joinPath("foo", "/bar") should return "/bar"

[timotheecour]

joinPath("foo", "/bar") should return "/bar"
It's almost always the right thing to do. Eg:

• https://github.com/citycide/glob/blob/b6de2fd1a9eff022bc0006e565840f431cb99015/src/glob.nim#L197 has to introduce this workaround:

proc maybeJoin (p1, p2: string): string =
  if p2.isAbsolute: p2 else: p1 / p2

Most languages I use do that, eg:
D, python, C#

• D: https://dlang.org/phobos/std_path.html#buildPath

rdmd --eval 'std.path.buildPath("a", "/b").writeln'
/b

• python:

os.path.join('a', '/b')
Out[3]: '/b'

• C#: https://msdn.microsoft.com/en-us/library/fyy7a5kt%28v=vs.110%29.aspx

For C++: as mentioned here (rust-lang/rust#16507 (comment)) The C++ committee has recently been dealing with this problem.
http://isocpp.org/files/papers/n4211.html issue 64

• NOTE: the current buggy behavior is not documented (no example like joinPath("foo", "/bar") is provided) so that makes fixing it less problematic.

• The PR that fixes it should make sure overload joinPath(parts: varargs[string]) is also fixed (with behavior: any absolute path found kills all the previous arguments) ; likewise with / operator

option B: if for some reason the breaking change is not acceptable, I suggest this: intoduce an optional third arg enum:

JoinBehavior = enum
  throwIfAbs # throw if 2nd arg is abs
  joinAsPath # behavior I suggest
  joinAsString # lagacy behavior
  
joinPath(string, string, joinBehavior = throwIfAbs) = string

joinPath("foo", "bar") # foo/bar
joinPath("foo", "/bar") # will throw
joinPath("foo", "/bar", joinAsPath) # /bar
joinPath("foo", "/bar", joinAsString) # foo/bar

option C: another option is to go ahead with the proposed semantic change but when a flag --transition=8268 is given, throw when 2nd arg is abs

But I prefer to keep it simple and correct and follow the semantics of D, python, C#

[timotheecour]

/cc @andreaferretti @dom96 would a PR to fix that be accepted?

[dom96]

Sure, I guess.

[Araq]

I think it's a terrible idea, but if everybody else does it this way... shrug

Copy all files to /backup:

# assume 'walkFiles' happens to return absolute paths:
for x in walkFiles("*.nim"):
  # yay, '/' is smart and makes this either a nop or a failure
  copyFile(x, "/backup" / x)

[timotheecour]

just use relativePath for that:

copyFile(x, "/backup" / x)
=>
copyFile(x, "/backup" / x.relativePath("/"))

where relativePath is defined here nim-lang/Nim#8166 (comment) (same as python's, D's versions)

[Araq]

You know... I think we need type Path = distinct string; type Filename = distinct string; type FileExt = distinct string and a new path handling module that uses Nim's type system.

[dom96]

You know... I think we need type Path = distinct string; type Filename = distinct string; type FileExt = distinct string and a new path handling module that uses Nim's type system.

Too late for v1 ;P

[github-actions]

This RFC is stale because it has been open for 1095 days with no activity. Contribute a fix or comment on the issue, or it will be closed in 7 days.

[ZoomRmc]

Precedents are convincing enough, but I'd argue that the current beheviour seems logical: "Interpret /bar in context of foo being a parent directory". Might be useful in a hypothetical chroot scenario.

Current implementation of relativePath is broken, btw.

[Zectbumo]

Please take into consideration the security risk when changing the library:

"htdocs" / "/etc/passwd"

current version:
file not found

after a surprise update that includes this proposal:
returns passwd contents

Btw,

NOTE: the current buggy behavior is not documented (no example like joinPath("foo", "/bar") is provided) so that makes fixing it less problematic.

This has since been documented.
https://nim-lang.org/docs/os.html#joinPath%2Cstring%2Cstring
assert joinPath("usr/", "/lib") == "usr/lib"

[Araq]

Rejected, it's a terrible idea and an invitation for security breaches.