You need pages 12 to 18 of a 200-page report. There are three buttons in front of you and they all appear to do it. They do not produce the same thing, and the difference matters more once you have the output.
The three operations
Split takes one file and gives you many. Either one file per page, or one file per range you define. Use it when you want separate documents, such as breaking a batch of scanned invoices into one file per invoice.
Extract takes one file and gives you one file containing the range you asked for. Use it when you want a shorter document, which is the case most of the time. Here, that is the split tool with a single range such as 12-18.
Delete takes one file and gives you one file with the pages you named removed, keeping everything else. Use it when what you want to get rid of is smaller than what you want to keep.
Extract and delete are complements. Extracting 12 to 18 from a 200-page file and deleting 1 to 11 and 19 to 200 produce an identical result. Reach for whichever one requires you to type fewer numbers, because every number typed is a chance to get it wrong.
The page numbering trap
This catches nearly everyone once.
A PDF numbers its pages by position in the file, starting at 1. The numbers printed on the pages are just ink. In a report with a cover, a blank verso, a contents page numbered i and body text starting at printed page 1, the printed page 1 is the file's page 5.
So "extract pages 12 to 18" means twelfth to eighteenth sheet in the file, which may be printed pages 8 to 14.
What breaks when you take pages away
Internal links and bookmarks. A link that pointed to "page 143 of this document" has nowhere to go once page 143 is gone. A bookmark tree referring to removed sections either disappears or points at the wrong place. This is not a fixable defect: you removed the destination.
Cross-references in the text. "See the table on page 47" survives as text and is now wrong. Nothing will warn you.
Form fields spanning the document. If a form's fields were defined at document level and you keep only some pages, filled values can be lost. Flatten before extracting if the form is filled.
Nothing else. Text stays sharp, images are copied byte-for-byte rather than re-encoded, fonts come along with the pages that use them. Extraction is lossless for page content, which is why extracting three pages from a large report is almost always a better way to hit an email size limit than compressing the whole thing.
Extracting images is a different job
"Extract images" and "convert pages to images" sound similar and are not related.
Converting pages to images renders each page, text and all, into a picture at a resolution you choose. You get a picture of the page.
Extracting images pulls the embedded image objects out of the file and gives you the originals. If a report contains eight photographs, you get eight image files at whatever resolution they were embedded at, with no page furniture around them and no re-encoding.
The second is what you want when you need the pictures. It also reveals something people forget: an image cropped inside the document is usually stored uncropped, because cropping in a PDF changes the visible window rather than the data. Extracting can therefore hand you more of the photograph than the document displays.
A short order of operations
- Look at the document first and note the file positions, not the printed numbers.
- Flatten any filled forms before splitting or extracting.
- Extract rather than compress when the goal is a smaller file and you only need part of it.
- Open the output and read the first and last page. Off-by-one errors are the common failure and they show up immediately at the boundaries.
In practice: split with one range to extract a section, or several ranges to cut a file into parts; delete pages when it is quicker to name the pages that go; and extract images when you need the pictures rather than the pages.