The object model

edit

A PDF file is a graph of objects: dictionaries, arrays, streams, names, strings, numbers and references between them. The high-level members (Pages, Information, Annotations...) read and write these objects; anything they do not cover can be done with the objects themselves. The types are in the Meziantou.Framework.Pdf.Objects namespace.

Read objects

var document = PdfDocument.Open(file);

// The fonts of the resources of the pages: the /BaseFont of each font dictionary
var fonts = new SortedSet<string>(StringComparer.Ordinal);
foreach (var page in document.Pages)
{
    if (page.Resources?.GetResolved("Font") is not PdfDictionary pageFonts)
        continue;

    foreach (var name in pageFonts.Keys)
    {
        if (pageFonts.GetResolved(name) is PdfDictionary font && font.GetName("BaseFont") is { } baseFont)
        {
            fonts.Add(baseFont.Value);
        }
    }
}

PdfPage.Dictionary is the page object, PdfPage.Resources its resources (inherited from the page tree when the page has none), PdfDocument.Catalog the root of the document, and PdfDocument.Trailer the trailer of the file.
The value of an entry is often a PdfReference to an indirect object. The indexer returns the value as it is stored; GetResolved follows the reference, and Get<T> and TryGetResolved<T> also check the type of the value.
GetName, GetText, GetNumber, GetInteger and GetBoolean read the common values. A string converts to a PdfName, so "Font" is the name /Font.
PdfPage.GetContentData returns the decoded content of the page, and PdfStream.GetDecodedData the decoded data of any stream.

Change objects

var document = new PdfDocument();
document.Pages.Add(PdfSize.Letter);

// The catalog is a dictionary of the object model: entries the API has no member for can be set directly
document.Catalog[PdfName.Create("PageLayout")] = PdfName.Create("TwoColumnLeft");
var viewerPreferences = new PdfDictionary
{
    [PdfName.Create("DisplayDocTitle")] = PdfBoolean.True,
};
document.Catalog[PdfName.Create("ViewerPreferences")] = document.Objects.Add(viewerPreferences);

Set an entry of a dictionary with its indexer, or with SetName, SetText and SetNumber; remove it with Remove.
PdfDocument.Objects.Add makes an object indirect and returns its reference, which other objects can point to. Remove removes an indirect object: the references to it then point to the null object.
A document can only refer to its own objects: a save throws InvalidOperationException for a reference that belongs to another document. Names cannot contain the null character: PdfName.Create throws ArgumentException. A name read from a file with #00 is kept, and written as it was read.
PdfStream.SetData replaces the data of a stream with data without filter, which the save compresses (CompressStreams); SetEncodedData sets data that is already encoded, with its filter.
The high-level members read and write the same objects: a change made through either is visible through both, and an incremental save writes the objects that changed.

The types

Type	PDF object
PdfDictionary	A dictionary, whose entries keep their order
PdfArray	An array; FromNumbers creates an array of numbers, such as a rectangle
PdfStream	A dictionary with data, such as a content stream, an image or a font file
PdfName	A name, such as `/Type`
PdfString	A string: FromText and ToText convert text, FromBytes and Bytes raw bytes
PdfInteger, PdfReal	Numbers, which derive from PdfNumber
PdfBoolean, PdfNull	`true` and `false`, and the null object
PdfReference	A reference to an indirect object; Resolve returns the object

PdfRectangle.ToArray and PdfRectangle.TryRead convert rectangles from and to arrays.