Document to PDF

The document/convert package lays out an open document and writes it as a PDF without needing Word or any other application on the machine. It works on a document you opened from disk and on one you built with document.New(), and it runs anywhere Go runs, which is what makes it the right answer on a server or in a container.

Both return a *creator.Creator from UniPDF rather than bytes. Nothing is written until you call a method on it, so you can set metadata or append pages before saving.

Converting a file

doc, err := document.Open("table.docx")
if err != nil {
    log.Fatalf("error opening document: %s", err)
}
defer doc.Close()

c := convert.ConvertToPdf(doc)

if err := c.WriteToFile("output/table.pdf"); err != nil {
    log.Fatalf("error converting document: %s", err)
}

Two license keys have to be set, not one. Conversion runs through UniPDF’s creator, so the example calls SetMeteredKey on both unioffice/v2/common/license and unipdf/v5/common/license. Setting only the UniOffice key gets you a watermarked or refused PDF.

WriteToFile is os.Create underneath and does not create directories. Writing to output/table.pdf fails with a “no such file or directory” error if output/ is not already there.

Keep defer doc.Close(). Opening a document extracts it into a temporary directory, which is also where embedded fonts are unpacked during conversion, and Close is what removes it.

Fields

The converter is not an application, so it evaluates only a few field codes. Everything else falls through.

An unevaluated field is not an error and produces no warning. If the file carries a cached result from the last time Word computed the field, that cached text renders as ordinary text. If it does not, which is the case for a document UniOffice just created, you get nothing. A table of contents added with AddField(document.FieldTOC) and never opened in Word therefore converts to a blank contents page. Generate a table of contents via OLE is the way out of that.

Limitations

Fonts are the usual cause of output that looks nearly right. Helvetica, Courier and Times New Roman map to the built-in PDF base fonts. Any other font name has to have been registered, and if it has not, the converter logs a debug line and silently falls back to Helvetica in the matching style. Line breaks move because the substitute has different metrics. Use custom fonts covers registering the real files.

The godoc on ConvertToPdf marks the package as beta and warns that breaking changes can occur, so pin the library version if the exact output matters to you.

Where a document sets no page size, A4 is assumed. Where it sets no font size, 12 points is assumed. Both are adjustable through the options struct.

Run the example

The example converts ten fixture documents in a loop, one per feature: a chart, a fld_simple field document, headers and footers, an image, mail merge fields, a table, portrait and landscape text, and two textbox layouts. Each is written into output/ under the same name. Comparing a fixture to its PDF is the fastest way to see what the renderer supports.

git clone https://github.com/unidoc/unioffice-examples.git
cd unioffice-examples/document/convert_to_pdf
go run main.go

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

Sample output

A table:

Landscape text:

Headers and footers, with the page number field substituted: