Set Custom XMP Metadata

xmputil has typed setters for pdf:, xmpMM:, pdfaid: and pdfuaid:. For any other namespace, take the underlying document with GetGoXmpDocument, get the model from the matching go-xmp package, set its fields, and marshal as usual. The example does this for xmpRights:, the rights management namespace.

FindModel and MakeModel are the pair to know. FindModel returns nil if the packet has no node for that namespace; MakeModel returns the existing model or creates and attaches an empty one. Use MakeModel when writing.

Doing it

goXmpDoc := xmpDoc.GetGoXmpDocument()

rights, err := xmprights.MakeModel(goXmpDoc)
if err != nil {
    return err
}

rights.Certificate = "56c69b3eaf10cff5c3bd8932f0169b"
rights.UsageTerms = xmp.NewAltString("My Custom Usage Terms")
rights.Owner = xmp.NewStringArray("Custom Owner")
rights.WebStatement = "My Custom Web Statement"

if err := rights.SyncToXMP(goXmpDoc); err != nil {
    return err
}

data, err := xmpDoc.MarshalIndent("", "\t")
if err != nil {
    return err
}

metadataStream, err := core.MakeStream(data, nil)
if err != nil {
    return err
}
return pdfWriter.SetCatalogMetadata(metadataStream)

The xmputil.Document and the go-xmp document are the same object, so edits made through the unwrapped handle are picked up by xmpDoc.Marshal and MarshalIndent. There is no write-back step.

Field types come from the xmp package rather than being plain strings. Multi-value properties are built with constructors: xmp.NewAltString for a language alternative such as UsageTerms, xmp.NewStringArray for an unordered bag such as Owner, xmp.NewStringList for an ordered sequence. Assigning a bare Go string to one of those fields will not compile.

SyncToXMP is what mirrors a model’s properties into related namespaces. It is a no-op for xmpRights, but meaningful elsewhere: pdf.PDFInfo.SyncToXMP copies title, author, subject and copyright into dc: and the creation and modification dates into xmp:. Calling it after you set fields keeps the packet internally consistent, and costs nothing when the model does not need it.

Limitations

Cross-namespace sync only fills blanks. pdf.PDFInfo.SyncToXMP copies into dc: and xmp: only where the target property is empty, so an existing dc:title is never overwritten from pdf:Title. If those have to agree, set the dc: model yourself.

Marshaling a document loaded from a file rewrites the whole packet. Namespaces go-xmp does not model are still preserved as nodes, but formatting and property order will not match the original bytes, which matters if something downstream is comparing packets.

core.MakeStream with a nil encoder writes the XML uncompressed, which is what you want for a metadata stream. SetCatalogMetadata rejects anything that is not a stream, and clears /Metadata if you pass nil.

Setting XMP does not update the /Info dictionary, and the rights namespace has no /Info equivalent at all, so there is nothing to mirror in that direction here.

Run the example

The example copies the input into a writer, loads the existing XMP packet or starts a new one, fills in the four xmpRights properties, and writes the packet to the output file. Read it back with the get custom XMP metadata example.

git clone https://github.com/unidoc/unipdf-examples.git
cd unipdf-examples/metadata
go run pdf_set_custom_xmp_metadata.go input.pdf output.pdf

If this is your first time using UniPDF, follow the getting started guide to create an API key and set up your development environment.