I wand to read biiiiig CSV-Files and want to insert them into a database. That already works:
if(($handleF = fopen($path."\\".$file, 'r')) !== false){
$i = 1;
// loop through the file line-by-line
while(($dataRow = fgetcsv($handleF,0,";")) !== false) {
// Only start at the startRow, otherwise skip the row.
if($i >= $startRow){
// Check if to use headers
if($lookAtHeaders == 1 && $i == $startRow){
$this->createUberschriften( array_map(array($this, "convert"), $dataRow ) );
} else {
$dataRow = array_map(array($this, "convert"), $dataRow );
$data = $this->changeMapping($dataRow, $startCol);
$this->executeInsert($data, $tableFields);
}
unset($dataRow);
}
$i++;
}
fclose($handleF);
}
My problem of this solution is, that it's very slow. But the files are too big to put it directly into the memory... So I wand to ask, if there a posibility to read, for example 10 lines, into the $dataRow array not only one or all. I want to get a better balance between the memory and the performance.
Do you understand what i mean? Thanks for help.
Greetz
V
EDIT: Ok, I still have to try to find a solution with the MSSQL-Database. My solution was to stack the data and than make a multiple-MSSQL-Insert:
while(($dataRow = fgetcsv($handleF,0,";")) !== false) {
// Only start at the startRow, otherwise skip the row.
if($i >= $startRow){
// Check if to use headers
if($lookAtHeaders == 1 && $i == $startRow){
$this->createUberschriften( array_map(array($this, "convert"), $dataRow ) );
} else {
$dataRow = array_map(array($this, "convert"), $dataRow );
$data = $this->changeMapping($dataRow, $startCol);
$this->setCurrentRow($i);
if(count($dataStack) > 210){
array_push($dataStack, $data);
#echo '<pre>', print_r($dataStack), '</pre>';
$this->executeInsert($dataStack, $tableFields, true);
// reset the stack
unset($dataStack);
$dataStack = array();
} else {
array_push($dataStack, $data);
}
unset($data);
}
$i++;
unset($dataRow);
}
}
Finaly I have to loop the Stack and build in mulitiple Insert in the method "executeInsert", to create a query like this:
INSERT INTO [myTable] (field1, field2) VALUES ('data1', 'data2'),('data2', 'datta3')...
That works much better. I still have to check the best balance, but therefor i can change only the value '210' in the code above. I hope that help's everybody with a similar problem. Attention: Don't forget to execute the method "executeInsert" again after readin the complete file, because it could happen that there are still some data in the stack and the method will only be executed when the stack reach the size of 210....
Greetz
V
I think your bottleneck is not reading the file. Which is a text file. Your bottleneck is the INSERT in the SQL table.
Do something, just comment the line that actually do the insert and you will see the difference.
I had this same issue in the past, where i did exactly what you are doing. reading a 5+ million lines CSV and inserting it in a Mysql table. The execution time was 60 hours which is unrealistic.
My solutions was switch to another db technology. I selected MongoDB and the execution time was reduced to 5 minutes. MongoDB performs really fast on this scenarios and also have a tool called mongoimport that will allow you to import a csv file firectly from the command line.
Give it a try if the db technology is not a limitation on your side.
Another solution will be spliting the huge CSV file into chunks and then run the same php script multiple times in parallel and each one will take care of the chunks with an specific preffix or suffix on the filename.
I don't know which specific OS are you using, but in Unix/Linux there is a command line tool called split that will do that for you and will also add any prefix or suffix you want to the filename of the chunks.