code for importing values from wikidata

diff --git a/wikidata/create.sql b/wikidata/create.sql
new file mode 100644 (file)
index 0000000..f6327bd
--- /dev/null
+++ b/wikidata/create.sql
@@ -0,0 +1,75 @@
+DROP TABLE entity;
+DROP TABLE entity_label;
+DROP TABLE entity_description;
+DROP TABLE entity_alias;
+DROP TABLE entity_link;
+DROP TABLE entity_property;
+
+CREATE TABLE entity (
+    entity_id           bigint,
+    title               text,
+    pid                 bigint,
+    qid                 bigint,
+    datatype            text,
+    CONSTRAINT pk_entity PRIMARY KEY(entity_id)
+);
+
+CREATE TABLE entity_label  (
+    entity_id           bigint,
+    language            text,
+    label               text,
+    CONSTRAINT pk_entity_label PRIMARY KEY(entity_id,language)
+);
+
+CREATE TABLE entity_description  (
+    entity_id           bigint,
+    language            text,
+    description         text,
+    CONSTRAINT pk_entity_description PRIMARY KEY(entity_id,language)
+);
+
+CREATE TABLE entity_alias  (
+    entity_id           bigint,
+    language            text,
+    alias               text,
+    CONSTRAINT pk_entity_alias PRIMARY KEY(entity_id,language,alias)
+);
+
+CREATE TABLE entity_link  (
+    entity_id           bigint,
+    target              text,
+    value               text,
+    CONSTRAINT pk_entity_link PRIMARY KEY(entity_id,target)
+);
+
+CREATE TABLE entity_link_hit  (
+    entity_id           bigint,
+    target              text,
+    value               text,
+    hits               bigint,
+    CONSTRAINT pk_entity_link_hit PRIMARY KEY(entity_id,target)
+);
+
+CREATE TABLE link_hit  (
+    target              text,
+    value               text,
+    hits               bigint,
+    CONSTRAINT pk_link_hit PRIMARY KEY(target,value)
+);
+
+CREATE TABLE entity_property  (
+    entity_id           bigint,
+    order_id            bigint,
+    pid                 bigint,
+    string              text,
+    toqid               bigint,
+    location            geometry,
+    datetime           timestamp with time zone,
+    CONSTRAINT pk_entity_property PRIMARY KEY(entity_id, order_id)
+);
+
+CREATE TABLE import_link_hit  (
+    target              text,
+    value               text,
+    hits               bigint
+);

diff --git a/wikidata/import.sh b/wikidata/import.sh
new file mode 100755 (executable)
index 0000000..c701b03
--- /dev/null
+++ b/wikidata/import.sh
@@ -0,0 +1,24 @@
+PSQL=/usr/lib/postgresql/9.2/bin/psql -d wikidata
+
+cat create.sql | $PSQL
+
+cat entity.csv | $PSQL -c "COPY entity from STDIN WITH CSV"
+cat entity_label.csv | $PSQL -c "COPY entity_label from STDIN WITH CSV"
+cat entity_description.csv | $PSQL -c "COPY entity_description from STDIN WITH CSV"
+cat entity_alias.csv | $PSQL -c "COPY entity_alias from STDIN WITH CSV"
+cat entity_link.csv | $PSQL -c "COPY entity_link from STDIN WITH CSV"
+cat entity_property.csv | $PSQL -c "COPY entity_property from STDIN WITH CSV"
+
+$PSQL -c "create index idx_entity_link_target on entity_link using btree (target,value)"
+$PSQL -c "create index idx_entity_qid on entity using btree (qid)"
+$PSQL -c "create table property_label_en as select pid,null::text as label from entity where pid is not null"
+$PSQL -c "update property_label_en set label = x.label from (select pid,label,language from entity join entity_label using (entity_id) where pid is not null and language = 'en') as x where x.pid = property_label_en.pid"
+$PSQL -c "create unique index idx_property_label_en on property_label_en using btree (pid)"
+$PSQL -c "alter table entity add column label_en text"
+$PSQL -c "update entity set label_en = label from entity_label where entity.entity_id = entity_label.entity_id and language = 'en'"
+$PSQL -c "alter table entity add column description_en text"
+$PSQL -c "update entity set description_en = description from entity_description where entity.entity_id = entity_description.entity_id and language = 'en'"
+
+cat totals.txt | $PSQL -c "COPY import_link_hit from STDIN WITH CSV DELIMITER ' '"
+$PSQL -c "insert into link_hit select target||'wiki', catch_decode_url_part(value), sum(hits) from import_link_hit group by target||'wiki', catch_decode_url_part(value)"
+$PSQL -c "insert into entity_link_hit select entity_id, target, value, coalesce(hits,0) from entity_link left outer join link_hit using (target, value)"

diff --git a/wikidata/parse.php b/wikidata/parse.php
new file mode 100755 (executable)
index 0000000..e1b0687
--- /dev/null
+++ b/wikidata/parse.php
@@ -0,0 +1,188 @@
+#!/usr/bin/php -Cq
+<?php
+
+$hFile = @fopen("wikidatawiki-20130623-pages-articles.xml", "r");
