From: Maxious Date: Sat, 28 Jan 2012 06:59:01 +0000 Subject: Scrape required and chart of complied features views X-Git-Url: https://maxious.lambdacomplex.org/git/?p=disclosr.git&a=commitdiff&h=3732ed2e5b49d875223990d6390ef382c5e27764 --- Scrape required and chart of complied features views Former-commit-id: 1f70b43713b7686e9f7a8a0f6a5aced655d53221 --- --- /dev/null +++ b/admin/refreshDesignDoc.php @@ -1,1 +1,7 @@ +get_db('disclosr-agencies'); +createAgencyDesignDoc(); +?> + --- a/alaveteli/exportAgencies.csv.php +++ b/alaveteli/exportAgencies.csv.php @@ -1,6 +1,21 @@ get_view("app", "byDeptStateName", null, true)->rows; //print_r($rows); foreach ($rows as $row) { - $tag[$row->value] = phrase_to_tag(dept_to_portfolio($row->key)); + $tag[$row->id] = phrase_to_tag(dept_to_portfolio($row->key)); } } catch (SetteeRestClientException $e) { setteErrorHandler($e); + die(); +} + +$foiEmail = Array(); +try { + $rows = $db->get_view("app", "foiEmails", null, true)->rows; + //print_r($rows); + foreach ($rows as $row) { + $foiEmail[$row->key] = $row->value; + } +} catch (SetteeRestClientException $e) { + setteErrorHandler($e); + die(); } $fp = fopen('php://output', 'w'); if ($fp && $db) { - header('Content-Type: text/csv'); + header('Content-Type: text/csv; charset=utf-8'); header('Content-Disposition: attachment; filename="export.' . date("c") . '.csv"'); header('Pragma: no-cache'); header('Expires: 0'); @@ -29,39 +57,43 @@ $agencies = $db->get_view("app", "byCanonicalName", null, true)->rows; //print_r($rows); foreach ($agencies as $agency) { - // print_r($agency); - $row = Array(); - $row["#id"] = $agency->value->_id; - $row["name"] = $agency->value->name; - if (isset($agency->value->foi_email)) { - $row["request_email"] = $agency->value->foi_email; - } else { - $row["request_email"] = "foi@".GetDomain($agency->value->website); - // - } - if (isset($agency->value->short_name)) { - $row["short_name"] = $agency->value->short_name; - } else { - $out = Array(); - preg_match_all('/[A-Z]/', $agency->value->name, $out); - $row["short_name"] = implode("",$out[0]); - } - $row["notes"] = ""; - $row["publication_scheme"] = $agency->value->infoPublicationSchemeURL; - $row["home_page"] = $agency->value->website; - if ($agency->value->orgType == "FMA-DepartmentOfState") { - $row["tag_string"] = $tag[$agency->value->_id]; - } else { - $row["tag_string"] = $tag[$agency->value->parentOrg];; - } - - fputcsv($fp, array_values($row)); - - if (isset($agency->value->foiBodies)) { - foreach ($agency->value->foiBodies as $foiBody) { - $row['name'] = $foiBody; - $row['short_name'] = ""; - fputcsv($fp, array_values($row)); + // print_r($agency); + + if (isset($agency->value->foiEmail) && $agency->value->foiEmail != "null" && !isset($agency->value->status)) { + $row = Array(); + $row["#id"] = $agency->id; + $row["name"] = trim($agency->value->name); + if (isset($agency->value->foiEmail)) { + $row["request_email"] = $agency->value->foiEmail; + } else { + if ($agency->value->orgType == "FMA-DepartmentOfState") { + $row["request_email"] = "foi@" . GetDomain($agency->value->website); + } else { + $row["request_email"] = $foiEmail[$agency->value->parentOrg]; + } + } + if (isset($agency->value->shortName)) { + $row["short_name"] = $agency->value->shortName; + } else { + $row["short_name"] = shortName($agency->value->name); + } + $row["notes"] = ""; + $row["publication_scheme"] = (isset($agency->value->infoPublicationSchemeURL) ? $agency->value->infoPublicationSchemeURL : ""); + $row["home_page"] = (isset($agency->value->website) ? $agency->value->website : ""); + if ($agency->value->orgType == "FMA-DepartmentOfState") { + $row["tag_string"] = $tag[$agency->value->_id] . " " . $agency->value->orgType; + } else { + $row["tag_string"] = $tag[$agency->value->parentOrg] . " " . $agency->value->orgType; + } + + fputcsv($fp, array_values($row)); + + if (isset($agency->value->foiBodies)) { + foreach ($agency->value->foiBodies as $foiBody) { + $row['name'] = iconv("UTF-8", "ASCII//TRANSLIT",$foiBody); + $row["short_name"] = shortName($foiBody); + fputcsv($fp, array_values($row)); + } } } } --- a/alaveteli/exportCategories.rb.php +++ b/alaveteli/exportCategories.rb.php @@ -1,19 +1,20 @@ get_db('disclosr-agencies'); try { $rows = $db->get_view("app", "byDeptStateName", null, true)->rows; //print_r($rows); foreach ($rows as $row) { - echo ' [ "'.phrase_to_tag(dept_to_portfolio($row->key)).'","'. dept_to_portfolio($row->key).'","'.dept_to_portfolio($row->key).'" ],'.PHP_EOL; + echo ' [ "' . phrase_to_tag(dept_to_portfolio($row->key)) . '","' . dept_to_portfolio($row->key) . '","part of the ' . dept_to_portfolio($row->key) . ' portfolio" ],' . PHP_EOL; } } catch (SetteeRestClientException $e) { setteErrorHandler($e); --- a/getAgency.php +++ b/getAgency.php @@ -4,17 +4,36 @@ include_header(); function displayValue($key, $value, $mode) { - global $db; + global $db, $schemas; if ($mode == "view") { + + echo ""; + + echo "" . $schemas['agency']["properties"][$key]['x-title'] . "
" . $schemas['agency']["properties"][$key]['description'] . ""; if (is_array($value)) { - echo "$key
    "; + echo "
      "; foreach ($value as $subkey => $subvalue) { - echo "
    1. $subvalue
      2. "; + if (isset($schemas['agency']["properties"][$key]['x-itemprop'])) { + echo '
    2. '; + } else { + echo "
    3. "; + } + echo "$subvalue
      4. "; } echo "
    "; } else { - echo "$key$value"; + if (isset($schemas['agency']["properties"][$key]['x-itemprop'])) { + echo ''; + } else { + echo ""; + } + if ((strpos($key, "URL") > 0 || $key == 'website') && $value != "") { + echo "view"; + } else { + echo "$value"; + } } + echo ""; } if ($mode == "edit") { if (is_array($value)) { @@ -32,22 +51,22 @@ if (strpos($key, "_") === 0) { echo""; } else if ($key == "parentOrg") { - echo ""; + echo ""; } else if (strpos($key, "has") === 0) { - echo ""; + echo ""; } else { echo ""; - if ((strpos($key,"URL") > 0 || $key == 'website')&& $value != "") { - echo "view"; + if ((strpos($key, "URL") > 0 || $key == 'website') && $value != "") { + echo "view"; } if ($key == 'abn') { - echo "view abn"; + echo "view abn"; } } } @@ -62,16 +81,14 @@ if (!isset($row[$defaultField])) { if ($schemas['agency']['properties'][$defaultField]['type'] == "string") { if (strpos($defaultField, "has") === 0) { - $row[$defaultField] = "false"; - + $row[$defaultField] = "false"; } else { - $row[$defaultField] = ""; + $row[$defaultField] = ""; } - } - if ($schemas['agency']['properties'][$defaultField]['type'] == "array") { - - $row[$defaultField] = Array(""); + if ($schemas['agency']['properties'][$defaultField]['type'] == "array") { + + $row[$defaultField] = Array(""); } } } @@ -89,20 +106,33 @@ //print_r($row); if (sizeof($_POST) > 0) { //print_r($_POST); + foreach ($_POST as $postkey => $postvalue) { + if ($postvalue == "") { + unset($_POST[$postkey]); + } + if (is_array($postvalue) && count($postvalue) == 1 && $postvalue[0] == "") { + unset($_POST[$postkey]); + } + } if (isset($_POST['_id']) && $db->get_rev($_POST['_id']) == $_POST['_rev']) { echo "Edited version was latest version, continue saving"; $newdoc = $_POST; $newdoc['metadata']['lastModified'] = time(); $row = $db->save($newdoc); } else { - echo "ALERT doc revised by someone else while editing."; + echo "ALERT doc revised by someone else while editing. Document not saved."; } } - $mode = "edit"; - $row = addDefaultFields(object_to_array($row)); + $mode = "view"; + if ($mode == "edit") { + $row = addDefaultFields(object_to_array($row)); + } else { + $row = object_to_array($row); + } + if ($mode == "view") { - echo ''; + echo '
    '; echo '"; echo ""; } @@ -127,40 +157,40 @@ }; - $value) { + echo displayValue($key, $value, $mode); + } + if ($mode == "view") { + echo "

    ' . $row['name'] . "

    Field NameField Value
    "; + } + if ($mode == "edit") { + echo ''; + } +} else { + + try { + /* $rows = $db->get_view("app", "showNamesABNs")->rows; + //print_r($rows); + foreach ($rows as $row) { + // print_r($row); + echo '
  1. ' . + (isset($row->value->name) && $row->value->name != "" ? $row->value->name : "NO NAME " . $row->value->abn) + . '
    2. '; + } */ + $rows = $db->get_view("app", "byName")->rows; + //print_r($rows); + foreach ($rows as $row) { + // print_r($row); + echo '
    3. '; } - foreach ($row as $key => $value) { - echo displayValue($key, $value, $mode); - } - if ($mode == "view") { - echo ""; - } - if ($mode == "edit") { - echo ''; - } - } else { - - try { - /*$rows = $db->get_view("app", "showNamesABNs")->rows; - //print_r($rows); - foreach ($rows as $row) { - // print_r($row); - echo '
  3. ' . - (isset($row->value->name) && $row->value->name != "" ? $row->value->name : "NO NAME " . $row->value->abn) - . '
    4. '; - }*/ - $rows = $db->get_view("app", "byName")->rows; - //print_r($rows); - foreach ($rows as $row) { - // print_r($row); - echo '
  4. ' . - $row->key - . '
    5. '; - } - } catch (SetteeRestClientException $e) { - setteErrorHandler($e); - } + } catch (SetteeRestClientException $e) { + setteErrorHandler($e); } - include_footer(); - ?> +} +include_footer(); +?> --- a/graph.php +++ b/graph.php @@ -1,21 +1,51 @@ - - - - + + + + }; + - - + --- a/include/common.inc.php +++ b/include/common.inc.php @@ -47,7 +47,7 @@ return trim(str_replace("Department of", "", str_replace("Department of the", "Department of", $deptName))); } function phrase_to_tag ($phrase) { - return str_replace(" ","_",str_replace(",","",strtolower($phrase))); + return str_replace(" ","_",str_replace("'","",str_replace(",","",strtolower($phrase)))); } function GetDomain($url) { @@ -62,7 +62,3 @@ } } -?> - - - --- a/include/couchdb.inc.php +++ b/include/couchdb.inc.php @@ -1,8 +1,8 @@ language = "javascript"; $obj->views->all->map = "function(doc) { emit(doc._id, 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') { - emit(doc.name, doc._id); + emit(doc.name, doc); } };"; - $obj->views->byDeptStateName->map = "function(doc) { + $obj->views->byDeptStateName->map = "function(doc) { if (doc.orgType == 'FMA-DepartmentOfState') { emit(doc.name, doc._id); } };"; - $obj->views->parentOrgs->map = "function(doc) { + $obj->views->parentOrgs->map = "function(doc) { if (doc.parentOrg) { emit(doc._id, doc.parentOrg); } @@ -34,10 +34,25 @@ } } };"; + + $obj->views->foiEmails->map = "function(doc) { + emit(doc._id, doc.foiEmail); +};"; + $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->getSuspended->map = 'function(doc) { if (doc.status == "suspended") { emit(doc._id, doc); } };'; - $obj->views->getScrapeRequired->map = "function(doc) { emit(doc.abn, doc); };"; + $obj->views->getScrapeRequired->map = "function(doc) { + +var lastScrape = Date.parse(doc.metadata.lastScraped); + +var today = new Date(); + +if (!lastScrape || lastScrape.getTime() + 1000 != today.getTime()) { + emit(doc._id, doc); +} + +};"; $obj->views->showNamesABNs->map = "function(doc) { emit(doc._id, {name: doc.name, abn: doc.abn}); };"; $obj->views->getConflicts->map = "function(doc) { if (doc._conflicts) { @@ -45,6 +60,26 @@ } }"; // http://stackoverflow.com/questions/646628/javascript-startswith + $obj->views->scoreHas->map = 'if(!String.prototype.startsWith){ + String.prototype.startsWith = function (str) { + return !this.indexOf(str); + } +} +if(!String.prototype.endsWith){ + String.prototype.endsWith = function(suffix) { +     return this.indexOf(suffix, this.length - suffix.length) !