does hardlink's --exclude prevent descending a directory?
#3244
-
|
Does/can I have two directories I want to hardlink where I could save a lot of time and computation by being able to prevent descending one particular subdirectory in each of the two directories. If I were using |
Beta Was this translation helpful? Give feedback.
Replies: 4 comments 7 replies
-
|
hardlink visits every regular file in the tree, and the --include / --exclude expressions are applied for each file. So, it does not prevent the visit of any subdirectory. The expression is applied to the file names, not to directory names (it internally ignores directories). |
Beta Was this translation helpful? Give feedback.
-
|
I can imagine implementing --exclude-dir, but it will require re-implementing the way it walks the files tree. Currently, it uses the nftw() function from libc. |
Beta Was this translation helpful? Give feedback.
-
|
Isn't |
Beta Was this translation helpful? Give feedback.
-
|
Ah, good point. I had no clue that nftw() can react to the callback's return value in a smart way (FTW_ACTIONRETVAL). Yes, it seems we can implement --exclude-subtree easily :-) |
Beta Was this translation helpful? Give feedback.
-
I thought so to. I was hoping to find some time soonish to create a PR, but not sure I will find the time before you will (given how simple it is). So please do go ahead if you find a minute. Not sure when I will. If I remember, I will update the answer here with a pointer to a PR/landing commit if/when that happens rather than the answer I did choose above (temporarily). |
Beta Was this translation helpful? Give feedback.
-
|
I wonder if, while you are at it, a |
Beta Was this translation helpful? Give feedback.
-
|
Should be implemented by PR: #3261 |
Beta Was this translation helpful? Give feedback.
-
|
Very nice. I seem to have run into a small usability issue though. When I execute something like: I'm getting matches for exclude such as : Should I be using RE anchors such as: (i.e. will those work as I have used them above?) Additionally, should/ do I [need to] further tighten my REs with: To avoid skipping things like |
Beta Was this translation helpful? Give feedback.
-
|
The expression is used with a full path when entering the directory. So, the directory name is the last item in the path -- it means it's a string after the last '/'. For example, if I have:
It makes sense to use '/' after the directory name only if you want to check the parental directory too. For example |
Beta Was this translation helpful? Give feedback.
-
So a full path including any prepending directories? I.e. when hardlinking the two directory trees Correct? |
Beta Was this translation helpful? Give feedback.
-
|
The command line argument "$SNAPROOT/$1".0/{pc,jenny}" is converted into two paths by the shell. The hardlink command sees two independent trees: /.snapshots/daily.0/pc/foo and /.snapshots/daily.0/pc/foo. During the initial stage, when hardlink scans the directories, each path is evaluated separately to filter out relevant files based on user-defined conditions. In the second stage, it works with all accepted files to identify duplicates and create links. Simple tip (I have to add this to the man page): If you are unsure about the path used in the regular expression, refer to the verbose output. The path shown in the verbose output is the exact string that hardlink uses when evaluating the expression (e.g. "Skipped (excluded subtree) /.snapshots/daily.0/pc/usr/include/boost/variant"). |
Beta Was this translation helpful? Give feedback.
Ah, good point.
I had no clue that nftw() can react to the callback's return value in a smart way (FTW_ACTIONRETVAL). Yes, it seems we can implement --exclude-subtree easily :-)