git word diff regex strange behaviour - regex

I'm using Git to version prose and have been trying git diff --word-diff to see changes within lines. I want to use the results generated in a script.
But the default way that --word-diff identifies a word seems flawed. So I've been experimenting with --word-diff-regex= options.
Problem
Here are the two main flaws I'm trying to deal with:
Added whitespace seems to be ignored. But whitespace can be quite important if trying to use the results programmatically.
For example, take this header from a Markdown (.md) file:
# Test file
Now, let's add some text to the end of it:
# Test file in Markdown
If I run git diff --word-diff on this:
# Test file {+in Markdown+}
But the space before the word "in" has not been included as part of the diff.
Empty lines are completely ignored.
Here's a standard git diff for the content of a file where I've removed a line and also added a couple of new lines -- one empty, the other with the text "Here's a new line."
This is a test file to see how word diff responds in certain situations.
-
I'll try removing lines and adding them to see what happens.
Here's another line so we can see what happens with line removals and additions. I want to see how `git diff --word-diff` handles it all!
+
+Here's a new line.
But here's git diff --word-diff for the same content:
This is a test file to see how word diff responds in certain situations.
I'll try removing lines and adding them to see what happens.
Here's another line so we can see what happens with line removals and additions. I want to see how `git diff --word-diff` handles it all!
{+Here's a new line.+}
The removed and added empty lines are completely ignored.
Desired results
Putting the two examples above together. Here's what I'd like to see:
# Test file{+ in Markdown+}
This is a test file to see how word diff responds in certain situations.
{--}
I'll try removing lines and adding them to see what happens.
Here's another line so we can see what happens with line removals and additions. I want to see how `git diff --word-diff` handles it all!
{++}
{+Here's a new line.+}
Things I've tried:
git diff --word-diff-regex='.' seems too granular for when new words share characters with existing words
git diff --word-diff-regex='[^ ]+|[ ]' seems to solve the first problem but, to be honest, I'm not actually sure why.
git diff --word-diff-regex='[^ ]+|[ ]|^$' -- I was hoping the ^$ on the end would help capture empty lines -- but it doesn't and, worse, it then seems to ignore the change that follows.
git diff --word-diff-regex='[^ ]+|[ ]|.{0}' creates same problem as the one before.
I'd be grateful if anyone could shed any light on how to do this, or at least share some knowledge on what's going on under the hood with --word-diff-regex.

The main thing that you're running into that's stopping you from having what you want, from https://git-scm.com/docs/diff-options, is:
A match that contains a newline is silently truncated(!) at the newline.
This is going to mean that word diffs are always going to ignore line diffs. I don't think you're going to achieve the results you want short of a custom diff generator.

Related

Trying to get git diff to ignore comments using regex, doesn't seem to be working

My goal is to get git diff to ignore C comments. I've been using a basic regex, and printing the diff to another file (it doesn't print anything otherwise). I've also tried the reverse, to get the diff to only show the comments (I'll later see if I can reverse engineer it). However, it doesn't behave as it's supposed to. Here's a few examples of what I've tried:
Trying to get the diff to show only the lines that begin with /*:
git diff -w -G'(^(/\**)' master > text.diff
Getting the diff to show lines that start with either * or / or end with the same:
git diff -w -G'(^[/\*])|($[^/\*])' master > text.diff
Getting the diff to show only non-comment lines (see How to make 'git diff' ignore comments):
git diff -w -G'(^[^\*# /])|(^#\w)|(^\s+[^\*#/])' master > text.diff
I'm running it under WSL and using git version 2.17.1 for reference.
My goal is to get git diff to ignore C comments ...
This is difficult, because:
Git doesn't understand C code;
parsing C comments requires lexical analysis across multiple lines;
git diff breaks up the input into lines too early.
Your best bet is therefore not to do this directly with git diff at all. Instead:
Extract the file(s) to be compared from wherever they live (two commits, one commit and a regular file, one commit and an index copy of a file, etc).
Use a C-comment-stripper that does understand how to analyze C source and detect (and remove1) the comments.
Run the output of step 2 through some diff engine (regular diff, git diff, whatever you like).
If you wrap all of this up as a tool that git difftool can run, you'll get something serviceable and convenient. It will require generating lots of temporary files.
(Note that your attempt to use -G here is ultimately doomed. The -G expression will look for a comment within the changed lines, rather than whether the changed line is or is not in the middle of a long comment. Languages that have only comment-to-end-of-line, such as sh/bash, are more tractable than C. Backslash-newline sequences will still foil things though. See also Erik Aronesty's answer to the linked question.)
1Remember that in ANSI C, comments always separate tokens, so for ANSI C, replace comments with white-space, but in many traditional K&R compilers, comments simply vanish. This technique is used in place of the new-in-1989 token-pasting operator in some very old C code. You might want to support this mode by making step 2 have an option to leave out the white-space.

