handle invalid input in file size and description markup for datagov import
handle invalid input in file size and description markup for datagov import


Former-commit-id: 6c34fd9f8a95d5075ab5fb36f9ee2721736dcaec

<?php <?php
   
require_once '../include/common.inc.php'; require_once '../include/common.inc.php';
//function createFOIDocumentsDesignDoc() { //function createFOIDocumentsDesignDoc() {
   
$foidb = $server->get_db('disclosr-foidocuments'); $foidb = $server->get_db('disclosr-foidocuments');
$obj = new stdClass(); $obj = new stdClass();
$obj->_id = "_design/" . urlencode("app"); $obj->_id = "_design/" . urlencode("app");
$obj->language = "javascript"; $obj->language = "javascript";
$obj->views->all->map = "function(doc) { emit(doc._id, doc); };"; $obj->views->all->map = "function(doc) { emit(doc._id, doc); };";
$obj->views->byDate->map = "function(doc) { emit(doc.date, doc); };"; $obj->views->byDate->map = "function(doc) { emit(doc.date, doc); };";
$obj->views->byDateMonthYear->map = "function(doc) { emit(doc.date, doc); };"; $obj->views->byDateMonthYear->map = "function(doc) { emit(doc.date, doc); };";
$obj->views->byDateMonthYear->reduce = "_count"; $obj->views->byDateMonthYear->reduce = "_count";
$obj->views->byAgencyID->map = "function(doc) { emit(doc.agencyID, doc); };"; $obj->views->byAgencyID->map = "function(doc) { emit(doc.agencyID, doc); };";
$obj->views->byAgencyID->reduce = "_count"; $obj->views->byAgencyID->reduce = "_count";
$obj->views->fieldNames->map = ' $obj->views->fieldNames->map = '
function(doc) { function(doc) {
for(var propName in doc) { for(var propName in doc) {
emit(propName, doc._id); emit(propName, doc._id);
} }
   
}'; }';
$obj->views->fieldNames->reduce = 'function (key, values, rereduce) { $obj->views->fieldNames->reduce = 'function (key, values, rereduce) {
return values.length; return values.length;
}'; }';
// allow safe updates (even if slightly slower due to extra: rev-detection check). // allow safe updates (even if slightly slower due to extra: rev-detection check).
$foidb->save($obj, true); $foidb->save($obj, true);
   
   
//function createDocumentsDesignDoc() { //function createDocumentsDesignDoc() {
$docdb = $server->get_db('disclosr-documents'); $docdb = $server->get_db('disclosr-documents');
   
$obj = new stdClass(); $obj = new stdClass();
$obj->_id = "_design/" . urlencode("app"); $obj->_id = "_design/" . urlencode("app");
$obj->language = "javascript"; $obj->language = "javascript";
$obj->views->web_server->map = "function(doc) {\n emit(doc.web_server, 1);\n}"; $obj->views->web_server->map = "function(doc) {\n emit(doc.web_server, 1);\n}";
$obj->views->web_server->reduce = "_sum"; $obj->views->web_server->reduce = "_sum";
$obj->views->byAgency->map = "function(doc) {\n emit(doc.agencyID, 1);\n}"; $obj->views->byAgency->map = "function(doc) {\n emit(doc.agencyID, 1);\n}";
$obj->views->byAgency->reduce = "_sum"; $obj->views->byAgency->reduce = "_sum";
$obj->views->byURL->map = "function(doc) {\n emit(doc.url, doc);\n}"; $obj->views->byURL->map = "function(doc) {\n emit(doc.url, doc);\n}";
$obj->views->agency->map = "function(doc) {\n emit(doc.agencyID, doc);\n}"; $obj->views->agency->map = "function(doc) {\n emit(doc.agencyID, doc);\n}";
$obj->views->byWebServer->map = "function(doc) {\n emit(doc.web_server, doc);\n}"; $obj->views->byWebServer->map = "function(doc) {\n emit(doc.web_server, doc);\n}";
   
