fs.existsSync(brokenSymlink) returns false

[fuweichin]

• Version: node/6.10.2; npm/3.10.10
• Platform: Windows 10 64-bit
• Subsystem: fs (File System)

fs.existsSync(linkName) returns false when linkName is a broken symbolic link path(whose target does not exist).

Is the result by design or like a bug?

[mscdex]

Sounds logical to me.

[bnoordhuis]

Yes, it's expected; symlinks are supposed to be transparent. I'll close the issue.

[kingces95]

Er.. so how do I test for the existence of a broken symlink?

[addaleax]

@kingces95 You can use fs.lstat and fs.stat/fs.exists on a symlink; if the former succeeds and the latter fails, you’ve got a broken symlink.

[tomasreimers]

Hi, sorry to resurrect this thread - I'm finding myself in need of something similar: I am trying to move a file to NEW_PATH, and if NEW_PATH already exists, the operation will fail.

The naive solution is to do something like:

if (fs.existsSync(NEW_PATH)) fs.unlinkSync(NEW_PATH);
fs.moveSync(OLD_PATH, NEW_PATH);

However, this will break if NEW_PATH is a broken symlink (b/c exists will return false, and move will fail), which doesn't feel super intuitive and potentially like a quick way for library authors to shoot themselves in the foot (as not everyone will be thinking about the case where there are broken symlinks).

Is there a function that is better suited than existsSync()? Or is there a more straightforward way than putting a try-catch around lstat(NEW_PATH)? Or is the belief that it is the correct behavior for libraries to fail here?

If those aren't the case, thoughts on adding a flag to existsSync options to check for symlinks? Or possibly a pathExists() method?

[addaleax]

Is there a function that is better suited than existsSync()?

You could not use existsSync and unconditionally use unlinkSync() (and then ignore ENOENT errors from that).

Or is there a more straightforward way than putting a try-catch around lstat(NEW_PATH)?

You might want to be more specific about the error you’re seeing, because fs.renameSync(OLD_PATH, NEW_PATH); just works for me.

If those aren't the case, thoughts on adding a flag to existsSync options to check for symlinks? Or possibly a pathExists() method?

I’d be against that – you already came up with good alternatives :)

[tomasreimers]

Appreciate it! And unlinking unconditionally is a good idea :)

I believe I was doing a copy, not a move--my mistake--, and that broke b/c it was trying to dereference the path.

Thank you!