[spawnia]

CSVArray::toArray() split the CSV into lines before parsing.
A quoted field with a line break became several broken rows.
It now cuts records only at line breaks outside quoted fields, then parses each record with str_getcsv.

Line breaks, blank lines and empty input behave as before

Lines still split at \r\n, \n and \r, mixed in one file too.
Blank lines are skipped and keep their row number, missing columns default to '', and '' returns [].
Row keys count records, which differs from line numbers only for multiline fields.

fgetcsv over php://memory was tried and rejected

fgetcsv ends records only at \n, so \r-only and mixed line breaks merged rows.
It returns [null] for a blank line, the same as for a quoted "" in a one-column file, so such rows were dropped.
It also needed PHPStan ignores, since Safe\fgetcsv is missing in Safe 1.x.

For agents 🤖: Copilot threads addressed, CI running, then review

• splitRecords() scans the input once and follows str_getcsv quoting: an enclosure opens only at the start of a field, after spaces, tabs, \v or \f. A doubled enclosure stays inside the field, and an escape character inside it skips the next character.
• 100,000 random inputs per setting split into the same records as fgetcsv, for escape \, " and '', and for delimiter equal to enclosure. With escape \n, fgetcsv itself is no valid reference; a test covers that case.
• 200,000 rows: 1.1 s and 122 MB peak, master 0.4 s and 122 MB.
• The unreachable Missing column headers. throw stays as on master.
• make rector fails locally after composer update: Undefined constant PHPUnitSetList::PHPUNIT_40. This comes from the newer Rector PHPUnit sets and is unrelated to this change.
• After merge and release, update the consumer project to the new version.

🤖 Generated with Claude Code

[github-code-quality]

Code coverage overview

Languages: PHP

PHP / code-coverage/phpunit

master 2949ba3	AE-4872-csv-multilin... 50bb216	+/-
89%	90%	+1%

View coverage of the 6 most impacted files.

File	master 2949ba3	AE-4872-csv-multilin... 50bb216	+/-
IlluminaSampleS...ingsSection.php	100%	100%	0%
InterOp/RunResult.php	100%	100%	0%
CSVArray.php	90%	96%	+6%
PHPStan/Rules/M...ypehintRule.php	0%	88%	+88%
PHPStan/Rules/M...ypehintRule.php	0%	100%	+100%
PHPStan/Rules/C...ypehintRule.php	0%	100%	+100%

ℹ️ Calculations represent line coverage. Results may change as calculations are updated.

---

_{Updated November 04, 2026 13:01 UTC}

[Copilot]

🟡 Changes recommended

Unresolved scanner edge cases split valid multiline CSV fields into separate rows, while retaining all line separators substantially increases peak memory.

4 open issues

• Scan line-break offsets incrementally to reduce peak memory · New
• Preserve escape state across captured line breaks · New
• Prioritize enclosure when delimiter and enclosure are identical · New
• Treat vertical tabs and form feeds as leading field whitespace · New

What changed in this PR

Updates CSVArray::toArray() to parse quoted multiline fields as single records, improving round-trip conversion with toCSV().

Changes:

• Adds quote-aware record splitting before parsing headers and values with str_getcsv().
• Expands coverage for line endings, enclosure and escape behavior, and empty input; enables the existing multiline round-trip test.

File	Description
tests/​CSVArrayTest.php	Adds parsing regression cases and enables multiline round-trip coverage.
src/​CSVArray.php	Introduces field-state tracking to join lines within enclosed CSV fields.

_{Review effort: Max}

[github-actions]

🎉 This PR is included in version 7.0.1 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