$obj->views->datasets->map = "function(doc) {\nif (doc.fieldName == \"data\") {\n emit(doc._id, doc);\n}\n}"; $obj->views->datasets->map = "function(doc) {\nif (doc.fieldName == \"data\") {\n emit(doc._id, doc);\n}\n}";
  $obj->views->datasetGroups->map = "function(doc) {\nif (doc.fieldName == \"data\") {\n doc.metadata[\"data.gov.au Category\"] && doc.metadata[\"data.gov.au Category\"].forEach(function(tag) {\n emit(tag, doc.url); \n });\n}\n}";
$obj->views->getValidationRequired->map = "function(doc) {\nif (doc.mime_type == \"text/html\" \n&& typeof(doc.validation) == \"undefined\") {\n emit(doc._id, doc._attachments);\n}\n}"; $obj->views->getValidationRequired->map = "function(doc) {\nif (doc.mime_type == \"text/html\" \n&& typeof(doc.validation) == \"undefined\") {\n emit(doc._id, doc._attachments);\n}\n}";
$docdb->save($obj, true); $docdb->save($obj, true);
   
   
   
   
//function createAgencyDesignDoc() { //function createAgencyDesignDoc() {
$db = $server->get_db('disclosr-agencies'); $db = $server->get_db('disclosr-agencies');
$obj = new stdClass(); $obj = new stdClass();
$obj->_id = "_design/" . urlencode("app"); $obj->_id = "_design/" . urlencode("app");
$obj->language = "javascript"; $obj->language = "javascript";
$obj->views->all->map = "function(doc) { emit(doc._id, doc); };"; $obj->views->all->map = "function(doc) { emit(doc._id, doc); };";
$obj->views->byABN->map = "function(doc) { emit(doc.abn, doc); };"; $obj->views->byABN->map = "function(doc) { emit(doc.abn, doc); };";
$obj->views->byCanonicalName->map = "function(doc) { $obj->views->byCanonicalName->map = "function(doc) {
if (doc.parentOrg || doc.orgType == 'FMA-DepartmentOfState') { if (doc.parentOrg || doc.orgType == 'FMA-DepartmentOfState') {
emit(doc.name, doc); emit(doc.name, doc);
} }
};"; };";
$obj->views->byDeptStateName->map = "function(doc) { $obj->views->byDeptStateName->map = "function(doc) {
if (doc.orgType == 'FMA-DepartmentOfState') { if (doc.orgType == 'FMA-DepartmentOfState') {
emit(doc.name, doc._id); emit(doc.name, doc._id);
} }
};"; };";
$obj->views->parentOrgs->map = "function(doc) { $obj->views->parentOrgs->map = "function(doc) {
if (doc.parentOrg) { if (doc.parentOrg) {
emit(doc._id, doc.parentOrg); emit(doc._id, doc.parentOrg);
} }
};"; };";
$obj->views->byName->map = 'function(doc) { $obj->views->byName->map = 'function(doc) {
if (typeof(doc["status"]) == "undefined" || doc["status"] != "suspended") { if (typeof(doc["status"]) == "undefined" || doc["status"] != "suspended") {
emit(doc.name, doc._id); emit(doc.name, doc._id);
if (typeof(doc.shortName) != "undefined" && doc.shortName != doc.name) { if (typeof(doc.shortName) != "undefined" && doc.shortName != doc.name) {
emit(doc.shortName, doc._id); emit(doc.shortName, doc._id);
} }
for (name in doc.otherNames) { for (name in doc.otherNames) {
if (doc.otherNames[name] != "" && doc.otherNames[name] != doc.name) { if (doc.otherNames[name] != "" && doc.otherNames[name] != doc.name) {
emit(doc.otherNames[name], doc._id); emit(doc.otherNames[name], doc._id);
} }
} }
for (name in doc.foiBodies) { for (name in doc.foiBodies) {
if (doc.foiBodies[name] != "" && doc.foiBodies[name] != doc.name) { if (doc.foiBodies[name] != "" && doc.foiBodies[name] != doc.name) {
emit(doc.foiBodies[name], doc._id); emit(doc.foiBodies[name], doc._id);
} }
} }
for (name in doc.positions) { for (name in doc.positions) {
if (doc.positions[name] != "" && doc.positions[name] != doc.name) { if (doc.positions[name] != "" && doc.positions[name] != doc.name) {
emit(doc.positions[name], doc._id); emit(doc.positions[name], doc._id);
} }
} }
} }
};'; };';
   