+
+$hFileEntity = fopen("entity.csv", "w");
+$hFileEntityLabel = fopen("entity_label.csv", "w");
+$hFileEntityDescription = fopen("entity_description.csv", "w");
+$hFileEntityAlias = fopen("entity_alias.csv", "w");
+$hFileEntityLink = fopen("entity_link.csv", "w");
+$hFileEntityProperty = fopen("entity_property.csv", "w");
+
+$iCount = 0;
+
+$sTitle = '';
+$iNS = false;
+$iID = false;
+
+if ($hFile) {
+       while (($sLine = fgets($hFile, 4000000)) !== false) {
+               if (substr($sLine, 0, 11) == '    <title>') {
+                       $sTitle = substr($sLine, 11, -9);
+               }
+               else if (substr($sLine, 0, 8) == '    <ns>') {
+                       $iNS = (int)substr($sLine, 8, -6);
+               }
+               else if (substr($sLine, 0, 8) == '    <id>') {
+                       $iID = (int)substr($sLine, 8, -6);
+               }
+               else if (substr($sLine, 0, 33) == '      <text xml:space="preserve">') {
+                       if ($iNS == -2) continue;
+                       if ($iNS == -1) continue;
+                       if ($iNS == 1) continue;
+                       if ($iNS == 2) continue;
+                       if ($iNS == 3) continue;
+                       if ($iNS == 4) continue;
+                       if ($iNS == 5) continue;
+                       if ($iNS == 6) continue;
+                       if ($iNS == 7) continue;
+                       if ($iNS == 8) continue;
+                       if ($iNS == 9) continue;
+                       if ($iNS == 10) continue;
+                       if ($iNS == 11) continue;
+                       if ($iNS == 12) continue;
+                       if ($iNS == 13) continue;
+                       if ($iNS == 14) continue;
+                       if ($iNS == 15) continue;
+                       if ($iNS == 121) continue;
+                       if ($iNS == 123) continue;
+                       if ($iNS == 829) continue;
+                       if ($iNS == 1198) continue;
+                       if ($iNS == 1199) continue;
+                       $sText = html_entity_decode(substr($sLine, 33, -8), ENT_COMPAT, 'UTF-8');
+                       $aArticle = json_decode($sText, true);
+
+                       if (array_diff(array_keys($aArticle), array("label", "description", "aliases", "links", "entity", "claims", "datatype")) != array()) {
+                               // DEBUG
+                               var_dump($sTitle);
+                               var_dump(array_keys($aArticle));
+                               var_dump($aArticle);
+                               exit;
+                       }
+
+                       $iPID = $iQID = null;
+                       if ($aArticle['entity'][0] == 'p') {
+                               $iPID = (int)substr($aArticle['entity'], 1);
+                       } else if ($aArticle['entity'][0] == 'q') {
+                               $iQID = (int)substr($aArticle['entity'], 1);
+                       } else {
+                               continue;
+                       }
+
+                       echo ".";
+
+                       fputcsv($hFileEntity, array($iID,$sTitle,$iPID,$iQID,@$aArticle['datatype']));
+
+                       foreach($aArticle['label'] as $sLang => $sLabel) {
+                               fputcsv($hFileEntityLabel, array($iID,$sLang,$sLabel));
+                               // echo "insert into entity_label values (".$iID.",'".pg_escape_string($sLang)."','".pg_escape_string($sLabel)."');\n";
+                       }
+
+                       foreach($aArticle['description'] as $sLang => $sLabel) {
+                               fputcsv($hFileEntityDescription, array($iID,$sLang,$sLabel));
+                               // echo "insert into entity_description values (".$iID.",'".pg_escape_string($sLang)."','".pg_escape_string($sLabel)."');\n";
+                       }
+
+                       foreach($aArticle['aliases'] as $sLang => $aLabels) {
+                               $aUniqueAlias = array();
+                               foreach($aLabels as $sLabel) {
+                                       if (!isset($aUniqueAlias[$sLabel]) && $sLabel) {
+                                       fputcsv($hFileEntityAlias, array($iID,$sLang,$sLabel));
+                                       // echo "insert into entity_alias values (".$iID.",'".pg_escape_string($sLang)."','".pg_escape_string($sLabel)."');\n";
+                                       $aUniqueAlias[$sLabel] = true;
+                               }
