How do you force git to keep a directory but ignore all the files inside it?

Asked

Viewed 405 times

6

I have a certain folder and logs which is essential to the functioning of my application. I would like to add the same in my repository, for those who replicate does not want to keep creating the same when doing the git clone. But the detail is that the log files should not go to the repository, but only the folder.

How to do this?

git

  • Git points to files, not directories, so it won’t be possible to add an empty directory/folder.

  • In a way, yes, but see another answer where the AR managed to "simulate" empty directory behavior by placing a hidden file

2 answers

3

To do this in the git just add a file .gitignore inside the desired folder and then determined that all the files except the .gitignore will be ignored inside that folder.

Structure:

logs/
    .gitignore

Content of .gitignore that’s inside the briefcase logs:

*
!.gitignore

Now just use the command git add logs/. Thus, the folder logs will be added to the repository, but the log files in that folder will be ignored.

I usually use this a lot for uploads folder that need to exist, but I don’t need the files that are in it.

  • 1

    Ingenious, but I must say not elegant. I’ve seen one workaround in the build of a project, always have the folder available to build in wild environments. I think the most elegant for the need to share an artifact would be to generate a compressed version with such a folder or have an "initializer" script that would provide this structure, not through version control.

2

There are some points I disagree with Reply from @Wallacemaxters, as well as the reply from @egomesbrandao also did not please me as a whole. I will give a quick explanation of how the folder and file scheme of a repository works git and then highlight why @egomesbrandao’s statement is correct and why the @Wallacemaxters solution answers the question.

I’m not gonna stick to repositories either Bare, the reading of the text of the reply should be made with this in mind.

The git works by identifying files, from the repository root (the location where the directory is located .git). The identification of a file, however, is not done in the traditional form of conventional file systems, but is more similar to what Amazon uses in S3.

So, to the .gitignore of the replies of @Wallacemaxters, the git identifies this file by name /logs/.gitignore. It is so much so that changes in the position/name of the file git suggests that you changed the file x to the archive y, regardless of the change being just a rename file within the same directory, or have moved to another directory keeping the name, or even name and directory change.

This file identification mode implies that there is no directory concept (within the internal identification of the git at least). One of the impacts of this is that you don’t have to worry about keeping directories. Move all files from /outer/inner/deep for /outer/inner implies that when the next clone, the briefcase /outer/inner/deep will not exist. And you did not need to say this explicitly.

The lack of this concept of directories in the git is the essence of @egomesbrandao’s answer, as well as the origin of one of the original question problems.

The solution proposed by @Wallacemaxters is to make the best use of this mode of operation git. Internally, folders do not exist. What exists is file identifier with bars in the middle. On the other hand, by "extracting" these files and putting them into the actual file system, directories are created so that the "directory name" is the same as the internal name used by git.

My point of disagreement with @Wallacemaxters' reply is: when giving git add logs, you have not added the directory (as this concept does not exist), but ALL the relevant content within the pointed directory. The .gitignore, in this case, it is a hidden file on Unix file systems, so the impression given is that the directory has been created empty. To maintain the directory structure, any file would be enough. Hidden would be better.

The @Wallacemaxters solution also points to the issue of maintaining empty folder contents logs in the actual file system (with the exception of .gitignore, of course). Why the file chosen to "populate" the folder was the .gitignore, as well as being hidden (ideal to make the directory exist) it also prevents recognition of new files.

In many places it is common to use a file .gitkeep empty. Unlike the @Wallacemaxters response, the archive .gitkeep There’s no special interpretation, so there’s no point in putting something inside it. The difference of this alternative is that the directory where it is will not be immediately ignored. This solution has been used in this answer.

  • 1

    I agree with your answer, it was an explanation of how my "gambiarra" worked. Laravel uses this in the application’s original repository.

Browser other questions tagged git

You are not signed in. Login or sign up in order to post.