From: Mike Hommey <mh@glandium.org>
Date: Thu, 13 Apr 2023 01:56:40 GMT
Subject: similarity index vs. whitespaces
Message-ID: <20230413015640.h6npzp47tnob7bq2@glandium.org>

Hi,

I hit an interesting case of copy-detection "failure" after renaming a
python script and changing indentation in it at the same time: git show
-C wouldn't detect it as a rename unless I lowered the similarity index
significantly. But from a human perspective, the similarity index feels
wrong.

Here's a small illustrative example:
$ git init qux
$ cd qux
$ cat > qux <<EOF
a
b
c
d
e
f
EOF
$ git add qux
$ git commit -a -m qux
$ git mv qux hoge
$ sed -i 's/[cde]/  \0/' hoge
$ git commit -a -m hoge
$ git diff -C HEAD^!
diff --git a/hoge b/hoge
new file mode 100644
index 0000000..9ab6fcc
--- /dev/null
+++ b/hoge
@@ -0,0 +1,6 @@
+a
+b
+  c
+  d
+  e
+f
diff --git a/qux b/qux
deleted file mode 100644
index 0fdf397..0000000
--- a/qux
+++ /dev/null
@@ -1,6 +0,0 @@
-a
-b
-c
-d
-e
-f

$ git diff -C10% HEAD^!
diff --git a/qux b/hoge
similarity index 33%
rename from qux
rename to hoge
index 0fdf397..9ab6fcc 100644
--- a/qux
+++ b/hoge
@@ -1,6 +1,6 @@
 a
 b
-c
-d
-e
+  c
+  d
+  e
 f

From a human perspective 33% similarity feels way too low here. I know
it's essentially counting lines in the diff, but that feels limited.

What do you think?

Mike