Why is this vim regex so expensive: s/\n/\\n/g

Attempting this on a sufficiently large file (say 80,000+ lines and about 500k+) will crash things or stall eventually both on my server and on my local Mac.
I've tried this at the command line as well, with the same result:
vim -es -c '%s/\n/\\n/g' -c wq $file
Also, the problem appears to be with the selection (\n) and not the replacement (\\n).
For my larger files I can of course split them and cat them back when finished, but the split points cannot be arbitrary in my case and must be adjusted manually for each and every split.
I appreciate that there are other ways to do this -- sed, etc. -- but I have similar and additional problems there, and I would like to be able to do this with vim.
I'm adding my comment as an answer:
Text editors usually don't like 'gigantic' lines (which is what you'll get with that replacement).
To test that if this is is due because of the 'big line' and not the substitution itself I did this test:
I created a simple ~500KB file with a script. No new line characters, just a single line. Then I tried to load the file with vim. Result? I had to kill it :-).
However, if on the same script I write some new lines every now and then, I have no problems opening the file.
Also, one thing you could try is the following: on vim, replace \n by \n\n if it is fast, then this should also confirm the 'big line' issue.

Ignore multiline comments git diff

I'm trying to find the significant differences in C/C++ source code in which only source code changes. I know you can use the git diff -G<regex> but it seems very limiting in the kind of regexes that can be run. For example, it doesn't seem to offer a way to ignore multiline comments in C/C++.
Is there any way in git or preferably libgit2 to ignore comments (including multiline), whitespaces, etc. before a diff is run? Or a way of determining if a line from the diff output is a comment or not?
git diff -w to ignore whitespace differences.
You cannot ignore multiline comments because git is a versioning tool, not a language dependent interpreter. It doesn't know your code is C++. It does not parse files for semantics, so it cannot interpret what is comment and what isn't. In particular, it relies on diff (or a configured difftool) to compare text files and it expects a line-by-line comparison.
I agree with #andrew-c that what you are really asking is to compare the two pieces of code without comments. More specifically helpful, you are asking to compare the lines of code where all multiline comments have been turned into empty lines. You keep the blank lines there so you have the correct line numbers to reference on a normal copy.
So you could manually convert the two code states to blank out multiline comments... or you might look at building your own diff wrapper that did the stripping for you. But the latter is not likely to be worth the effort.
You can achieve this using git attributes and diff filters as described in Viewing git filters output when using meld as a diff tool to call a sed script, which however is pretty complex on its own if you want it to handle all cases like comment delimiters inside string literals etc.

git smart line and word diff

I'd like to git diff and combine the regular line-by-line diff with git diff --word-diff. The problem with line-by-line diffs is that they're unnecessary if I change one or two words and leave the line mostly intact--the chunking is too coarse. On the other hand if I change entire lines and use --word-diff, sometimes the diff algorithm will get confused and spit out incredibly confusing diffs, with lots of words inserted and deleted to "morph" one line into another.
Is there a way to specify that git should be smart about this and only --word-diff if it actually makes sense to do so (on a line-by-line basis, of course)?
The smartest thing I have found for git diff --word-diff or git diff --color-words are the predefined patterns that come with git (as used in --word-diff-regex or diff.wordregex). They might not be perfect, but give quite good results AFAICT.
A list of predefined diff drivers (they all have predefined word regexes too) is given in the docs for .gitattributes. It is further stated that
you still need to enable this with the attribute mechanism, via .gitattributes
So to activate the python pattern for all *.py files, you could issue the following command in your repo root:
echo "*.py diff=python" >> .gitattributes
If you are interested in what the different preset patterns actually look like, take a look at git's source code

Doing a 'diff/st' and ignoring the first line if it matches a specific criterion

In a repository for a well known open source project, all files contain a version string with a timestamp as their first line:
<?php // $Id: index.php,v 1.201.2.10 2009-04-25 21:18:24 stronk7 Exp $
Even if I don't really understand why they do this - since the files are already under version control -, I have to live with this.
The main problem is that if I try to 'st' or 'diff' a release to get an idea of what was changed from the previous one, every single file contained in the repository is obviously marked as modified and the diffs become unreadable and unmanageable.
I'm wondering if there's a way to ignoring the first lines doing a diff/st when they match a regexp.
The project is under cvs - cvs, yes, you've read correctly - and included in a bigger mercurial repository.
I don't know about cvs, but with hg you can use any external diff tool with the bundled extdiff extension, and any modern tool should have the ability to let you ignore diffs that match certain patterns.
I swear by Beyond Compare, which allows arbitrary syntax definition.
kdiff3 has preprocessor commands that you can pipe the input through.
If you try
man diff
you'll find
--ignore-matching-lines=RE Ignore changes whose lines all match RE.
search "ignore matching lines" on the web gives examples :
diff --unified --recursive --new-file
--ignore-matching-lines='[$]Author.[$]'
--ignore-matching-lines='[$]Date.[$]' ...
(http://www.cygwin.com/ml/cygwin-apps/2005-01/msg00000.html)
Thus try :
diff --ignore-matching-lines='[<][?]php [/][/] [$]Id:'