$obj->views->foiEmails->map = "function(doc) { $obj->views->foiEmails->map = "function(doc) {
emit(doc._id, doc.foiEmail); emit(doc._id, doc.foiEmail);
};"; };";
   
$obj->views->byLastModified->map = "function(doc) { emit(doc.metadata.lastModified, doc); }"; $obj->views->byLastModified->map = "function(doc) { emit(doc.metadata.lastModified, doc); }";
$obj->views->getActive->map = 'function(doc) { if (doc.status == "active") { emit(doc._id, doc); } };'; $obj->views->getActive->map = 'function(doc) { if (doc.status == "active") { emit(doc._id, doc); } };';
$obj->views->getSuspended->map = 'function(doc) { if (doc.status == "suspended") { emit(doc._id, doc); } };'; $obj->views->getSuspended->map = 'function(doc) { if (doc.status == "suspended") { emit(doc._id, doc); } };';
$obj->views->getScrapeRequired->map = "function(doc) { $obj->views->getScrapeRequired->map = "function(doc) {
   
var lastScrape = Date.parse(doc.metadata.lastScraped); var lastScrape = Date.parse(doc.metadata.lastScraped);
   
var today = new Date(); var today = new Date();
   
if (!lastScrape || lastScrape.getTime() + 1000 != today.getTime()) { if (!lastScrape || lastScrape.getTime() + 1000 != today.getTime()) {
emit(doc._id, doc); emit(doc._id, doc);
} }
   
};"; };";
$obj->views->showNamesABNs->map = "function(doc) { emit(doc._id, {name: doc.name, abn: doc.abn}); };"; $obj->views->showNamesABNs->map = "function(doc) { emit(doc._id, {name: doc.name, abn: doc.abn}); };";
$obj->views->getConflicts->map = "function(doc) { $obj->views->getConflicts->map = "function(doc) {
if (doc._conflicts) { if (doc._conflicts) {
emit(null, [doc._rev].concat(doc._conflicts)); emit(null, [doc._rev].concat(doc._conflicts));
} }
}"; }";
$obj->views->getStatistics->map = $obj->views->getStatistics->map =
"function(doc) { "function(doc) {
if (doc.statistics) { if (doc.statistics) {
for (var statisticSet in doc.statistics) { for (var statisticSet in doc.statistics) {
for (var statisticPeriod in doc.statistics[statisticSet]) { for (var statisticPeriod in doc.statistics[statisticSet]) {
emit([statisticSet,statisticPeriod], doc.statistics[statisticSet][statisticPeriod]['value']); emit([statisticSet,statisticPeriod], doc.statistics[statisticSet][statisticPeriod]['value']);
} }
} }
} }
}"; }";
$obj->views->getStatistics->reduce = '_sum'; $obj->views->getStatistics->reduce = '_sum';
// http://stackoverflow.com/questions/646628/javascript-startswith // http://stackoverflow.com/questions/646628/javascript-startswith
$obj->views->score->map = 'if(!String.prototype.startsWith){ $obj->views->score->map = 'if(!String.prototype.startsWith){
String.prototype.startsWith = function (str) { String.prototype.startsWith = function (str) {
return !this.indexOf(str); return !this.indexOf(str);
} }
} }
   
