What is count in a change object?

[rococode]

What do those counts mean? Unless I'm missing something it doesn't seem to be mentioned in the spec at all. In the image above I'm using word diff.

Edit: Oh, it seems it's a count of the type that was diffed? So in this case it's a count of "words" (except spaces are counted too) and for a diffChars it counts characters. What's a potential use case for that info?

[Dzenly]

I also was trying to analyze it. It seems that 'count' is the tokens count.

Tokens can be such as:

• word
• whitespaces groups
• punctuation signs
• etc.

E.g. for diffWordsWithSpace you can set breakpoint to

node_modules/diff/lib/diff/base.js:38

on the string:

var newLen = newString.length,

and see newString content, it keeps tokens.

[ExplodingCabbage]

What's a potential use case for that info?

You can add up the numbers of added/removed tokens up to show summary stats like GitHub does on pull requests:

It's also useful for understanding how the underlying algorithm works (in particular how it is tokenizing stuff), especially if you're getting results that intuitively seem like a sub-optimal diff! For instance, in the issue description you figured out from these numbers that diffWords treats spaces between words as tokens (something which I think is kinda broken behaviour, tbh, with bad consequences beyond the count values).

I can also imagine hypothetical uses like showing in some kind of review tool that the user has so far reviewed (or scrolled past) x% of all inserted content and y% of all deleted content, though I've not seen an example of such functionality in the wild.

[ExplodingCabbage]

Anyway, you're right that the docs don't mention this, and that seems bad. We should document count. I'll probably do this and #149 in one swoop since both of these issues depend upon documenting a more fundamental point: that each diffing function exposed by the library has its own way of splitting the input into "tokens" and that the core diffing algorithm then acts on the series of tokens. It probably makes sense to write up all of that in one go.

[ExplodingCabbage]

By the way, @rococode, I'd love any feedback you have on the updates to the README I've made to address this. :)