added shell actions to download all 11 activities files from ORCID

This commit is contained in:
Enrico Ottonello 2020-04-28 20:25:09 +02:00
parent a1861b9eaa
commit 1edcd53581
2 changed files with 454 additions and 9 deletions

View File

@ -23,6 +23,7 @@ import org.mortbay.log.Log;
public class ActivitiesDecompressor {
private static final int MAX_XML_WORKS_PARSED = -1;
private static final int XML_WORKS_PARSED_COUNTER_LOG_INTERVAL = 100000;
public static void parseGzActivities(Configuration conf, String inputUri, Path outputPath)
throws Exception {
@ -127,7 +128,7 @@ public class ActivitiesDecompressor {
Log.warn(e);
}
if ((counter % 100000) == 0) {
if ((counter % XML_WORKS_PARSED_COUNTER_LOG_INTERVAL) == 0) {
Log.info("Current xml works parsed: " + counter);
}

View File

@ -1,9 +1,75 @@
<workflow-app name="Generate Orcid Authors DOIs Data" xmlns="uri:oozie:workflow:0.5">
<workflow-app name="Gen Orcid Authors DOIs" xmlns="uri:oozie:workflow:0.5">
<parameters>
<property>
<name>workingPath_activities</name>
<description>the working dir base path</description>
</property>
<property>
<name>shell_cmd_0</name>
<value>wget -O /tmp/ORCID_2019_activites_0.tar.gz https://orcid.figshare.com/ndownloader/files/18017660 ; hdfs dfs -copyFromLocal /tmp/ORCID_2019_activites_0.tar.gz /data/orcid_activities/ORCID_2019_activites_0.tar.gz
</value>
<description>the shell command that downloads and puts to hdfs orcid activity file 0</description>
</property>
<property>
<name>shell_cmd_1</name>
<value>wget -O /tmp/ORCID_2019_activites_1.tar.gz https://orcid.figshare.com/ndownloader/files/18017675 ; hdfs dfs -copyFromLocal /tmp/ORCID_2019_activites_1.tar.gz /data/orcid_activities/ORCID_2019_activites_1.tar.gz
</value>
<description>the shell command that downloads and puts to hdfs orcid activity file 1</description>
</property>
<property>
<name>shell_cmd_2</name>
<value>wget -O /tmp/ORCID_2019_activites_2.tar.gz https://orcid.figshare.com/ndownloader/files/18017717 ; hdfs dfs -copyFromLocal /tmp/ORCID_2019_activites_2.tar.gz /data/orcid_activities/ORCID_2019_activites_2.tar.gz
</value>
<description>the shell command that downloads and puts to hdfs orcid activity file 2</description>
</property>
<property>
<name>shell_cmd_3</name>
<value>wget -O /tmp/ORCID_2019_activites_3.tar.gz https://orcid.figshare.com/ndownloader/files/18017765 ; hdfs dfs -copyFromLocal /tmp/ORCID_2019_activites_3.tar.gz /data/orcid_activities/ORCID_2019_activites_3.tar.gz
</value>
<description>the shell command that downloads and puts to hdfs orcid activity file 3</description>
</property>
<property>
<name>shell_cmd_4</name>
<value>wget -O /tmp/ORCID_2019_activites_4.tar.gz https://orcid.figshare.com/ndownloader/files/18017831 ; hdfs dfs -copyFromLocal /tmp/ORCID_2019_activites_4.tar.gz /data/orcid_activities/ORCID_2019_activites_4.tar.gz
</value>
<description>the shell command that downloads and puts to hdfs orcid activity file 4</description>
</property>
<property>
<name>shell_cmd_5</name>
<value>wget -O /tmp/ORCID_2019_activites_5.tar.gz https://orcid.figshare.com/ndownloader/files/18017987 ; hdfs dfs -copyFromLocal /tmp/ORCID_2019_activites_5.tar.gz /data/orcid_activities/ORCID_2019_activites_5.tar.gz
</value>
