public function getHtmlFiles($data, $project_id, $last)
{
$reference = new Reference();
$responce = [];
//массиы для хранения ошибок
$responce['errors'] = [];
//массив для хранения "успешных" урлов
$refs_handled = [];
//счетчик урлов
$ref_counts = [];
//максимальное количество запасных урлов, которые можно взять на замену
$max_extra_urls = 2;
foreach ($data as $query => $urls_array) {
$ref_count = 0;
// $max_ref_count = intval(( floor( ( count($urls_array) - 1 )/10.0 ) )*10);
// if( $max_ref_count < 10 ){ $max_ref_count = 10; }
if (!is_null($last)) {
$max_ref_count = $last[$query];
} else {
//вдруг last для этого запроса не установлен почему-то
$max_ref_count = 10;
}
foreach ($urls_array as $url) {
if ($ref_count < $max_ref_count) {
//получить html
$html_content = $reference->getHtml($url, $query, $ref_count, $project_id);
//если ошибка - берем запасной урл
if (strpos($html_content, "ERROR: Can not")) {
$responce['errors'][] = $html_content;
if ($max_extra_urls > 0) {
$max_ref_count++;
$max_extra_urls--;
}
} else {
$filename = $query . "_" . $ref_count . ".html";
//"очистка" html и запись в папку html_wo_garbage
$cleaned_content = $reference->removeGarbage($html_content, $project_id, $filename);
if (strpos($cleaned_content, "ERROR: Can not")) {
$responce['errors'][] = $cleaned_content;
//если ошибка - замена урла на запасной
if ($max_extra_urls > 0) {
$max_ref_count++;
$max_extra_urls--;
}
} else {
$responce[$query][] = $filename;
$refs_handled[$query][] = $url;
$ref_count += 1;
}
}
}
}
$ref_counts[$query] = $ref_count;
}
$storekeeper = new Storekeeper();
//запись в файл reports/Handled_referencies.txt тех урлов, которые реально будут обрабатываться
$storekeeper->storeReferencies($refs_handled, $project_id);
//запись в файл reports/errors.txt ошибок при получении и первичной обработке html-документов
$storekeeper->storeErrors($responce['errors'], $project_id);
unset($responce['errors']);
return [$responce, $ref_counts];
//the array of filenames with the query-text as the key
// return [$responce, $max_ref_count];//the array of filenames with the query-text as the key
}
public function handle_referencies(Request $request, Reference $reference, Lemmatizator $lemmatizator, Storekeeper $storekeeper)
{
$this->authorize('client-usage', $request->user());
$raw_data = $request->all();
// file_put_contents("/var/www/test.txt", print_r($raw_data,true));
// validates if there is any notEmpty queries in the set
// validates if the sample url is valid
// validates if project_id or new project name is present
if (isset($raw_data['ref'])) {
$raw_data['ref'] = $this->removeSecondScreenDuplicates($raw_data['ref']);
}
if (!empty($raw_data['ref'])) {
$raw_data['ref'] = $this->removeQueriesWithNoRefs($raw_data['ref']);
}
if (isset($raw_data['sample_url'])) {
$raw_data['sample_url'] = trim($raw_data['sample_url']);
}
$validator = Validator::make($raw_data, ['sample_url' => 'required_without:sample_html|url', 'sample_html' => 'required_without:sample_url|mimes:html,htm,htmls,odt,oth,ott,doc,dot,docx,dotx', 'ref' => 'required', 'query' => 'required_without:project_id', 'project_name' => 'required_without:project_id|max:20|alpha_num_dash_lat', 'project_id' => 'required_without:project_name']);
//validates if mentioned only one of url or local file for sample
$validator->after(function ($validator) {
if (Input::get('sample_url') != '' && Input::hasFile('sample_html')) {
$validator->errors()->add('sample_url', 'Choose smth one between remote url or local file');
}
});
//make $inputed_responce using ref_dirty data to fill form fields with already edited by user data if validation fails
if ($validator->fails()) {
$id = isset($raw_data['project_id']) ? $raw_data['project_id'] : null;
$inputed_responce = [];
foreach (Input::get('ref_dirty') as $query => $dirty_refs) {
if ($dirty_refs) {
$query = $query == '0' ? "" : $query;
$inputed_responce[$query] = $dirty_refs;
} else {
$inputed_responce = [];
break;
}
}
