d_type/dirEntType only exposed in Internals?

[sjshuck]

The only module that exposes struct dirent { ...; unsigned char d_type; ... } is an internal one. Reason I'm interested is, recursive walking of filesystems in Haskell are slow compared to Python's os.walk(). Haskell ignores the d_type and then does one or more separate system call per file to determine whether it's a file or directory. Example

I might want to work on this, but I don't know between unix, streaming-common, and conduit which is the right place to begin. It's not really a technical question—I could do it from anywhere—I just want to know if this has been attempted before, and if so, what issues were encountered.

It looks like there's not a sum type in existence for d_type values corresponding to https://github.com/bminor/musl/blob/86373b4999bfd9a9379bc4a3ca877b1c80a2a340/include/dirent.h#L51-L59 or glibc's equivalent. I'd say there should be, but it would clash conceptually with this.

What are some thoughts?

[Bodigrim]

I might want to work on this, but I don't know between unix, streaming-common, and conduit which is the right place to begin. It's not really a technical question—I could do it from anywhere—I just want to know if this has been attempted before, and if so, what issues were encountered.

I think https://hackage.haskell.org/package/directory-ospath-streaming is the top choice for recursive walking of filesystems. It seems to employ DirEnt in some modules.

[sjshuck]

That looks pretty good, thanks @Bodigrim . And I was wrong about there not being a sum type for d_type, or at least there's equivalent functionality in the form of pattern synonyms here.

The package you linked still depends on System.Posix.Directory.Internals. Is that okay, long-term?

[Bodigrim]

The package you linked still depends on System.Posix.Directory.Internals.

IIRC directory-ospath-streaming has provisions to support unix both before and after the introduction of System.Posix.Directory.Internals. I suppose if it disappears in favor of another mechanism, the package will add yet another set of conditionals.

Is that okay, long-term?

There is a long discussion at #251, which I honestly cannot grasp in full, because I know virtually nothing about these interfaces.

[sjshuck]

I'm happy enough with the current status that I can close this. Thanks again.