我正在尝试使用laravel队列将数据从CSV文件导入数据库。这些CSV文件很大,大约有50万行。
我在某处了解到,使用laravel队列我们不需要考虑连接超时,但这看起来并不正确。也许我错了。
请检查我的工作代码,如果这些方法有任何问题。我正在使用“ League \ Csv”读取CSV文件。
public function __construct($data,$error_arr,$error_row_numbers) {
$this->data = $data;
$this->error_arr = $error_arr;
$this->error_row_numbers = $error_row_numbers;
}
/**
* Execute the job.
*
* @return void
*/
public function handle()
{
$offset = $this->data['offset'];
$limit = $this->data['limit'];
$filename = $this->data['file_name'];
$service = new Service();
$table = 'committees';
$dbase = new Committee();
//map_data have array about which csv column
//should be inserted in which column of database
$map_data = $this->data['map_data'];
//get all columns name of a table
$db_header_obj = new Committee();
$db_header = $db_header_obj->getTableColumns();
$csv_file_path = storage_path('app/files/committee/').$filename;
if (!ini_get("auto_detect_line_endings")) {
ini_set("auto_detect_line_endings", TRUE);
}
$csv = Reader::createFromPath($csv_file_path, 'r');
$csv->setOutputBOM(Reader::BOM_UTF8);
$csv->addStreamFilter('convert.iconv.ISO-8859-15/UTF-8');
$csv->setHeaderOffset(0);
$csv_header = $csv->getHeader();
$rec_arr = array();
$records = array();
$records_arr = array();
$stmt = (new Statement())
->offset($offset)
->limit($limit)
;
$records = $stmt->process($csv);
foreach ($records as $record)
{
$rec_arr[] = array_values($record);
}
//trim index if the value of an array is empty
$records_arr = $service->trimArray($rec_arr);
if(count($records_arr)>0)
{
foreach($records_arr as $ck => $cv){
$committee_arr = array();
foreach ($map_data as $mk => $mv) {
if(isset($mv)){
$data_type = $service->getDatabaseColumnType($table,$mv);
//if data is one of datetime data type
//then format the csv data to mysql datetime format
if($data_type == 'date' || $data_type == 'datetime' || $data_type == 'timestamp'){
$datetime = (array)$cv[$mk];
$dt = array_shift($datetime);
$dt = date('Y-m-d h:i:s', strtotime($dt));
$committee_arr[$mv] = $dt;
}else{
$committee_arr[$mv] = $cv[$mk];
}
}
}
$error_encountered = false;
DB::beginTransaction();
if(!empty($committee_arr['com_id'])){
try{
$committee_row = Committee::updateOrCreate(
['com_id' => $committee_arr['com_id']],
$committee_arr
);
if ($committee_row->wasRecentlyCreated === true) {
$committee_row->created_by = $this->data['user_id'];
}else{
$committee_row->updated_by = $this->data['user_id'];
}
$committee_row->save();
} catch (\Exception $e) {
$error_encountered = true;
$this->error_arr[] = $e->getMessage();
$this->error_row_numbers[] = $this->data['row_value'];
}
}
DB::commit();
//just to keep track which row is currently processing
//so that user can be notified in which row of csv
//there is an error
$this->data['row_value'] = $this->data['row_value'] + 1;
}
//offset just to start fectch next chunk of data from csv
$this->data['offset'] = $offset + $limit;
//Call to same job but with increased offset value
$committeeInsertJob = (new StoreCommittee($this->data,$this->error_arr,$this->error_row_numbers))->delay(Carbon::now()->addSeconds(3));
