diff tools cannot handle moving code blocks
- Adding a line: ✅
- Removing a line: ✅
- Modifying a line: ✅
- Moving a codeblock: ❌ i see you've rewritten everything, let me just highlight it all.
RIP reviewers on my PR.
(Meme created by my coworker)
569
Comments35
I’m surprised that after almost 20 years of versioning C code, git still manages to assign the closing brace of a function wrongly.
Because text is text and all
}are the same.I'm convinced there must be a way of using ASTs to do more intelligent diffing of a given programming language, but I'm far too lazy to find out for myself.
You mean like Difftastic?
Hell yeah, being lazy paid off. Thanks.
You're welcome. And have fun trying to break it!
There is, but your dif tool would have to be language aware and likely be slower to show difs.
There have been some attempts at semantic diffs, but it’s very uhh… difficult to gain traction with such a thing.
Diffing algorithms on trees might not be as efficient, especially if they have to find arbitrary node moves.
I wouldn't expect it to be, but I think modern processors can handle the load!
It's not necessarily about the load, it's about the algorithmic complexity. Going from lists (lines in a file, characters in a line) to trees introduces a potentially exponential increase in complexity due to the number of ways the same list of elements can be organized into a tree.
Also, you're underestimating the amount of processing. It's not about pure CPU computations but RAM access or even I/O. Even existing non-semantic diff implementations are unexpectedly inadequate in terms of performance. You clearly haven't tried diffing multi-GB log files.
Log files wouldn't fall under the banner of compiled languages or ASTs, so I'm not sure how that example applies.
And I'm aware that it can lead to O(n²) complexity but, as others have provided, there are already tools that do this, so it is within the capabilities of modern processors
Yes there will be cases where the size of the search space will make it prohibitive to run in reasonable times but this is - by merit of the existing tools and the fact that they seem to work quite well - an edge case.
Log files themselves don't, but I'm just comparing it with simpler files with simpler structure with simpler algorithms with better complexity.
That's awesome. I wonder how it'd handle moving plus a small change.
Too bad GitHub doesn't support it yet afaict. But at least it's not all diff tools.
Me adding one if statement around around a code block. Ah shit I guess I own the whole function now.
At least a good diff tool will ignore whitespace diffs.
laughs in IntelliJ
that's not even a joke, I'm using intellij community as a merge and diff tool exclusively. it doesn't support the language I want but even without it it's better then anything else.
Beyond Compare is awesome and handles this and many other things quite well.
VSCode has had that feature for some months now. Maybe it's still hidden behind an off-by-default setting, but it's there and I use it.
I'll have to take a look to see if I can use it to view (enterprise) GitHub PRs, because that'd be a huge help
Good question. Maybe GitLens can help with that, if not the official GitHub extensions.
Most diff tools have an option to ignore leading or trailing whitespace changes.
Me omw to shift the entire codebase to the right by one tab and claim authorship over every line in the project with a completely untraceable commit
I think OP meant moving a code block up or down in a file, not left or right
Yes, thank you. I probably should have been more clear
If you're going to rearrange code, have that as a separate check in from changing code.
Maybe specify a different diff algorithm? Histogram for example?
Give me some love for adding an indent level either showing nothing changed or you rewrote everything too.
Try difftastic! https://github.com/Wilfred/difftastic
That logo tho. oh noooooo
It doesn't actually detect moved code, though, like
git diffcan? I gave it a shot and also there's a couple issues open about it, e.g. https://github.com/Wilfred/difftastic/issues/520 .Other than that, difftastic is quite nice.
How do you expect it to be shown though?
For example, on side by side views, you can draw a box around it on both sides, and draw a line connecting the two
I think Sublime Merge does this with the conflict resolution tool