ cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, o.name, c.code, c.name; /*EOS*/
create table ${observatory_db_name}.result_deposited_funder stored as parquet as create table ${observatory_db_name}.result_deposited_funder stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,
@ -534,7 +534,7 @@ group by r.green, r.gold, case when rl.type is not null then true else false end
cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, p.funder; /*EOS*/ cc_licence, r.authors > 1, rpc.count > 1, rfc.count > 1, p.funder; /*EOS*/
create table ${observatory_db_name}.result_deposited_funder_country stored as parquet as create table ${observatory_db_name}.result_deposited_funder_country stored as parquet as
select select /*+ COALESCE(100) */
count(distinct r.id) as total, count(distinct r.id) as total,
r.green, r.green,
r.gold, r.gold,

10
dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step8.sql
View File

@ -39,21 +39,21 @@ where dtrce.datainfo.deletedbyinference = false and dtrce.datainfo.invisible = f
DROP TABLE IF EXISTS ${stats_db_name}.datasource_languages purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.datasource_languages purge; /*EOS*/
CREATE TABLE ${stats_db_name}.datasource_languages STORED AS PARQUET AS CREATE TABLE ${stats_db_name}.datasource_languages STORED AS PARQUET AS
SELECT substr(d.id, 4) AS id, langs.languages AS language SELECT /*+ COALESCE(100) */ 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 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; /*EOS*/ where d.datainfo.deletedbyinference=false and d.datainfo.invisible=false; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.datasource_oids purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.datasource_oids purge; /*EOS*/
CREATE TABLE ${stats_db_name}.datasource_oids STORED AS PARQUET AS CREATE TABLE ${stats_db_name}.datasource_oids STORED AS PARQUET AS
SELECT substr(d.id, 4) AS id, oids.ids AS oid SELECT /*+ COALESCE(100) */ substr(d.id, 4) AS id, oids.ids AS oid
FROM ${openaire_db_name}.datasource d LATERAL VIEW explode(d.originalid) oids AS ids FROM ${openaire_db_name}.datasource d LATERAL VIEW explode(d.originalid) oids AS ids
where d.datainfo.deletedbyinference=false and d.datainfo.invisible=false; /*EOS*/ where d.datainfo.deletedbyinference=false and d.datainfo.invisible=false; /*EOS*/
DROP TABLE IF EXISTS ${stats_db_name}.datasource_organizations purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.datasource_organizations purge; /*EOS*/
CREATE TABLE ${stats_db_name}.datasource_organizations STORED AS PARQUET 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 SELECT /*+ COALESCE(100) */ substr(r.target, 4) AS id, substr(r.source, 4) AS organization
FROM ${openaire_db_name}.relation r FROM ${openaire_db_name}.relation r
WHERE r.reltype = 'datasourceOrganization' and r.datainfo.deletedbyinference = false and r.source like '20|%' and r.datainfo.invisible=false; /*EOS*/ WHERE r.reltype = 'datasourceOrganization' and r.datainfo.deletedbyinference = false and r.source like '20|%' and r.datainfo.invisible=false; /*EOS*/
@ -62,10 +62,10 @@ WHERE r.reltype = 'datasourceOrganization' and r.datainfo.deletedbyinference = f
DROP TABLE IF EXISTS ${stats_db_name}.datasource_sources purge; /*EOS*/ DROP TABLE IF EXISTS ${stats_db_name}.datasource_sources purge; /*EOS*/
create table if not exists ${stats_db_name}.datasource_sources STORED AS PARQUET 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 select /*+ COALESCE(100) */ 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 from ${openaire_db_name}.datasource d lateral view explode(d.collectedfrom) cfrom as cf
where d.datainfo.deletedbyinference = false and d.datainfo.invisible=false; /*EOS*/ where d.datainfo.deletedbyinference = false and d.datainfo.invisible=false; /*EOS*/
CREATE OR REPLACE VIEW ${stats_db_name}.datasource_results AS CREATE OR REPLACE VIEW ${stats_db_name}.datasource_results AS
SELECT /*+ COALESCE(100) */ datasource AS id, id AS result SELECT datasource AS id, id AS result
FROM ${stats_db_name}.result_datasources; /*EOS*/ FROM ${stats_db_name}.result_datasources; /*EOS*/