Checks a document against the PDF/A, PDF/UA and PDF/X standards, and converts an editor's document to PDF/A.

{:ok, doc} = PdfElixide.Document.open("report.pdf")
{:ok, report} = PdfElixide.Compliance.validate(doc, :pdf_a_2b)
report.compliant?
#=> false
Enum.map(report.errors, & &1.code)
#=> ["XMP-001", …]

validate/2 returns a PdfElixide.Compliance.Report listing the errors and warnings found. A document that fails a standard is an ordinary report with compliant?: false, not an error; {:error, %PdfElixide.Error{}} means the document could not be read far enough to check it.

What a report establishes

Validation runs this library's own checks for the requested standard, which cover a subset of each standard's requirements. compliant?: true means those checks found no errors; it is not a conformance certification. Use a dedicated conformance validator where a claim of compliance has to hold up.

PDF/A and PDF/UA reports list checks that were only partly performed among their warnings, under code "WARN-007", so even a report with no errors carries warnings.

A PDF/A report does not compare the level a document declares with the one requested, so a document declaring PDF/A-2b can be reported compliant with PDF/A-1b. Compare report.declared with the standard where the declaration matters. Nor does it check that the document information dictionary agrees with the XMP metadata, or that the file carries a document identifier (/ID), both of which PDF/A requires. A PDF/A-1 report does not check that the XMP metadata is stored uncompressed, which PDF/A-1 also requires.

PDF/A and PDF/X forbid encryption, so an encrypted document always fails them. PDF/UA-2 is not offered: this library performs no check specific to it.

PDF/X checks read each page's boxes from the page's own dictionary. A page that inherits its /MediaBox from the page tree is reported as missing one ("XBOX-004"), and its boxes are not checked for consistency.

Converting to PDF/A

convert/3 rewrites the document an editor holds and returns a PdfElixide.Compliance.Conversion describing what changed:

{:ok, editor} = PdfElixide.Editor.open("report.pdf")
{:ok, conversion} = PdfElixide.Compliance.convert(editor, :pdf_a_2b)
if conversion.report.compliant? do
  {:ok, pdf} = PdfElixide.Editor.to_binary(editor)
end

Conversion fixes supported findings and reports anything left unresolved, so it can finish with compliant?: false. Convert last, then validate the bytes you write.

See the PDF/A conversion guide for the fixes, editor restrictions, metadata caveats and font-embedding trade-offs.

Memory and locking

Validation reads a private copy of the document, parsed from its bytes, so it holds about two extra copies of the file in memory for its duration. It takes a shared read on the handle, and so runs alongside other reads — with one exception: a document that cannot be opened without a password is validated in place and takes the handle exclusively. See the Concurrency guide.

Conversion takes the editor exclusively. It holds several copies of the document at once: the written bytes it starts from, the converted document and the editor rebuilt from it.