+                       }
+               }
+
+               foreach($aArticle['links'] as $sLang => $sLabel) {
+                       fputcsv($hFileEntityLink, array($iID,$sLang,$sLabel));
+                       // echo "insert into entity_link values (".$iID.",'".pg_escape_string($sLang)."','".pg_escape_string($sLabel)."');\n";
+               }
+
+
+               if (isset($aArticle['claims'])) {
+
+                       foreach($aArticle['claims'] as $iClaim => $aClaim) {
+
+                               $bFail = false;
+                       if ($aClaim['m'][0] == 'novalue') continue;
+                       if ($aClaim['m'][0] == 'somevalue') continue;
+                       $iPID = (int)$aClaim['m'][1];
+                       if ($aClaim['m'][0] != 'value') $bFail = true;
+                       if ($aClaim['m'][2]== 'wikibase-entityid') {
+
+                               if ($aClaim['m'][3]['entity-type'] != 'item') $bFail = true;
+                               fputcsv($hFileEntityProperty, array($iID,$iClaim,$iPID,null,$aClaim['m'][3]['numeric-id'],null,null));
+                               // echo "insert into entity_property values (nextval('seq_entity_property'),".$iID.",".$iPID.",null,".$aClaim['m'][3]['numeric-id'].",null);\n";
+
+                       } elseif ($aClaim['m'][2] == 'globecoordinate') {
+
+                               if ($aClaim['m'][3]['globe'] != 'http://www.wikidata.org/entity/Q2') $bFail = true;
+                               fputcsv($hFileEntityProperty, array($iID,$iClaim,$iPID,null,null,"SRID=4326;POINT(".((float)$aClaim['m'][3]['longitude'])." ".((float)$aClaim['m'][3]['latitude']).")",null));
+                               // echo "insert into entity_property values (nextval('seq_entity_property'),".$iID.",".$iPID.",null,null,ST_SetSRID(ST_MakePoint(".((float)$aClaim['m'][3]['longitude']).", ".((float)$aClaim['m'][3]['latitude'])."),4326));\n";
+
+                       } elseif ($aClaim['m'][2] == 'time') {
+                               // TODO!
+/*
+               if ($aClaim['m'][3]['calendarmodel'] == 'http://www.wikidata.org/entity/Q1985727') {
+                       // Gregorian
+                       if (preg_match('#(\\+|-)0*([0-9]{4})-([0-9]{2}-[0-9]{2}T[0-9]{2}:[0-9]{2}:[0-9]{2})Z#', $aClaim['m'][3]['time'], $aMatch)) {
+                               if ((int)$aMatch[2] < 4700 && ) {
+                                       $sDateString = $aMatch[2].'-'.$aMatch[3].($aClaim['m'][3]['timezone']>=0?'+':'').$aClaim['m'][3]['timezone'].($aMatch[1]=='-'?' bc':'');
+                                       fputcsv($hFileEntityProperty, array($iID,$iClaim,$iPID,null,null,null,$sDateString));
+                               }
+                       } else {
+//                             $bFail = true;
+                       }
+               } elseif ( $aClaim['m'][3]['calendarmodel'] != 'http://www.wikidata.org/entity/Q1985786') {
+/ *
+                       // Julian
+                       if (preg_match('#(\\+|-)0*([0-9]{4})-([0-9]{2})-([0-9]{2})T([0-9]{2}:[0-9]{2}:[0-9]{2})Z#', $aClaim['m'][3]['time'], $aMatch)) {
+var_dump($aMatch);
+exit;
+$iDayCount = juliantojd(2, 11, 1732);
+var_dump($iDayCount, jdtogregorian($iDayCount));
+                       } else {
+                               $bFail = true;
+exit;
+                       }
+exit;
+* /
+               } else {
+//                     $bFail = true;
+               }
+*/
+                       } elseif ($aClaim['m'][2] == 'string') {
+
+                               // echo "insert into entity_property values (nextval('seq_entity_property'),".$iID.",".$iPID.",'".pg_escape_string($aClaim['m'][3])."',null,null);\n";
+                               fputcsv($hFileEntityProperty, array($iID,$iClaim,$iPID,$aClaim['m'][3],null,null,null));
+
+                       } else {
+
+                               $bFail = true;
+
+                       }
+
+                       // Don't care about sources:    if ($aClaim['refs'] != array()) $bFail = true;
+
+                       if ($bFail) {
+                               var_dump($sTitle);
+                               var_dump($aClaim);
+                       } else {
+                               // process
+                       }
+
+               }
+
+                       }
+               }
+       }
+       fclose($hFile);
+       fclose($hFileEntity);
+       fclose($hFileEntityLabel);
+       fclose($hFileEntityDescription);
+       fclose($hFileEntityAlias);
+       fclose($hFileEntityLink);
+       fclose($hFileEntityProperty);
+}