<description>the shell command that downloads and puts to hdfs orcid activity file 5</description>
</property>
<property>
<name>shell_cmd_6</name>
<value>wget -O /tmp/ORCID_2019_activites_6.tar.gz https://orcid.figshare.com/ndownloader/files/18018053 ; hdfs dfs -copyFromLocal /tmp/ORCID_2019_activites_6.tar.gz /data/orcid_activities/ORCID_2019_activites_6.tar.gz
</value>
<description>the shell command that downloads and puts to hdfs orcid activity file 6</description>
</property>
<property>
<name>shell_cmd_7</name>
<value>wget -O /tmp/ORCID_2019_activites_7.tar.gz https://orcid.figshare.com/ndownloader/files/18018023 ; hdfs dfs -copyFromLocal /tmp/ORCID_2019_activites_7.tar.gz /data/orcid_activities/ORCID_2019_activites_7.tar.gz
</value>
<description>the shell command that downloads and puts to hdfs orcid activity file 7</description>
</property>
<property>
<name>shell_cmd_8</name>
<value>wget -O /tmp/ORCID_2019_activites_8.tar.gz https://orcid.figshare.com/ndownloader/files/18018248 ; hdfs dfs -copyFromLocal /tmp/ORCID_2019_activites_8.tar.gz /data/orcid_activities/ORCID_2019_activites_8.tar.gz
</value>
<description>the shell command that downloads and puts to hdfs orcid activity file 8</description>
</property>
<property>
<name>shell_cmd_9</name>
<value>wget -O /tmp/ORCID_2019_activites_9.tar.gz https://orcid.figshare.com/ndownloader/files/18018029 ; hdfs dfs -copyFromLocal /tmp/ORCID_2019_activites_9.tar.gz /data/orcid_activities/ORCID_2019_activites_9.tar.gz
</value>
<description>the shell command that downloads and puts to hdfs orcid activity file 9</description>
</property>
<property>
<name>shell_cmd_X</name>
<value>wget -O /tmp/ORCID_2019_activites_X.tar.gz https://orcid.figshare.com/ndownloader/files/18018182 ; hdfs dfs -copyFromLocal /tmp/ORCID_2019_activites_X.tar.gz /data/orcid_activities/ORCID_2019_activites_X.tar.gz
</value>
<description>the shell command that downloads and puts to hdfs orcid activity file X</description>
</property>
</parameters>
<start to="ResetWorkingPath"/>
@ -18,16 +84,47 @@
<delete path='${workingPath_activities}/output'/>
<mkdir path='${workingPath_activities}/output'/>
</fs>
<ok to="fork_generate_orcid_authors_dois"/>
<ok to="fork_gen_orcid_authors_dois"/>
<error to="Kill"/>
</action>
<fork name = "fork_generate_orcid_authors_dois">
<path start = "Generate_Orcid_Authors_DOIs_0"/>
<path start = "Generate_Orcid_Authors_DOIs_1"/>
<fork name = "fork_gen_orcid_authors_dois">
<path start = "check_exist_on_hdfs_activities_0"/>
<path start = "check_exist_on_hdfs_activities_1"/>
<path start = "check_exist_on_hdfs_activities_2"/>
<path start = "check_exist_on_hdfs_activities_3"/>
<path start = "check_exist_on_hdfs_activities_4"/>
<path start = "check_exist_on_hdfs_activities_5"/>
<path start = "check_exist_on_hdfs_activities_6"/>
<path start = "check_exist_on_hdfs_activities_7"/>
<path start = "check_exist_on_hdfs_activities_8"/>
<path start = "check_exist_on_hdfs_activities_9"/>
<path start = "check_exist_on_hdfs_activities_X"/>
</fork>
<action name="Generate_Orcid_Authors_DOIs_0">
<decision name="check_exist_on_hdfs_activities_0">
<switch>
<case to="Gen_Orcid_Authors_DOIs_0">
${fs:exists(concat(workingPath_activities,'/ORCID_2019_activites_0.tar.gz'))}
</case>
<default to="Download_0" />
</switch>
</decision>