function(doc) { function(doc) {
count = 0; count = 0;
if (doc["status"] != "suspended") { if (doc["status"] != "suspended") {
for(var propName in doc) { for(var propName in doc) {
if(typeof(doc[propName]) != "undefined" && doc[propName] != "") { if(typeof(doc[propName]) != "undefined" && doc[propName] != "") {
count++; count++;
} }
} }
portfolio = doc.parentOrg; portfolio = doc.parentOrg;
if (doc.orgType == "FMA-DepartmentOfState") { if (doc.orgType == "FMA-DepartmentOfState") {
portfolio = doc._id; portfolio = doc._id;
} }
if (doc.orgType == "Court-Commonwealth" || doc.orgType == "FMA-DepartmentOfParliament") { if (doc.orgType == "Court-Commonwealth" || doc.orgType == "FMA-DepartmentOfParliament") {
portfolio = doc.orgType; portfolio = doc.orgType;
} }
emit(count+doc._id, {id:doc._id, name: doc.name, score:count, orgType: doc.orgType, portfolio:portfolio}); emit(count+doc._id, {id:doc._id, name: doc.name, score:count, orgType: doc.orgType, portfolio:portfolio});
} }
}'; }';
$obj->views->scoreHas->map = 'if(!String.prototype.startsWith){ $obj->views->scoreHas->map = 'if(!String.prototype.startsWith){
String.prototype.startsWith = function (str) { String.prototype.startsWith = function (str) {
return !this.indexOf(str); return !this.indexOf(str);
} }
} }
if(!String.prototype.endsWith){ if(!String.prototype.endsWith){
String.prototype.endsWith = function(suffix) { String.prototype.endsWith = function(suffix) {
    return this.indexOf(suffix, this.length - suffix.length) !== -1;     return this.indexOf(suffix, this.length - suffix.length) !== -1;
}; };
} }
function(doc) { function(doc) {
if (typeof(doc["status"]) == "undefined" || doc["status"] != "suspended") { if (typeof(doc["status"]) == "undefined" || doc["status"] != "suspended") {
for(var propName in doc) { for(var propName in doc) {
if(typeof(doc[propName]) != "undefined" && (propName.startsWith("has") || propName.endsWith("URL"))) { if(typeof(doc[propName]) != "undefined" && (propName.startsWith("has") || propName.endsWith("URL"))) {
emit(propName, 1); emit(propName, 1);
} }
} }
emit("total", 1); emit("total", 1);
} }
}'; }';
$obj->views->scoreHas->reduce = '_sum'; $obj->views->scoreHas->reduce = '_sum';
$obj->views->fieldNames->map = ' $obj->views->fieldNames->map = '
function(doc) { function(doc) {
for(var propName in doc) { for(var propName in doc) {
emit(propName, doc._id); emit(propName, doc._id);
} }
}'; }';
$obj->views->fieldNames->reduce = '_count'; $obj->views->fieldNames->reduce = '_count';
// allow safe updates (even if slightly slower due to extra: rev-detection check). // allow safe updates (even if slightly slower due to extra: rev-detection check).
$db->save($obj, true); $db->save($obj, true);
?> ?>
   
file:b/disclosr.iml (new)
  <?xml version="1.0" encoding="UTF-8"?>
  <module type="WEB_MODULE" version="4">
  <component name="FacetManager">
  <facet type="Python" name="Python">
  <configuration sdkName="" />
  </facet>
  </component>
  <component name="NewModuleRootManager" inherit-compiler-output="true">
  <exclude-output />
  <content url="file://$MODULE_DIR$" />
  <orderEntry type="inheritedJdk" />
  <orderEntry type="sourceFolder" forTests="false" />
  </component>
  </module>
 
 
  import ckanclient
  import couchdb
  from ckanclient import CkanApiError
  import re
 
 
  class LoaderError(Exception):
  pass
 
  # Instantiate the CKAN client.
  #ckan = ckanclient.CkanClient(base_location='http://localhost:5000/api', api_key='b47b24cd-591d-40c1-8677-d73101d56d1b')
  api_key = 'ff34526e-f794-4068-8235-fcbba38cd8bc'
  ckan = ckanclient.CkanClient(base_location='http://data.disclosurelo.gs/api',
  api_key=api_key)
  couch = couchdb.Server('http://127.0.0.1:5984/')
  #couch = couchdb.Server('http://192.168.1.113:5984/')
 
  # https://github.com/okfn/ckanext-importlib
  def munge(name):
  # convert spaces to underscores
  name = re.sub(' ', '_', name).lower()
  # convert symbols to dashes
  name = re.sub('[:]', '_-', name).lower()
  name = re.sub('[/]', '-', name).lower()
  # take out not-allowed characters
  name = re.sub('[^a-zA-Z0-9-_]', '', name).lower()
  # remove double underscores
  name = re.sub('__', '_', name).lower()
  return name
 
 
  def name_munge(input_name):
  return munge(input_name.replace(' ', '').replace('.', '_').replace('&', 'and'))
 
 
  docsdb = couch['disclosr-documents']
 
  if __name__ == "__main__":
  groups = {}
  for doc in docsdb.view('app/datasetGroups'):
  group_name = doc.key
  if group_name != "Not specified":
  pkg_name = filter(lambda x: x in '0123456789abcdefghijklmnopqrstuvwxyz-_',
  doc.value.replace("http://data.gov.au/dataset/", '').replace('/', '')[:100]);
  if group_name in groups.keys():
  groups[group_name] = list(set(groups[group_name] + [pkg_name]))
  else:
  groups[group_name] = [pkg_name]
 
