dfuse is a D language binding for the high level fuse library. It allows to write a fuse filesystem for Linux or Mac OS (using osxfuse) in D.
Fuse is a library and kernel extension to write filesystems in userspace. These filesystems are easy to implement and have access to userland components like HTTP libraries, etc. For more information about fuse see: http://fuse.sourceforge.net.
A simple filesystems implementing a directory listing can be found in the examples/ directory. You can build the examples using:
$ make examples
$ mkdir /mnt/simplefs
$ ./simplefs /mnt/simplefs
dfuse provides a high level interface for libfuse. To implement a filesystem, extend the Operations class in the dfuse.fuse module:
import dfuse.fuse;
class MyFS : Operations
{
override void getattr(const(char)[] path, ref stat_t s)
{
/* implementation */
throw new FuseException(EOPNOTSUPP);
}
override string[] readdir(const(char)[] path)
{
return [/*...list of files...*/];
}
override ulong read(const(char)[] path, ubyte[] buf, ulong offset)
{
/* implementation */
throw new FuseException(EOPNOTSUPP);
}
}
A minimal filesystem implements Operations.getattr()
, Operations.readdir()
, Operations.read()
. See dfuse/fuse.d for implementation specific details.
To mount a filesystem use a Fuse object and call mount:
import dfuse.fuse;
int main(string[] args)
{
/* foreground=true, threading=false */
auto fs = new Fuse("MyFS", true, false);
fs.mount(new MyFS(), "/mnt", ["allow_other"]);
}
Error conditions are handled by throwin a FuseException with the appropriate error number. See man 3 errno
for more information about errno.
dfuse requires:
- Mac OS X or Linux
- fuse >= 2.8.0 or osxfuse >= 2.6.0
- DMD/Druntime/Phobos >= 2.065
dfuse comes with a standard makefile that assumes that DMD (the D-compiler) is in your $PATH.
In order to compile dfuse on Linux:
$ make dfuse
or
$ make dfuse ENABLE_DEBUG=1
to build a debug version
MacOS supports two inode sizes which are both supported by OSXfuse, however when compiling dfuse you have to be aware which OSXfuse should be linked.
By default dfuse is trying to build with a 32bit inode size and link against osxfuse_i32 which is part of OSXfuse for compatibility. Please note that your library itself will still be 64bit on a 64bit system. The setting only affects the size of the inode.
To build just run
$ make dfuse
If you want to compile with 64bit inodes you need a at least DMD, Druntime, Phobos in version 2.066:
$ make dfuse ENABLE_64BIT=1
dfuse comes with experimental support for dub, a package manager for D. See the dub documentation how to build and use dub.
At the moment the dfuse makefile doesn't support an install target. It is recommended to just include the library in a project at this point.
dfuse is a simple D wrapper. It exposes a lowelevel interface to the libfuse C functions in c/fuse/fuse.d. The lowlevel interface preserves C types.
A highlevel interface is provided by fs/fuse.d. The D interface initializes fuse filsystems operations structure and installs it's own handlers. Every dfuse handler converts C types to D types and is trapping FuseExceptions used for error handling. The handlers keep track of the initialized Operations object and call the appropriate method once types are converted and pass the result into the D layer.
The user facing interface is the Operations class in fs/fuse.d. It provides default implementations for all handlers and every method can be invidually overwritten to provide an interface.
If you encounter issues or bugs with dfuse, please file an issue on github. Please ensure that you maintain a constructive feedback atmosphere and if possible attach a reproduction step. If you have any questions, feel free to write to the D mailinglist or ask in IRC.
Pull requests are highly appreciated!
- Website: https://github.com/facebook/dfuse/wiki
- Mailing list: The D Mailinglist
- irc: irc.freenode.net #d
dfuse is Boost-licensed. We also provide an additional patent grant.