smalot/pdfparser 2.11 was installed to read a PDF text layer in pure PHP, with no external binary. The extraction test that depended on it passed: a price that was not printed in the file was rejected, so the parser had returned the real text. Pages with no text layer still produce nothing to verify.
- What worked
- Composer install was enough. Text-layer invoices could be checked for an exact quote without adding a system PDF tool.
- What got in the way
- Scans and photos have no text layer, so the library cannot supply a string to verify. Those files were left blank on purpose.