A PDF file is far more than a sequence of pages stored end-to-end. Internally, the Portable Document Format organizes its data as a collection of numbered objects — fonts, images, page descriptions, metadata — linked together through a structure called the cross-reference table (xref table). This table acts as an index, mapping every object number to its byte offset within the file so that a PDF reader can jump directly to any object without scanning the entire document. At the top of the object hierarchy sits the document catalog, which points to the page tree, outline (bookmarks), and other top-level structures.

When two or more PDFs are merged, the process is fundamentally an exercise in reconciling separate object namespaces. Each source file has its own set of numbered objects starting from 1, so the merger must renumber every object in subsequent files to avoid collisions. Every internal reference — a page pointing to its font dictionary, an annotation linking to a destination — must be updated to reflect the new numbering scheme. The cross-reference table of the output file is then rebuilt to index all objects from every source document in a single unified table.

The page tree structure deserves special attention during merging. PDF organizes pages in a tree (not a flat list) where intermediate nodes can define inheritable attributes like media size, rotation, and default resources. A naive merge might simply graft each source document's page tree as a subtree of the new root, but a well-implemented merger flattens inherited attributes onto individual pages first. This ensures that page-level properties like dimensions and rotation are preserved correctly regardless of how the original documents structured their page trees.

Object streams add another layer of complexity. Modern PDFs often pack multiple small objects into compressed object streams to reduce file size. During merging, these streams must be either decoded and re-encoded or carefully spliced so that the output document remains valid. Font and image resources shared across pages within a single source document ideally remain shared rather than being duplicated, keeping the merged file as small as possible.

Merging also involves combining or choosing among document-level metadata: the title, author, creation date, and XMP metadata blocks from each source file. Bookmark trees, named destinations, and interactive form fields must be merged without conflicts. A robust merger handles all of these edge cases to produce a single, coherent PDF that a reader can navigate, search, and render exactly as though it had been created as one document from the start.