== -1; + }; +} +function(doc) { +if (typeof(doc["status"]) == "undefined" || doc["status"] != "suspended") { +for(var propName in doc) { + if(typeof(doc[propName]) != "undefined" && (propName.startsWith("has") || propName.endsWith("URL"))) { + emit(propName, 1); + } +} + emit("total", 1); + } +}'; $obj->views->score->map = 'if(!String.prototype.startsWith){ String.prototype.startsWith = function (str) { return !this.indexOf(str); @@ -67,19 +102,17 @@ return $db->save($obj, true); } +if (php_uname('n') == "vanille") { -if( php_uname('n') == "vanille") { + $server = new SetteeServer('http://192.168.178.21:5984'); +} else +if (php_uname('n') == "KYUUBEY") { -$server = new SetteeServer('http://192.168.178.21:5984'); -} else - if( php_uname('n') == "KYUUBEY") { - -$server = new SetteeServer('http://192.168.1.148:5984'); + $server = new SetteeServer('http://192.168.1.148:5984'); } else { $server = new SetteeServer('http://127.0.0.1:5984'); } + function setteErrorHandler($e) { echo $e->getMessage() . "
    " . PHP_EOL; } -?> - --- a/include/template.inc.php +++ b/include/template.inc.php @@ -19,11 +19,11 @@ Disclosr - - + + @@ -44,7 +44,7 @@
    Agencies - About/FAQ + About/FAQ
    @@ -55,7 +55,10 @@
    + function include_footer() { + global $basePath; + ?> +
    @@ -63,14 +66,11 @@ - - + + - - + "Representation of government agency and online transparency measures", "type" => "object", "properties" => Array( - "name" => Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "short_name" => Array("type" => "string", "required" => false, "x-title"=> "Agency Short Name", "description" => "Agency Name shortened, usually to an acronym"), - "otherNames" => Array("type" => "array", "required" => true, "x-title" => "Agency Past/Other Names", "description" => "Agency Names", + "name" => Array("type" => "string", "required" => true, "x-itemprop" => "name", "x-title" => "Name", "description" => "Name, most recent and broadest"), + "shortName" => Array("type" => "string", "required" => false, "x-title" => "Short Name", "description" => "Name shortened, usually to an acronym"), + "foiEmail" => Array("type" => "string", "required" => false, "x-title" => "FOI Contact Email", "description" => "FOI contact email if not foi@"), + "sameAs" => Array("type" => "array", "required" => false, "x-itemprop"=>"http://www.w3.org/2002/07/owl#sameAs","x-title" => "Same As", "description" => "Same as other URLs/URIs for this entity", "items" => Array("type" => "string")), - "foiBodies" => Array("type" => "array", "required" => true, "x-title" => "FOI Bodies", "description" => "Organisational units within this agency that are subject to FOI Act but are not autonomous", + "otherNames" => Array("type" => "array", "required" => true, "x-title" => "Past/Other Names", "description" => "Other names for organisation", "items" => Array("type" => "string")), - "orgType"=> Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "parentOrg"=> Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "website"=> Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "abn" => Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "contractListURL" => Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "grantsReportingURL" => Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "annualReportURL" => Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "consultanciesURL" => Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "legalExpenditureURL"=> Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "recordsListURL"=> Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "FOIDocumentsURL"=> Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "infoPublicationSchemeURL"=> Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "appointmentsURL"=> Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "hasRSS"=> Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "hasMailingList"=> Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "hasTwitter"=> Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "hasFacebook"=> Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - "hasYouTube"=> Array("type" => "string", "required" => true, "x-title"=> "Agency Name", "description" => "Agency Name, most recent and broadest"), - + "foiBodies" => Array("type" => "array", "required" => true, "x-title" => "FOI Bodies","x-itemprop"=>"members", "description" => "Organisational units within this agency that are subject to FOI Act but are not autonomous", + "items" => Array("type" => "string")), + "orgType" => Array("type" => "string", "required" => true, "x-title" => "Organisation Type", "description" => "Org type based on legal formation via FMA/CAC legislation etc."), + "parentOrg" => Array("type" => "string", "required" => true, "x-title" => "Parent Organisation", "description" => "Parent organisation, usually a department of state"), + "website" => Array("type" => "string", "required" => true, "x-title" => "Website", "x-itemprop" => "url", "description" => "Website URL"), + "abn" => Array("type" => "string", "required" => true, "x-title" => "Australian Business Number", "description" => "ABN from business register"), + "contractListURL" => Array("type" => "string", "required" => true, "x-title" => "Contract Listing", "description" => "Departmental and agency contracts, mandated by the Senate"), + "grantsReportingURL" => Array("type" => "string", "required" => true, "x-title" => "Grants Awarded", + "description" => "Departmental and agency grants mandated by the Senate and Commonwealth grants guidelines "), + "annualReportURL" => Array("type" => "string", "required" => true, "x-title" => "Annual Report(s)", "description" => ""), + "consultanciesURL" => Array("type" => "string", "required" => true, "x-title" => "Consultants Hired", "description" => ""), + "legalExpenditureURL" => Array("type" => "string", "required" => true, "x-title" => "Legal Services Expenditure", "description" => "Legal Services Expenditure mandated by Legal Services Directions 2005"), + "recordsListURL" => Array("type" => "string", "required" => true, "x-title" => "Files/Records Held", "description" => "Indexed lists of departmental and agency files, mandated by the Senate"), + "FOIDocumentsURL" => Array("type" => "string", "required" => true, "x-title" => "FOI Documents Released", "description" => ""), + "infoPublicationSchemeURL" => Array("type" => "string", "required" => true, "x-title" => "Information Publication Scheme", "description" => ""), + "appointmentsURL" => Array("type" => "string", "required" => true, "x-title" => "Agency Appointments/Boards", "description" => "Departmental and agency appointments and vacancies , mandated by the Senate"), + "advertisingURL" => Array("type" => "string", "required" => true, "x-title" => "Approved Advertising Campaigns", "description" => " Agency advertising and public information projects, mandated by the Senate "), + "hasRSS" => Array("type" => "string", "required" => true, "x-title" => "Has RSS", "description" => ""), + "hasMailingList" => Array("type" => "string", "required" => true, "x-title" => "Has Mailing List", "description" => ""), + "hasTwitter" => Array("type" => "string", "required" => true, "x-title" => "Has Twitter", "description" => ""), + "hasFacebook" => Array("type" => "string", "required" => true, "x-title" => "Has Facebook", "description" => ""), + "hasYouTube" => Array("type" => "string", "required" => true, "x-title" => "Has YouTube", "description" => ""), + "hasFlickr" => Array("type" => "string", "required" => true, "x-title" => "Has Flickr", "description" => ""), + "hasCCBY" => Array("type" => "string", "required" => true, "x-title" => "Has CC-BY", "description" => "Has any page licenced Creative Commons - Attribution"), ), - /*"org":{"type":"object", - "properties":{ - "organizationName":{"type":"string"}, - "organizationUnit":{"type":"string"}}, - } - }*/ + /* "org":{"type":"object", + "properties":{ + "organizationName":{"type":"string"}, + "organizationUnit":{"type":"string"}}, + } + } */ ); ?> --- /dev/null +++ b/scrape.py @@ -1,1 +1,76 @@ +#http://packages.python.org/CouchDB/client.html +import couchdb +import urllib2 +from BeautifulSoup import BeautifulSoup +import re +#http://diveintopython.org/http_web_services/etags.html +class NotModifiedHandler(urllib2.BaseHandler): + def http_error_304(self, req, fp, code, message, headers): + addinfourl = urllib2.addinfourl(fp, headers, req.get_full_url()) + addinfourl.code = code + return addinfourl + +def scrapeAndStore(URL, depth, agency): + URL = "http://www.google.com" + req = urllib2.Request(URL) + etag = 'y' + last_modified = 'y' + #if there is a previous version sotred in couchdb, load caching helper tags + if etag: + req.add_header("If-None-Match", etag) + if last_modified: + req.add_header("If-Modified-Since", last_modified) + + opener = urllib2.build_opener(NotModifiedHandler()) + url_handle = opener.open(req) + headers = url_handle.info() # the addinfourls have the .info() too + etag = headers.getheader("ETag") + last_modified = headers.getheader("Last-Modified") + web_server = headers.getheader("Server") + file_size = headers.getheader("Content-Length") + mime_type = headers.getheader("Content-Type") + + if hasattr(url_handle, 'code'): + if url_handle.code == 304: + print "the web page has not been modified" + else: + #do scraping + html = url_handle.read() + # http://www.crummy.com/software/BeautifulSoup/documentation.html + soup = BeautifulSoup(html) + links = soup.findAll('a') # soup.findAll('a', id=re.compile("^p-")) + for link in links: + print link['href'] + #for each unique link + #if html mimetype + # go down X levels, + # diff with last stored attachment, store in document + #if not + # remember to save parentURL and title (link text that lead to document) + + #store as attachment epoch-filename + else: + print "error %s in downloading %s", url_handle.code, URL + #record/alert error to error database + + + + + + + + + + + +couch = couchdb.Server('http://192.168.1.148:5984/') + +# select database +agencydb = couch['disclosr-agencies'] + +for row in agencydb.view('app/getScrapeRequired'): #not recently scraped agencies view? + agency = agencydb.get(row.id) + print agency['name'] +scrapeAndStore("A",1,1) + --- a/unimplemented/exportAgencies.csv.php +++ /dev/null @@ -1,65 +1,1 @@ -prepare('select * from "UNSPSCcategories" where "UNSPSC"::text like \'%00000\';'); -$unspscresult->execute(); -foreach ($unspscresult->fetchAll() as $row) { - $unspsc[$row['UNSPSC']] = $row['Title']; -} - -$query = $conn->prepare(' -SELECT "CNID",contractnotice."agencyName",agency_nametoabn.abn as "agencyABN", -EXTRACT(EPOCH FROM "publishDate") as "publishDate", -EXTRACT(EPOCH FROM "contractStart") as "contractStart", -EXTRACT(EPOCH FROM "contractEnd") as "contractEnd", -value,description,category, -"supplierName",(case when "supplierABN" != 0 THEN "supplierABN"::text ELSE "supplierName" END) as supplierID, -(\'https://www.tenders.gov.au/?event=public.advancedsearch.keyword&keyword=CN\'::text || "CNID"::text) as sourceURL -FROM contractnotice join agency_nametoabn on contractnotice."agencyName"=agency_nametoabn."agencyName" -where "childCN" is null' - , array(PDO::ATTR_CURSOR => PDO::FETCH_ORI_NEXT)); -$query->execute(); -$errors = $conn->errorInfo(); -if ($errors[2] != "") { - die("Export terminated, db error" . print_r($errors, true)); -} - -$num_fields = $query->columnCount(); -$headers = Array(); -for ($i = 0; $i < $num_fields; $i++) { // for each column in query, make a CSV header - $meta = $query->getColumnMeta($i); - $headers[] = $meta['name']; -} -$fp = fopen('php://output', 'w'); -if ($fp && $query) { - header('Content-Type: text/csv'); - header('Content-Disposition: attachment; filename="export.' . date("c") . '.csv"'); - header('Pragma: no-cache'); - header('Expires: 0'); - fputcsv($fp, $headers); - while ($row = $query->fetch(PDO::FETCH_NUM, PDO::FETCH_ORI_NEXT)) { - foreach ($row as $key => &$colvalue) { - - $colvalue = preg_replace('/[^[:print:]]/', '', utf8_encode($colvalue)); - if ($headers[$key] == "publishDate" || $headers[$key] == "contractStart" - || $headers[$key] == "contractEnd") { - $colvalue = date("Y-m-d", $colvalue); - } - /* if ($headers[$key] == "CNID") { - $colvalue = str_replace("A","", $colvalue); -}*/ - if ($headers[$key] == "cat1" || $headers[$key] == "cat2" - || $headers[$key] == "cat3") { - $colvalue = $unspsc[$colvalue]; - } - } - fputcsv($fp, array_values($row)); - } - die; -} -?> - --- a/unimplemented/scrape.py +++ /dev/null @@ -1,64 +1,1 @@ -#http://packages.python.org/CouchDB/client.html -import couchdb -import urllib2 -from BeautifulSoup import BeautifulSoup -import re -couch = couchdb.Server() # Assuming localhost:5984 -# If your CouchDB server is running elsewhere, set it up like this: -# couch = couchdb.Server('http://example.com:5984/') - -# select database -agencydb = couch['disclosr-agencies'] - -for row in agencydb.view('app/getScrapeRequired'): #not recently scraped agencies view? - agency = agencydb.get(row.id) - print agency['agencyName'] - -#http://diveintopython.org/http_web_services/etags.html -class NotModifiedHandler(urllib2.BaseHandler): - def http_error_304(self, req, fp, code, message, headers): - addinfourl = urllib2.addinfourl(fp, headers, req.get_full_url()) - addinfourl.code = code - return addinfourl - -def scrapeAndStore(URL, depth, agency): - URL = "http://www.hole.fi/jajvirta/weblog/" - req = urllib2.Request(URL) - - #if there is a previous version sotred in couchdb, load caching helper tags - if etag: - req.add_header("If-None-Match", etag) - if last_modified: - req.add_header("If-Modified-Since", last_modified) - - opener = urllib2.build_opener(NotModifiedHandler()) - url_handle = opener.open(req) - headers = url_handle.info() # the addinfourls have the .info() too - etag = headers.getheader("ETag") - last_modified = headers.getheader("Last-Modified") - web_server = headers.getheader("Server") - file_size = headers.getheader("Content-Length") - mime_type = headers.getheader("Content-Type") - - if hasattr(url_handle, 'code') and url_handle.code == 304: - print "the web page has not been modified" - else: - print "error %s in downloading %s", url_handle.code, URL - #record/alert error to error database - - #do scraping - html = ? - # http://www.crummy.com/software/BeautifulSoup/documentation.html - soup = BeautifulSoup(html) -links = soup.findAll('a') # soup.findAll('a', id=re.compile("^p-")) -for link in links: - print link['href'] - #for each unique link - #if html mimetype - # go down X levels, - # diff with last stored attachment, store in document - #if not - # remember to save parentURL and title (link text that lead to document) - - #store as attachment epoch-filename