How to automatically remove all .orig files in Mercurial working tree?

Question

During merges mercurial leaves .orig file for any unresolved file. But after manually resolving problems and marking a file correct it does not delete the .orig file. Can it be automatically removed by some command?

I work on a Mac so I can use something like:

find . -iname '*.orig' -exec rm '{}' ';'

and alias it or something, but I'd rather use something like hg cleanup...

UPDATE:

Since some time now, Purge extension is bundled with Mercurial and solves this problem nicely.

Solution 1

Personally, I use

$ rm **/*.orig

if I get tired of the .orig files. This works in Zsh and in Bash 4 after you run shopt -s globstar.

But if you use another shell or want a built-in solution, then maybe you'll like the purge extension (link updated 2016-08-25). That lets you remove all untracked files with

$ hg purge

You can remove all untracked and ignored files with

$ hg purge --all

An advantage of using hg purge is that this will also cleanup directories that become empty after removing the files. The rm command line will just leave the empty directories behind.

Solution 2

btw. find utility has an action -delete so you can type only:

find <path-to-files> -name '*.orig' -delete

Solution 3

If you just want to delete the .orig files, and you happen to be on a Windows computer, the following seems to work well:

D:\workspace>hg purge -I **/*.orig --all

This will delete all untracked files that end in .orig, but won't delete other untracked files, as the other answers would.

You can test this before running it by putting the --print flag in as well.

Solution 4

The following will remove .orig files in your entire working copy hierarchy and also works for paths containing spaces:

find . -name *.orig | while read -d $'\n' file; do rm -v "$file"; done;

I use an alias in my .bash_profile:

alias clearorig='echo "Removing .orig files..."; find . -name *.orig | \
while read -d $'\''\n'\'' file; do rm -v "$file"; done;'

Solution 5

I'm working in Powershell and didn't see the answer here:

# NOTE: be in the root of your repository
# fetch all .orig files recursively
$orig = (dir *.orig -recurse) ;

# remove each .orig file
foreach ($item in $orig) { del $($item.FullName) ; }

# afterwards I make sure to remove the references to the .orig files in the repo
# then commit & push
hg addremove ; 
hg commit -m "remove .orig" ;
hg push ;

View more solutions

Comments

[rkj]

rm is quite short :). But the purge removes all untracked files - and some file I use are not tracked on purpose - private project configuration, database configuration, etc...

[rkj]

can you be more more specific? Are you proposing removing all files on commit hook? :)

[dfa]

you can execute find . -iname '.orig' -exec rm '{}' ';' *after a merge, see the doc for details

[Martin Geisler]

Why am I getting downvotes for this? Is it because this only works in shells like Zsh that understand **? I said personally since this is what I use myself, and then I gave a general, cross-platform answer in the form of the purge extension.

[Martin Geisler]

This is not a good answer: file names with spaces will break the rm invocation. Please let me know if you fix the answer so that I can remove the down vote.

[Benbob]

@Martin It doesn't recurse into sub-directories is my guess.

[Martin Geisler]

@Keyo: the rm command line does recurse, that's the purpose of the ** part. It wont remove directories that become empty after removing the .orig files. That's a slight advantage of using hg purge. It's rare that directories come and go in my projects, so I hadn't thought of this before.

[Gary]

You need to add double quotes "" around the variable $f. This is a common problem in a lot of scripts that operate on files and directories, including Apple documentation (which I notified them to update years ago and still has incompatible script examples) :P

[Collin Klopfenstein]

@MartinGeisler fixed! :)

[Martin Geisler]

@CollinKlopfenstein: Imagine you have files named foo.txt and bar baz.txt. The find call will return foo.txt\nbar baz.txt\n and the for loop will split this into three files: foo.txt, bar, and baz.txt. Only the first name is correct. This is why you need find -print0 or should let find execute rm.

[miguelSantirso]

I agree this is the best option if you have purge installed

[Mecki]

How is that piping through several processes any better than find . -name '*.orig' -exec rm '{}' ';' or even simpler find . -name '*.orig' -delete, where all work is performed by find and rm directly or just by find on its own?

[Mecki]

How is that piping through several processes any better than find . -name '*.orig' -exec rm '{}' ';' or even simpler find . -name '*.orig' -delete, where all work is performed by find and rm directly or just by find on its own?

[mlissner]

Correct me if I'm wrong here, but won't this have all the usual issues that you have when you send output to xargs without using the -0 flags? From the man page: "-0 Input items are terminated by a null character instead of by whitespace, and the quotes and backslash are not special (every character is taken literally). Disables the end of file string, which is treated like any other argument. Useful when input items might contain white space, quote marks, or backslashes. The GNU find -print0 option produces input suitable for this mode."

[mlissner]

@Mecki, pretty sure it's not, and that it's prone to errors when files contain weird characters...like newlines, for example.

[Gary]

@mlissner, you are correct. In my case, we typically didn't put spaces or newlines in filenames, so for the general case this solution above wouldn't work.