  # add dataset to group(s)
  for group_name in groups.keys():
  if group_name != "Not specified":
  group_url = name_munge(group_name[:100])
  print group_name
  print groups[group_name]
  try:
  # Update the group details
  group_entity = ckan.group_entity_get(group_url)
  print "group "+group_name+" exists"
  if 'packages' in group_entity.keys():
  group_entity['packages'] = list(set(group_entity['packages'] + groups[group_name]))
  else:
  group_entity['packages'] = groups[group_name]
  ckan.group_entity_put(group_entity)
  except CkanApiError, e:
  if ckan.last_status == 404:
  print "group "+group_name+" does not exist, creating"
  group_entity = {
  'name': group_url,
  'title': group_name,
  'description': group_name,
  'packages': groups[group_name]
  }
  #print group_entity
  ckan.group_register_post(group_entity)
  elif ckan.last_status == 409:
  print "group already exists"
  else:
  raise LoaderError('Unexpected status %s adding to group under \'%s\': %r' % (
  ckan.last_status, pkg_name, e.args))
 
  # coding=utf-8
import ckanclient import ckanclient
import couchdb import couchdb
from ckanclient import CkanApiError from ckanclient import CkanApiError
import re import re
import html2text # aaronsw :( import html2text # aaronsw :(
import ckanapi # https://github.com/open-data/ckanapi import ckanapi # https://github.com/open-data/ckanapi
  import scrape
  import datetime, os, hashlib
  import urllib2
   
class LoaderError(Exception): class LoaderError(Exception):
pass pass
   
  import tempfile
  def add_package_resource_cachedurl(ckan, package_name, url, name, format, license_id, size,**kwargs):
  if "xls" in url:
  format = "xls"
  if "pdf" in url:
  format = "pdf"
  if "xlsx" in url:
  format = "xlsx"
  (returned_url, mime_type, content) = scrape.fetchURL(scrape.docsdb,
  url, "dataset_resource", "AGIMO", False)
  if mime_type in ["application/vnd.ms-excel","application/msexcel","application/x-msexcel","application/x-ms-excel","application/x-excel","application/x-dos_ms_excel","application/xls","application/x-xls"]:
  format = "xls"
  if mime_type in ["application/xlsx","application/x-xlsx","application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"]:
  format = "xlsx"
   
  if content != None:
  tf = tempfile.NamedTemporaryFile(delete=False)
  tfName = os.path.abspath(tf.name)
  print tfName
  tf.seek(0)
  tf.write(content)
  tf.flush()
  ckan.add_package_resource (package_name, tfName, name=name, format=format, license_id=license_id)
  else:
  print "fetch error"
  return ckan.add_package_resource(package_name, url, name=name, resource_type='data',
  format=format,
  size=size, mimetype=mime_type, license_id=license_id)
   
# Instantiate the CKAN client. # Instantiate the CKAN client.
#ckan = ckanclient.CkanClient(base_location='http://localhost:5000/api', api_key='b47b24cd-591d-40c1-8677-d73101d56d1b') api_key = 'ff34526e-f794-4068-8235-fcbba38cd8bc'
api_key = 'b3ab75e4-afbb-465b-a09d-8171c8c69a7a' server = 'data.disclosurelo.gs'
ckan = ckanclient.CkanClient(base_location='http://data.disclosurelo.gs/api',  
  ckan = ckanclient.CkanClient(base_location='http://' + server + '/api',
api_key=api_key) api_key=api_key)
ckandirect = ckanapi.RemoteCKAN('http://data.disclosurelo.gs', api_key=api_key) ckandirect = ckanapi.RemoteCKAN('http://' + server, api_key=api_key)