<action name="Download_0">
<shell xmlns="uri:oozie:shell-action:0.1">
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<exec>bash</exec>
<argument>-c</argument>
<argument>${shell_cmd_0}</argument>
<capture-output/>
</shell>
<ok to="Gen_Orcid_Authors_DOIs_0"/>
<error to="Kill"/>
</action>
<action name="Gen_Orcid_Authors_DOIs_0">
<java>
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
@ -41,7 +138,29 @@
<error to="Kill"/>
</action>
<action name="Generate_Orcid_Authors_DOIs_1">
<decision name="check_exist_on_hdfs_activities_1">
<switch>
<case to="Gen_Orcid_Authors_DOIs_1">
${fs:exists(concat(workingPath_activities,'/ORCID_2019_activites_1.tar.gz'))}
</case>
<default to="Download_1" />
</switch>
</decision>
<action name="Download_1">
<shell xmlns="uri:oozie:shell-action:0.1">
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<exec>bash</exec>
<argument>-c</argument>
<argument>${shell_cmd_1}</argument>
<capture-output/>
</shell>
<ok to="Gen_Orcid_Authors_DOIs_1"/>
<error to="Kill"/>
</action>
<action name="Gen_Orcid_Authors_DOIs_1">
<java>
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
@ -55,6 +174,331 @@
<error to="Kill"/>
</action>
<decision name="check_exist_on_hdfs_activities_2">
<switch>
<case to="Gen_Orcid_Authors_DOIs_2">
${fs:exists(concat(workingPath_activities,'/ORCID_2019_activites_2.tar.gz'))}
</case>
<default to="Download_2" />
</switch>
</decision>
<action name="Download_2">
<shell xmlns="uri:oozie:shell-action:0.1">
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<exec>bash</exec>
<argument>-c</argument>
<argument>${shell_cmd_2}</argument>
<capture-output/>
</shell>
<ok to="Gen_Orcid_Authors_DOIs_2"/>
<error to="Kill"/>
</action>
<action name="Gen_Orcid_Authors_DOIs_2">
<java>
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<main-class>eu.dnetlib.doiboost.orcid.OrcidAuthorsDOIsDataGen</main-class>
<arg>-d</arg><arg>${workingPath_activities}/</arg>
<arg>-n</arg><arg>${nameNode}</arg>
<arg>-f</arg><arg>ORCID_2019_activites_2.tar.gz</arg>
<arg>-o</arg><arg>output/authors_dois_2.seq</arg>
</java>
<ok to="join_node"/>
<error to="Kill"/>
</action>
<decision name="check_exist_on_hdfs_activities_3">
<switch>
<case to="Gen_Orcid_Authors_DOIs_3">
${fs:exists(concat(workingPath_activities,'/ORCID_2019_activites_3.tar.gz'))}
</case>
<default to="Download_3" />
</switch>
</decision>
<action name="Download_3">
<shell xmlns="uri:oozie:shell-action:0.1">
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<exec>bash</exec>
<argument>-c</argument>
<argument>${shell_cmd_3}</argument>
<capture-output/>
</shell>
<ok to="Gen_Orcid_Authors_DOIs_3"/>
<error to="Kill"/>
</action>
<action name="Gen_Orcid_Authors_DOIs_3">
<java>
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<main-class>eu.dnetlib.doiboost.orcid.OrcidAuthorsDOIsDataGen</main-class>
<arg>-d</arg><arg>${workingPath_activities}/</arg>
<arg>-n</arg><arg>${nameNode}</arg>
<arg>-f</arg><arg>ORCID_2019_activites_3.tar.gz</arg>
<arg>-o</arg><arg>output/authors_dois_3.seq</arg>
</java>
<ok to="join_node"/>
<error to="Kill"/>
</action>
<decision name="check_exist_on_hdfs_activities_4">
<switch>
<case to="Gen_Orcid_Authors_DOIs_4">
${fs:exists(concat(workingPath_activities,'/ORCID_2019_activites_4.tar.gz'))}
</case>
