Is Compare reliable?

I’ve begun using compare a lot to see different revisions in different versions. The granularity I’ve chosen is at sentence level, as ‘paragraph’ is too wide, and ‘word’ too narrow.

However, I’ve noticed that while all actual changes are successfully highlighted, there are sometimes also some highlighted sentences where nothing has changed. I look closely and occasionally see there is a small detail such as punctuation, which would account for it. Other times - no matter how hard I look - there seems to be no difference between the struck out red and the underlined blue.

As all the actual changes are there, I can live with this (reluctantly), but I do wonder why it is happening.

Do you have a reproducible example? I always use word level compare, and I haven’t really noticed this. But is could be possible there is a buglet somewhere in that compare has to strip RTF first I assume and perhaps the stripping introduces some non-visible character differences?

If you have some diff software that supports RTF you could try comparing directly (the files are all in the Package contents)…

Here’s one example:

Screen Shot 2017-03-16 at 12.01.06.png

I’ve made it bigger so it’s easier to see, but you can see the lack of change it’s highlighted as a “change”.

There are two spaces between " Fun. And", but only one before the second " And", right?

That’s possible. (Not sure). If that is the reason, surely it must be possible to exclude spacing along with font, font size, formatting etc, that is already excluded? I don’t want spacing to prompt a change in the comparison, unless I chose to, which would be almost never. After all, line spacing is ignored already.

I tested it. If you have Granularity set to Word it won’t bother about extra spaces, but if you untick Word the whole sentence will be marked as changed.

Yes, you’re absolutely right. I’ll leave the granularity at ‘Word’ level from now on.

Note that the Format -> Options -> Show Invisibles command will make spaces, paragraph breaks, and other invisible characters visible. That’s a big help when trying to detect this sort of change.


But if you don’t want those to be taken into account for changes, would it help?

No. It doesn’t affect how Compare works, it just makes it easier to identify what has actually changed when Scrivener’s markup alone is too mysterious.


Being able to ignore white space differences for Compare would be a valid wish list item (though pretty low priority, given the workarounds).

Agreed. I’ve adopted ‘word granularity’ which solves the problem I had with it.