Julia Neumann

I take these jobs because they look trivial until you handle the failure cases, and the failure cases are where the data loss happens. Merging appends documents. The things to check are the things each document carries with it: form field definitions that collide by name, annotations that no longer point at anything after a reorder, and bookmarks whose destinations now resolve to the wrong page. A merge can produce a file that opens cleanly and is still wrong. Splitting sounds simpler and has an ambiguity worth naming. Page ranges from a print job mean the printed sheet numbering, which is not the page numbering in the reader when the document has foldouts. I always check which numbering the range is using. Reordering is a document-level operation, so anything that referenced a page position has to move with it. Cross-references, link annotations and outline entries are the three that break. Rotation is worth doing in the file rather than by turning the page, because a rotated page still prints on portrait paper with the content sideways, and the recipient has to know why. Encryption survives most of these operations and has to be reapplied afterwards, which is the step people forget when splitting a protected document. Metadata handling is the other quiet trap in a merge. Combining documents from two sources produces a file whose properties usually come from the first one, which is a small piece of provenance that can mislead when the files come from different places. Splitting a document also affects its outline and its form values, so the resulting file may need those rebuilt rather than merely preserved.