Split CSV by parsed records, not by raw lines: a quoted field can legally contain a line break. PHP’s CSV parser reads one logical record at a time, so the process can remain memory-bounded and preserve valid quoting.
Last updated: September 26, 2026.
$input = new SplFileObject('customers.csv', 'r');
$input->setFlags(SplFileObject::READ_CSV | SplFileObject::SKIP_EMPTY);
$input->setCsvControl(',', '"', '');
$header = $input->fgetcsv(',', '"', '');
$rowsPerFile = 50000;
$part = 0;
$rowInPart = $rowsPerFile;
foreach ($input as $row) {
if ($row === false || $row === [null]) continue;
if ($rowInPart >= $rowsPerFile) {
$part++;
$output = new SplFileObject(sprintf('customers-%03d.csv', $part), 'w');
$output->fputcsv($header, ',', '"', '', "
");
$rowInPart = 0;
}
$output->fputcsv($row, ',', '"', '', "
");
$rowInPart++;
}Passing an empty escape character follows standard doubled-quote escaping and avoids relying on the deprecated default behavior in PHP 8.4 and later.
Validate the input contract
Confirm the delimiter, enclosure, encoding, header presence, and line-ending expectations before splitting. If the delimiter is unknown, use the delimiter detection procedure on a representative sample. Convert encoding before parsing when the source contract requires it.
Protect the output directory
Generate filenames yourself in a dedicated non-public directory. Do not concatenate an uploaded filename into a filesystem path. Check disk space, handle write failures, and write to temporary names before renaming completed parts. Record part number, row count, and checksum in a manifest so downstream processing can detect missing or repeated files.
Verify every part
- The header appears exactly once per part.
- The sum of data rows equals the accepted input count.
- Every row has the expected number of fields when the format requires it.
- Quoted commas, quotes, and embedded newlines round-trip correctly.
- No empty final part is produced.
The PHP SplFileObject CSV reference describes the delimiter and escape rules. For general iteration and error handling, review reading large CSV files without exhausting memory.
