Reorder the Pages in a PDF

How it works

qpdf rewrites the document's page tree — the list inside the file naming which page object appears in which position — and copies every page object across unchanged. The order changes; the pages do not.

Moved, not redrawn

There are two ways to reorder a PDF and they produce very different files.

The wrong way is to render each page and assemble a new document from the results. It looks correct on screen and it destroys everything that is not ink: the text layer becomes an image, bookmarks vanish, internal links point nowhere, form fields stop being fields.

The right way is to rewrite the page tree — the list inside the file that says which page object comes in which position. The page objects themselves are untouched. That is what happens here, through qpdf, and it is why the output is usually within a few bytes of the input and why a contents page still works afterwards.

Dragging is not the only way

Every page card has arrow buttons alongside the drag handle, and they are not a fallback for browsers that cannot drag.

HTML5 drag and drop cannot be operated with a keyboard, is awkward on a touchscreen, and is invisible to a screen reader — a page built only around it is unusable for anyone in those three groups. The arrows do the same job, are reachable by Tab, and announce which page they move.

What this deliberately does not do

It does not delete pages, and it does not rotate them.

That is a choice rather than a gap. A grid where dragging can both reposition and remove is a grid where a slightly long drag destroys a page, and there is no undo on a file you are about to download. /delete-pdf-pages and /rotate-pdf each do one thing where the consequence of a mistake is visible before you commit.

Why the order came out wrong in the first place

Three causes account for almost every reordered document.

A sheet-fed scanner fed a face-down stack produces the pages in reverse, because the sheet at the top of the pile is the last one it reaches. A duplex scan run as two passes — all the fronts, then all the backs — produces pages 1, 3, 5, … followed by 2, 4, 6, …, and interleaving them by hand is what this grid is for. And a merge of files whose names sorted differently than you expected puts sections in the wrong place, because 10.pdf sorts before 2.pdf in every file manager that sorts as text.

All three leave a document whose content is complete and whose sequence is not, which is the cheapest kind of problem to fix and the most annoying to live with.

Check before you download

The thumbnails are rendered at low resolution, which is enough to recognize a page and not enough to read it. If two pages look alike — a form and its duplicate, two pages of the same table — open the original alongside and check by content rather than by thumbnail.

The button that saves stays disabled until something has actually moved, so an unchanged document cannot be downloaded as though it had been rearranged.