Sure, there's a lot of efficiency to Ctrl-F a text string and just find all the places in a document. I won't deny that it takes me longer to pull up the index, search for the function name in the index, then flip to the page. But then I can just leave the book open at that page on the desk (or my lap). I never have to Alt-Tab, or fiddle with the location of windows to switch between looking at documentation and looking at the code I'm working on.
This difference was more stark when I was trying to close-read a different specification to ensure that I understood it well enough to make sure a PR implemented it correctly. I needed to have three different parts of the specification open simultaneously to bounce between all of them. With physical paper, that's just a swish of a hand away. With a PDF reader, well, goto that other section, scroll down to the piece I wanted, now goto the first section again and scroll down again and wait what was that back thing again goto and scroll and scroll and goto and descent into insanity. Trying to use multiple windows ameliorates the problem somewhat, but it also takes an inordinate amount of time to set the view up correctly, and I often end up running into the "focus doesn't follow the eye gaze" problem of typing in the wrong window and ruining the view.
I pretty much just use screenshots in snagit for that stuff.
Of course a good index is hard (read expensive) to write and so many books didn't have good indexes.
The index is so good I've shared my happiness about it several times.
Let me give you an example of a high-quality index entry from the Software Architecture in Practice (Bass et al. 2021) [1]:
Availability
analytic model space, 259
analyzing, 255–259
broker pattern, 240
calculations, 259
CAP theorem, 523
CIA approach, 147
cloud, 521
design checklist, 96–98
detect faults tactic, 87–91
general scenario, 85–86
introduction, 79–81
planning for failure, 82–85
prevent faults tactic, 94–95
recover-from-faults tactics, 91–94
summary, 98–99
tactics overview, 87
As you see, it lists a number of taxonomic terms that are merely related to each other and you might not think about Ctrl+F-ing for them unless you already want to search for them. You may come to this entry knowing about CAP and navigate away to analytic model space, 259.
[1]: https://www.goodreads.com/book/show/14786083-software-archit...
It is quite disheartening to see a comment about importance of horse shoes being downvoted. In professional blacksmith shops, horse shoeing is a job done by a qualified farrier and not as trivial as one may think. The importance of horseshoeing for horse wellbeing is also highlighted in certain key equestrian literature.
Some are. Far, far, far, far, far too many aren't.
The half-assedry of PDF creation is a major frustration.
There are scanned-in books whose index pages don't precisely match the digital pages. Good PDFs will account for that offset themselves, but manual recalculation may be necessary.
Worse are books half-assedly converted from print to digital. These often include an index (useful for all the reasons others have mentioned elsewhere in this thread), but the "faithful" reproduction of the print text means that the page enumeration in the index bears a nonconstant relationship to the digital text. The offsets are not constant.
Then there are ePubs with the above feature. The sane thing to do would be to link the index entry to occurances. Often you'll find, again, print-edition page mentions which are of little use in locating the passage within your digital edition.
One of the underlying problems is that the print notion of "page" is increasingly archaic. For languages / typographies in which paragraphs are a useful convention, paragraph numbering might be preferable (this should be constant across formats). Direct symlinks are of course useful, but these conceal information revealed in a conventional (print) index such as passages where a topic is discussed at some length, or clusters of appearances, as well as cross-references or associated references which a well-constructed index will reveal.
There's a pre-digital publishing trend of loose-leaf or removeable bindings, with publications prepared in sections or with periodic supplements, which began in the late 19th / early 20th century. Those would typically be organised and numbered by section. The concept is somewhat trite now, but I think of it as a significant stage and evolution of publishing, somewhere between fixed-format codices, periodicals, and eventually databases and wikis.