<default to="Download_4" />
</switch>
</decision>
<action name="Download_4">
<shell xmlns="uri:oozie:shell-action:0.1">
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<exec>bash</exec>
<argument>-c</argument>
<argument>${shell_cmd_4}</argument>
<capture-output/>
</shell>
<ok to="Gen_Orcid_Authors_DOIs_4"/>
<error to="Kill"/>
</action>
<action name="Gen_Orcid_Authors_DOIs_4">
<java>
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<main-class>eu.dnetlib.doiboost.orcid.OrcidAuthorsDOIsDataGen</main-class>
<arg>-d</arg><arg>${workingPath_activities}/</arg>
<arg>-n</arg><arg>${nameNode}</arg>
<arg>-f</arg><arg>ORCID_2019_activites_4.tar.gz</arg>
<arg>-o</arg><arg>output/authors_dois_4.seq</arg>
</java>
<ok to="join_node"/>
<error to="Kill"/>
</action>
<decision name="check_exist_on_hdfs_activities_5">
<switch>
<case to="Gen_Orcid_Authors_DOIs_5">
${fs:exists(concat(workingPath_activities,'/ORCID_2019_activites_5.tar.gz'))}
</case>
<default to="Download_5" />
</switch>
</decision>
<action name="Download_5">
<shell xmlns="uri:oozie:shell-action:0.1">
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<exec>bash</exec>
<argument>-c</argument>
<argument>${shell_cmd_5}</argument>
<capture-output/>
</shell>
<ok to="Gen_Orcid_Authors_DOIs_5"/>
<error to="Kill"/>
</action>
<action name="Gen_Orcid_Authors_DOIs_5">
<java>
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<main-class>eu.dnetlib.doiboost.orcid.OrcidAuthorsDOIsDataGen</main-class>
<arg>-d</arg><arg>${workingPath_activities}/</arg>
<arg>-n</arg><arg>${nameNode}</arg>
<arg>-f</arg><arg>ORCID_2019_activites_5.tar.gz</arg>
<arg>-o</arg><arg>output/authors_dois_5.seq</arg>
</java>
<ok to="join_node"/>
<error to="Kill"/>
</action>
<decision name="check_exist_on_hdfs_activities_6">
<switch>
<case to="Gen_Orcid_Authors_DOIs_6">
${fs:exists(concat(workingPath_activities,'/ORCID_2019_activites_6.tar.gz'))}
</case>
<default to="Download_6" />
</switch>
</decision>
<action name="Download_6">
<shell xmlns="uri:oozie:shell-action:0.1">
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<exec>bash</exec>
<argument>-c</argument>
<argument>${shell_cmd_6}</argument>
<capture-output/>
</shell>
<ok to="Gen_Orcid_Authors_DOIs_6"/>
<error to="Kill"/>
</action>
<action name="Gen_Orcid_Authors_DOIs_6">
<java>
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<main-class>eu.dnetlib.doiboost.orcid.OrcidAuthorsDOIsDataGen</main-class>
<arg>-d</arg><arg>${workingPath_activities}/</arg>
<arg>-n</arg><arg>${nameNode}</arg>
<arg>-f</arg><arg>ORCID_2019_activites_6.tar.gz</arg>
<arg>-o</arg><arg>output/authors_dois_6.seq</arg>
</java>
<ok to="join_node"/>
<error to="Kill"/>
</action>
<decision name="check_exist_on_hdfs_activities_7">
<switch>
<case to="Gen_Orcid_Authors_DOIs_7">
${fs:exists(concat(workingPath_activities,'/ORCID_2019_activites_7.tar.gz'))}
</case>
<default to="Download_7" />
</switch>
</decision>
<action name="Download_7">
<shell xmlns="uri:oozie:shell-action:0.1">
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<exec>bash</exec>
<argument>-c</argument>
<argument>${shell_cmd_7}</argument>
<capture-output/>
</shell>
<ok to="Gen_Orcid_Authors_DOIs_7"/>
<error to="Kill"/>
</action>
<action name="Gen_Orcid_Authors_DOIs_7">
<java>
