document the difference between -tmd and -g*.md

[lamyergeier]

What version of ripgrep are you using?

ripgrep 12.0.1
-SIMD -AVX (compiled)
+SIMD +AVX (runtime)

How did you install ripgrep?

linuxbrew

What operating system are you using ripgrep on?

Ubuntu 20.04

Describe your bug.

rg does not respects .gitignore. See example below:

What is the actual behavior?

Test-Glob() {
	mkdir -p Temp
	cd Temp || exit
	git init > /dev/null
	cp "../Data/.gitignore" .
	mkdir -p a/.Commit && touch a/.Commit/a.md
	touch a/a.md
	echo "PWD= ${PWD}"
	echo "${Blue:=}------------Cat gitignore--------------------${Normal:=}"
	cat .gitignore
	echo "${Blue:=}------------List files not ignored--------------------${Normal:=}"
	git check-ignore -v ./**/*
	echo "${Blue:=}------------Find--------------------${Normal:=}"
	find . -type f -name "*.md"
	echo "${Blue:=}------------Ripgrep Glob--------------------${Normal:=}"
	rg --glob "*.md" --files
	echo "${Blue:=}------------Ripgrep--------------------${Normal:=}"
	rg --files
	cd ..
	#Trash Temp > /dev/null
}

PWD= /home/nikhil/Documents/Git/Cs/Architecture/ArchitectureSrc/Hardware/MemoryUnit/Harddisk/FileSystem/File/ListSearch/AckAgRipgrep/Src/Temp
------------Cat gitignore--------------------
# Ignore everything
*

# Unignore all dirs
!*/

# Unignore
!*.md

# Ignore directory
/**/.Commit/**
------------List files not ignored--------------------
.gitignore:8:!*.md      ./a/a.md
------------Find--------------------
./a/a.md
./a/.Commit/a.md
------------Ripgrep Glob--------------------
a/a.md
a/.Commit/a.md
------------Ripgrep--------------------
a/a.md

What is the expected behavior?

rg should ignore ./a/.Commit/a.md

[BurntSushi]

Thank you for the detailed bug report, but this is correct behavior. Glob overrides always have the highest precedent. So if you use -g '*.md' and ripgrep would otherwise descend into a directory containing a .md file (i.e., the directory itself wasn't ignored), then ripgrep will search ever .md file in that directory.

Interestingly, the same is not true for file types. If you use -tmd instead for example, then a/.Commit/a.md is ignored.

This does seem like something that the docs could be more explicit about, so I'll mark this as a doc bug.

[lamyergeier]

Hi I knew about -tmd, but was using glob only because I wanted to search for random extensions (non-standard).

Not complaining, but I thought rg followed the behavior of .gitignore, as you mentioned it here: #689 (comment)

You said: However, if you had read the README or the blog post introducing ripgrep, then it should have been very clear that respecting gitignore by default is a fundamental design decision.

If I am mistaken, I request you to reconsider this as by this we could support random extensions and also respect .gitigore

---

Also please see the following, when I modified .gitignore (Is this consistent with your comment? I am not sure.):

Test-Glob() {
	mkdir -p Temp
	cd Temp || exit
	git init > /dev/null
	echo '/**/.Commit/**' > .gitignore
	mkdir -p a/.Commit && touch a/.Commit/a.md
	touch a/a.md
	echo "PWD= ${PWD}"
	echo "${Blue:=}------------Cat gitignore--------------------${Normal:=}"
	cat .gitignore
	echo "${Blue:=}------------List files not ignored--------------------${Normal:=}"
	git check-ignore -v ./**/*
	echo "${Blue:=}------------Find--------------------${Normal:=}"
	find . -type f -name "*.md"
	echo "${Blue:=}------------Ripgrep Glob--------------------${Normal:=}"
	rg --glob "*.md" --files
	echo "${Blue:=}------------Ripgrep--------------------${Normal:=}"
	rg --files
	cd ..
}

------------Cat gitignore--------------------
/**/.Commit/**
------------List files not ignored--------------------
------------Find--------------------
./a/a.md
./a/.Commit/a.md
------------Ripgrep Glob--------------------
a/a.md
------------Ripgrep--------------------
a/a.md

[BurntSushi]

but I thought rg followed the behavior of .gitignore

It does. The fact that rules in a .ignore override .gitignore does not mean ripgrep doesn't follow the behavior of gitignore. And similarly for the --glob flag. ripgrep does more than just gitignore anyway. It also filters out hidden and binary files by default.

I request you to reconsider this

No, the behavior isn't changing. It would be a breaking change that feels big enough to me that I'm not comfortable with it, and while I see that there are problems with the current behavior, there would also be problems with your proposed behavior. Ultimately, I judged that the status quo was the least bad option and went with that.

Also please see the following, when I modified .gitignore (Is this consistent with your comment? I am not sure.):

Yes, because /**/.Commit/** ends up ignoring .Commit itself, which means ripgrep doesn't descend into that directory. Use the --debug flag to see how filtering rules are applied:

DEBUG|ignore::walk|crates/ignore/src/walk.rs:1730: ignoring ./a/.Commit: Ignore(IgnoreMatch(Hidden))

[lamyergeier]

Sorry for the confusion: I am really not sure why /**/.Commit/** is ignored in second case but not in the first case?

For the first case you said:

Glob overrides always have the highest precedent. So if you use -g '*.md' and ripgrep would otherwise descend into a directory containing a .md file (i.e., the directory itself wasn't ignored), then ripgrep will search ever .md file in that directory.

What i understood (from your comments): If glob finds *.md even in a gitignored directory, it will still print it.

Going by this explanation, if glob overrides always has the highest precedent, then rg should also print a/.Commit/a.md in the second case. Isn't it?

[BurntSushi]

It has to descend into the directory in the first place. Look at the --debug output. With your second gitignore file, ripgrep doesn't descend into a/.Commit at all. So there is no opportunity for *.md to match it.

This is why I wrote, "and ripgrep would otherwise descend into the directory."