return redirect()->action('WebsiteController@referencies_handler', [$id])->withErrors($validator)->withInput(Input::except('ref', 'query', 'ref_dirty'))->with('inputed_responce', $inputed_responce);
}
if (isset($raw_data['last'])) {
$last = $raw_data['last'];
} else {
$last = null;
}
//get the project_id from INPUT or set the new one
if (isset($raw_data['project_id'])) {
$project_id = strip_tags($raw_data['project_id']);
} else {
// mkdir empty dirs which are necessary for query handling
// initialize project
$new_project = $storekeeper->initializeStoring($raw_data['project_name'], $raw_data['query']);
$project_id = $new_project->id;
}
//set exact flag to the project if it is checked
if (isset($raw_data['exact'])) {
$this_project = Project::find($project_id);
$this_project->exact = true;
$this_project->save();
}
//store additional queries to file at project dir
if (isset($raw_data['dop_queries'])) {
$raw_query['dop_queries'] = strip_tags($raw_data['dop_queries']);
$dop_queries_array = preg_split("/[,;\\.]*[\t\n\f\r]+\\s*/", trim($raw_data['dop_queries']), -1, PREG_SPLIT_NO_EMPTY);
} else {
$dop_queries_array = [];
}
//check if dop_queries duplicate main queries
foreach ($dop_queries_array as $num => $dop_query) {
$dop_query = $this->clean_query($dop_query);
if (in_array($dop_query, array_keys($raw_data['ref']))) {
unset($dop_queries_array[$num]);
}
}
if (!empty($dop_queries_array)) {
foreach ($dop_queries_array as $key => $raw_query) {
$dop_query = trim(array_keys($lemmatizator->splitToWords($raw_query))[0]);
$dop_queries_array[$key] = $dop_query;
}
$dop_queries = implode("\n", $dop_queries_array);
$project_dir = Project::where('id', $project_id)->first()->dir;
file_put_contents(storage_path("tmp/" . $project_dir . "/reports/dop_queries.txt"), $dop_queries);
}
// obtain an array of referencies instead of text from textarea
// text is splitted against different white-spaces and the commas
foreach ($raw_data['ref'] as $ref => $urls_text) {
//splitToWords ЗДЕСЬ используется только для "очистки" запроса
//меняется е на ё, убираются стоп-слова и т.д.
//splitToWords возвращает массив, в котором ключами являются тексты запросов, а значениями массивы составляющих их слов
//используем splitToWords для того, чтобы "очистка" была одинаковой на всех этапах обработки
$query = array_keys($lemmatizator->splitToWords($ref))[0];
//заменяем запрос на очищенный в переменной last
if ($query != $ref && !is_null($last)) {
$last[$query] = $last[$ref];
unset($last[$ref]);
}
//make array of urls instead of text
$data[$query] = preg_split("/[\\s,]+/", $urls_text);
}
$responce = [];
$refs_handled = [];
foreach ($data as $query => $urls_array) {
foreach ($urls_array as $key => $url) {
//remove empty strings if any
if ($url == "") {
unset($data[$query][$key]);
}
}
}
if ($raw_data['sample_url'] || Input::hasFile('sample_html')) {
if ($raw_data['sample_url']) {
$src_type = 'http';
$sample_url = Input::get('sample_url');
} else {
$src_type = 'local';
$sample_url = Input::file('sample_html')->getRealPath();
}
//get html of sample from file or from www
try {
$sample_content = $reference->getHtml($sample_url, 'sample', 0, $project_id, $src_type);
$sample_cleaned_content = $reference->removeGarbage($sample_content, $project_id, "sample_0.html");
} catch (\Exception $e) {
$mes = $e->getMessage();
$inputed_responce = [];
foreach (Input::get('ref') as $query => $refs) {
$query = $query == '0' ? "" : $query;
$inputed_responce[$query] = $refs;
}
return redirect()->action('WebsiteController@referencies_handler', [$project_id])->with('alert', "Cannot fetch html from the sample url: " . $sample_url . "\n" . $mes)->withInput(Input::except('ref', 'query'))->with('inputed_responce', $inputed_responce);
}
} else {
$sample_file_name = 'sample_0.html';
$sample_cleaned_content = '';
}
// return '<pre>'.print_r($last,true).'</pre>';
$this->dispatch(new HandleDataJob($sample_cleaned_content, $data, $project_id, $last));
return redirect()->action('UserController@show_user', [$request->user()->id])->with('info', "Your Project is in process. Gather report when it'll appear in reports column.");
}