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<main-class>eu.dnetlib.doiboost.orcid.OrcidAuthorsDOIsDataGen</main-class>
<arg>-d</arg><arg>${workingPath_activities}/</arg>
<arg>-n</arg><arg>${nameNode}</arg>
<arg>-f</arg><arg>ORCID_2019_activites_7.tar.gz</arg>
<arg>-o</arg><arg>output/authors_dois_7.seq</arg>
</java>
<ok to="join_node"/>
<error to="Kill"/>
</action>
<decision name="check_exist_on_hdfs_activities_8">
<switch>
<case to="Gen_Orcid_Authors_DOIs_8">
${fs:exists(concat(workingPath_activities,'/ORCID_2019_activites_8.tar.gz'))}
</case>
<default to="Download_8" />
</switch>
</decision>
<action name="Download_8">
<shell xmlns="uri:oozie:shell-action:0.1">
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<exec>bash</exec>
<argument>-c</argument>
<argument>${shell_cmd_8}</argument>
<capture-output/>
</shell>
<ok to="Gen_Orcid_Authors_DOIs_8"/>
<error to="Kill"/>
</action>
<action name="Gen_Orcid_Authors_DOIs_8">
<java>
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<main-class>eu.dnetlib.doiboost.orcid.OrcidAuthorsDOIsDataGen</main-class>
<arg>-d</arg><arg>${workingPath_activities}/</arg>
<arg>-n</arg><arg>${nameNode}</arg>
<arg>-f</arg><arg>ORCID_2019_activites_8.tar.gz</arg>
<arg>-o</arg><arg>output/authors_dois_8.seq</arg>
</java>
<ok to="join_node"/>
<error to="Kill"/>
</action>
<decision name="check_exist_on_hdfs_activities_9">
<switch>
<case to="Gen_Orcid_Authors_DOIs_9">
${fs:exists(concat(workingPath_activities,'/ORCID_2019_activites_9.tar.gz'))}
</case>
<default to="Download_9" />
</switch>
</decision>
<action name="Download_9">
<shell xmlns="uri:oozie:shell-action:0.1">
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<exec>bash</exec>
<argument>-c</argument>
<argument>${shell_cmd_9}</argument>
<capture-output/>
</shell>
<ok to="Gen_Orcid_Authors_DOIs_9"/>
<error to="Kill"/>
</action>
<action name="Gen_Orcid_Authors_DOIs_9">
<java>
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<main-class>eu.dnetlib.doiboost.orcid.OrcidAuthorsDOIsDataGen</main-class>
<arg>-d</arg><arg>${workingPath_activities}/</arg>
<arg>-n</arg><arg>${nameNode}</arg>
<arg>-f</arg><arg>ORCID_2019_activites_9.tar.gz</arg>
<arg>-o</arg><arg>output/authors_dois_9.seq</arg>
</java>
<ok to="join_node"/>
<error to="Kill"/>
</action>
<decision name="check_exist_on_hdfs_activities_X">
<switch>
<case to="Gen_Orcid_Authors_DOIs_X">
${fs:exists(concat(workingPath_activities,'/ORCID_2019_activites_X.tar.gz'))}
</case>
<default to="Download_X" />
</switch>
</decision>
<action name="Download_X">
<shell xmlns="uri:oozie:shell-action:0.1">
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<exec>bash</exec>
<argument>-c</argument>
<argument>${shell_cmd_X}</argument>
<capture-output/>
</shell>
<ok to="Gen_Orcid_Authors_DOIs_X"/>
<error to="Kill"/>
</action>
<action name="Gen_Orcid_Authors_DOIs_X">
<java>
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<main-class>eu.dnetlib.doiboost.orcid.OrcidAuthorsDOIsDataGen</main-class>
<arg>-d</arg><arg>${workingPath_activities}/</arg>
<arg>-n</arg><arg>${nameNode}</arg>
<arg>-f</arg><arg>ORCID_2019_activites_X.tar.gz</arg>
<arg>-o</arg><arg>output/authors_dois_X.seq</arg>
</java>
<ok to="join_node"/>
<error to="Kill"/>
</action>
<join name = "join_node" to = "End"/>
<end name="End"/>