How to Split a PDF: Range or Every N Pages

Split, extract, remove, reorder: choosing the right tool

Four different jobs sound almost identical: "I only need these pages", "take out the surplus", "break it into parts", "change the order". The tools for them look alike too, and picking the wrong one is easy.

It is easier to sort this out by the shape of the result rather than by the wording of the task. The question that settles half the doubt: how many files should exist at the end, and do the remaining pages have to go somewhere.

Choosing by the shape of the output

From that table follows the rule most often broken. If the output has to be a single file, split-pdf is not the tool: it always hands back an archive, even when there are only two parts. For one file made of chosen pages, extract-pages exists.

The opposite mistake happens too: a document is run through extraction one page at a time, taking several passes where a single split would have done.

Extract or remove: a question of arithmetic

These two produce nearly the same result by different routes, and the choice between them is purely practical.

Extraction assembles a new document out of the pages you ticked. Removal leaves the document as it was, minus the pages you ticked.

Six pages needed out of forty: tick six in extraction. Three pages to drop out of forty: tick three in remove-pages. Count which list is shorter.

There is a substantive difference too. Removal keeps a document a document: the file properties survive and the outline is rebuilt with new page numbers. Extraction creates a new file, which keeps only those bookmarks whose pages made the selection.

Removal also has a keep-selected mode in which the ticks mean the opposite. That makes it resemble extraction, with one distinction: the surviving pages are forced back into the original order, whatever order you ticked them in.

When you really need reordering

The job "swap these pages around" is solved by none of splitting, extracting or removing. Not one of the three can change the order: all they do is choose what stays.

Order belongs to organize-pdf, which is also the only one that can duplicate a page when the same sheet has to appear in the document twice.

An easy tell for this kind of job: the page count of the output is not lower than the input's. If the number of pages stays the same or grows, this is reordering rather than selection.

Regular cutting versus meaningful cutting

Two tools cut a document into parts, and they differ in where the boundaries come from.

split-every-n-pages cuts by counter: a new part every N pages. The boundaries owe nothing to the content, and nothing has to be marked up. It has a "start from page" field, which is handier than it sounds: the cover and the contents can stay in the first part if you begin cutting at the third sheet.

Splitting by ranges cuts by meaning: you say where the section boundaries fall. More work, but the parts match the content rather than the arithmetic.

The rule for choosing is simple. If the document is uniform, a register, a data dump printout, a long scan, cut by counter. If the document has structure, a report with sections, a contract with annexes, mark up the ranges.

The ceiling on the number of parts

Splitting has a limit on how many parts it will make, and it is not there out of spite: every part is created, compressed and packed into an archive.

The check runs before the first file is written. A doomed job fails immediately, and the message names both how many parts would have been produced and what the maximum is. That beats waiting five minutes for a refusal.

The cure is coarser parts. Instead of splitting into single pages, take every tenth page. Instead of two dozen small ranges, consolidate into five.

Worth remembering separately: the ceiling counts parts, not pages. A thousand-page document splits into ten parts quite happily and does not split into a thousand.

What happens to the table of contents

All four operations carry the outline across, but in different ways, and for documents with navigation that is worth knowing.

On removal, bookmarks are renumbered to the new page positions. An entry pointing at a deleted page is dropped while its children move one level up, so a whole branch is not lost.

On extraction, only the bookmarks belonging to the extracted pages reach the new file. The structure can come out noticeably flatter if you took pages from the middles of sections.

On reordering, bookmarks follow their pages to the new positions.

On splitting, each part receives its own share of the outline. Parts that caught no entries have no outline at all.

Names of the parts, and why they look like that

Splitting packs its output into an archive, and the names inside were not chosen for looks. In nearly every mode they carry the source page numbers: `pages_001_010.pdf`, `range_001_pages_001_003.pdf`, `page_001.pdf`.

The leading zeros are not decoration. Without them the tenth part would sort before the second in a file list, and reassembly would need the order restored by hand. With them it is enough to merge the files sorted by name.

Extraction with one file per page numbers the parts by the source document rather than by the order of your selection. Pick pages 3, 7 and 12 and you get files numbered 3, 7 and 12, not 1, 2 and 3. That is convenient: the file name tells you where the page came from.

Reversibility and what it costs

A last consideration, better held in mind before the operation than after.

Splitting is reversible: the parts merge back together, and the order restores itself from the names, because part names carry the source page numbers with leading zeros.

Removal and extraction are irreversible in the sense that the dropped pages are absent from the result. Your original file of course remains, but nothing can be recovered from the output.

Reordering is reversible only by hand: the tool does not store the previous order.

The practical conclusion is single and dull: keep the source until the result has been checked and sent.

One more difference people forget

Extraction has a preserve-page-size switch. On by default, it leaves every page its own format.

Switched off, it brings every selected page to the format of the first one in the selection: the content is scaled as a whole, proportions kept, and centred on the new sheet. Nothing gets stretched; a landscape page inside a portrait selection simply becomes smaller and sits in the middle.

The key detail: the model is the first page you selected, not the first page of the document and not A4. If your selection begins with an odd-sized sheet, everything else is brought to it.