dispatch($committeeInsertJob);
}else{
//Store activity just to keep track of activity
$activity = new Activity();
$activity->url = $this->data['url'];
$activity->action = 'store';
$activity->description = $table;
$activity->user_id = $this->data['user_id'];
$activity->created_at = date('Y-m-d H:i:s');
$activity->save();
$arr_data = [
'filename' => $filename,
'user_name' => $this->data['user_name'],
'error' => $this->error_arr,
'error_row_numbers' => $this->error_row_numbers
];
//Notify user that the job is complete
Mail::to($this->data['user_email'])->send(new CSVImportJobComplete($arr_data));
}
if (!ini_get("auto_detect_line_endings")) {
ini_set("auto_detect_line_endings", FALSE);
}
}
错误:来自(存储内部的laravel.log)
[2019-04-05 07:13:23] local.ERROR: PDOStatement::execute(): MySQL server has gone away (SQL: insert into `jobs` (`queue`, `attempts`, `reserved_at`, `available_at`, `created_at`, `payload`) values (default, 0, , 1554448406, 1554448403, ....................................................(long list)
来自:命令终端
$ php artisan queue:work --tries=3
[2019-04-05 07:09:11][1] Processing: App\Jobs\StoreCommittee
[2019-04-05 07:09:33][1] Processed: App\Jobs\StoreCommittee
[2019-04-05 07:09:36][2] Processing: App\Jobs\StoreCommittee
[2019-04-05 07:09:58][2] Processed: App\Jobs\StoreCommittee
[2019-04-05 07:10:01][3] Processing: App\Jobs\StoreCommittee
[2019-04-05 07:10:23][3] Processed: App\Jobs\StoreCommittee
[2019-04-05 07:10:26][4] Processing: App\Jobs\StoreCommittee
[2019-04-05 07:10:48][4] Processed: App\Jobs\StoreCommittee
[2019-04-05 07:10:51][5] Processing: App\Jobs\StoreCommittee
[2019-04-05 07:11:13][5] Processed: App\Jobs\StoreCommittee
[2019-04-05 07:11:17][6] Processing: App\Jobs\StoreCommittee
[2019-04-05 07:11:40][6] Processed: App\Jobs\StoreCommittee
[2019-04-05 07:11:43][7] Processing: App\Jobs\StoreCommittee
[2019-04-05 07:12:05][7] Processed: App\Jobs\StoreCommittee
[2019-04-05 07:12:08][8] Processing: App\Jobs\StoreCommittee
[2019-04-05 07:12:31][8] Processed: App\Jobs\StoreCommittee
[2019-04-05 07:12:34][9] Processing: App\Jobs\StoreCommittee
[2019-04-05 07:12:57][9] Processed: App\Jobs\StoreCommittee
[2019-04-05 07:13:00][10] Processing: App\Jobs\StoreCommittee
dell@DESKTOP-UQ2 MINGW64 /d/wamp64/www/project(master)
$
(it stops without any error or failed notifications)
我在工作逻辑上有什么可以改进的地方吗?我该如何处理所有这些连接丢失,超时或其他问题?我不认为增加超时是解决方案。由于无法保证它将在此固定时间内完成。
有没有办法,可以关闭连接,然后在每个工作队列之间重新建立连接?
答案 0 :(得分:0)
您解析了CSV文件,并尝试通过单个查询发送全部内容。 MySQL包含阻止其接受太大查询的变量。叫做max_allowed_packet
原因是您要这么做。但是,在处理关于数据量太大的查询时,您可能会遇到许多与网络/ MySQL相关的变量之一。
仅准备一次语句。准备好的语句只使用一次,执行多次
解析CSV并遍历记录
将值绑定到准备好的语句并在循环中执行它
要使所有操作更快,请使用事务。在事务中包装每1000条记录。这样一来,您就可以编写简单的插入查询,但是它们很快,因为MySQL会复用写操作
您正在使用Laravel,因此上述步骤非常简单
$csv = collect([]); // This is the array holding your CSV records
// Split the array into chunks. Let's assume you want to insert 1000 records in one attempt
$chunk_count = ceil($csv->count() / 1000);
$csv->chunk($chunk_count)->map(function($chunk) {
\DB::beginTransaction();
// Create a record
$chunk->map(function($data) {
StoreCommittee::create($data);
});
\DB::commit();
});