/**
* Function to begin parsing the local copy of the pubchem substances directory
**/
function parse_substances()
{
$ignore = array(".", "..");
$input_dir = $this->getParameterValue('indir') . "/substances/";
$gz = false;
$this->CreateDirectory($this->getParameterValue('outdir') . "/substances/");
parent::setDatasetURI("bio2rdf_dataset:bio2rdf-" . $this->getPcsPrefix() . "-" . date("Ymd"));
$graph_uri = parent::getGraphURI();
//set graph URI to dataset uri
if (parent::getParameterValue('dataset_graph') == true) {
parent::setGraphURI(parent::getDatasetURI());
}
$dataset_description = '';
if ($handle = opendir($input_dir)) {
while (false !== ($file = readdir($handle))) {
if (in_array($file, $ignore)) {
continue;
}
echo "Processing file: " . $input_dir . $file . PHP_EOL;
$suffix = parent::getParameterValue('output_format');
$outfile = realpath($this->getParameterValue('outdir')) . "/substances/" . basename($file, ".xml.gz") . "." . $suffix;
if (strstr(parent::getParameterValue('output_format'), "gz")) {
$gz = true;
}
echo "... into " . $outfile . PHP_EOL;
parent::setCheckpoint('file');
$this->setWriteFile($outfile, $gz);
$this->parse_substance_file($input_dir, $file);
$this->getWriteFile()->close();
}
closedir($handle);
$source_file = (new DataResource($this))->setURI("http://www.ncbi.nlm.nih.gov/pcsubstance")->setTitle("PubChem Substance")->setRetrievedDate(date("Y-m-d\\TG:i:s\\Z", filemtime($input_dir)))->setFormat("text/xml")->setFormat("application/zip")->setPublisher("http://ncbi.nlm.nih.gov/")->setHomepage("http://pubchem.ncbi.nlm.nih.gov/")->setRights("use")->setRights("restricted-by-source-license")->setLicense("ftp://ftp.ncbi.nlm.nih.gov/pubchem/README")->setDataset("http://identifiers.org/pubchem.substance/");
$prefix = $this->getPcsPrefix();
$bVersion = parent::getParameterValue('bio2rdf_release');
$date = date("Y-m-d\\TG:i:s\\Z");
$output_file = (new DataResource($this))->setURI("http://download.bio2rdf.org/release/{$bVersion}/{$prefix}")->setTitle("Bio2RDF v{$bVersion} RDF version of {$prefix} (generated at {$date})")->setSource($source_file->getURI())->setCreator("https://github.com/bio2rdf/bio2rdf-scripts/blob/master/pubchem/pubchem.php")->setCreateDate($date)->setHomepage("http://download.bio2rdf.org/release/{$bVersion}/{$prefix}/{$prefix}.html")->setPublisher("http://bio2rdf.org")->setRights("use-share-modify")->setRights("by-attribution")->setRights("restricted-by-source-license")->setLicense("http://creativecommons.org/licenses/by/3.0/")->setDataset(parent::getDatasetURI());
if ($gz) {
$output_file->setFormat("application/gzip");
}
if (strstr(parent::getParameterValue('output_format'), "nt")) {
$output_file->setFormat("application/n-triples");
} else {
$output_file->setFormat("application/n-quads");
}
$dataset_description .= $source_file->toRDF() . $output_file->toRDF();
//set graph URI back to default
parent::setGraphURI($graph_uri);
// write the dataset description
$this->setWriteFile($this->getParameterValue('outdir') . "/substances/" . $this->getBio2RDFReleaseFile());
$this->getWriteFile()->write($dataset_description);
$this->getWriteFile()->close();
} else {
echo "unable to read directory contents: " . $input_dir . "\n";
exit;
}
}