On Thu, Dec 1, 2011 at 3:41 PM, Simon Hengel wrote:

...

...

How to get this path in Haskell?

If I understand you correctly, you want

  takeDirectory `fmap` getProgName

I think getProgName does not give you the full path, but only the program name.

Neither does $0, does it? It depends on how the program is called. You can always use System.Directory.getCurrentDirectory with System.FilePath.{isRelative,replaceDirectory} if you somehow need the full path. Note, however, that not even this is generally guaranteed to be correct. Cheers, -- Felipe.

On Thu, Dec 01, 2011 at 07:02:09PM +0100, Simon Hengel wrote:

On Thu, Dec 01, 2011 at 03:53:37PM -0200, Felipe Almeida Lessa wrote:

...

On Thu, Dec 1, 2011 at 3:41 PM, Simon Hengel wrote:

...

...

...

How to get this path in Haskell?

If I understand you correctly, you want

  takeDirectory `fmap` getProgName

I think getProgName does not give you the full path, but only the program name.

Neither does $0, does it? It depends on how the program is called. $0 depend everything you need to find your program (say, the relative or absolute path used), but getProgName does not. Here is an example:

./foo/foo

Here $0 will be "./foo/foo", but getProgName will be "foo". s/depend/contains

On Thu, Dec 1, 2011 at 14:26, dokondr wrote:

On the contrary, standard shell variable $0 - contains a full path to the program location in the directory structure, no matter from what directory the program was called

If the shell found it by $PATH search, $0 will be simply the program name with no directory and you will have to repeat the PATH search yourself. (And the pathological case: the user did "PATH=something yourprog", where "something" does not contain the directory holding "yourprog".) There is no 100% reliable way to get the executable without using something like /proc/self/exe (only on Linux). -- brandon s allbery allbery.b@gmail.com wandering unix systems administrator (available) (412) 475-9364 vm/sms

Quoth wren ng thornton ,

There was a discussion about this recently over on libraries@, IIRC. The short answer is that, at present, there is no function to give you $0. We'd like to add such a function, but it's not been done yet.

Part of the problem is that, as Alexey says, the first element of argv is just whatever is passed to exec, which is not guaranteed to be a complete path, a canonical path, or any other specific thing we'd desire. It's not at all straightforward to determine the actual location of the executable, especially not in a platform-independent manner. argv[0] can't be trusted, scanning through $PATH isn't guaranteed to find it (and even if you find something of the right name, it's not guaranteed to be the correct executable), etc etc.

This isn't your problem, though. I mean, if the proposed function is to return the path to the executable file, then indeed you have a big problem with argv[0]. argv[0] can be anything - or nothing (it can be a null string.) But if we've turned to the question of whether to return argv[0], that's much simpler: you don't need to consider why a programmer might want it. It's appalling to think that library developers would withhold access to standard POSIX 1003.1 features while they decide whether they approve of them. Donn

Part of the problem is that, as Alexey says, the first element of argv is just whatever is passed to exec, which is not guaranteed to be a complete

That's true even for regular fork/exec. -----Original message----- From: Richard O'Keefe To: wren ng thornton Cc: haskell-cafe Sent: Sun, Dec 4, 2011 15:54:15 PST Subject: Re: [Haskell-cafe] How to get a file path to the program invoked? On 4/12/2011, at 7:32 PM, wren ng thornton wrote: path, a canonical path, or any other specific thing we'd desire. It's not at all straightforward to determine the actual location of the executable, especially not in a platform-independent manner. argv[0] can't be trusted, scanning through $PATH isn't guaranteed to find it (and even if you find something of the right name, it's not guaranteed to be the correct executable), etc etc. In particular, with posix_spawnp(), the $PATH that is used to find the executable and the $PATH in the environment that the executable starts with can be two different things. _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

It's not a poor practice at all. Example: gcc, which uses the executable's path as the base directory from which other files are located. MacOS also does something similar. -----Original message----- From: Paul R To: dokondr Cc: Simon Hengel , haskell-cafe Sent: Sun, Dec 4, 2011 15:26:28 PST Subject: Re: [Haskell-cafe] How to get a file path to the program invoked? dokondr> On the contrary, standard shell variable $0 - contains a full dokondr> path to the program location in the directory structure, no dokondr> matter from what directory the program was called. I don't think the comparison makes sense, as shell script invocation and executable run are very different mechanisms. Whenever you invoke a shell script, what really happens is that a program in your path (sh, bash ...) gets started with an argument that is the path to the script to load (your script actually). In this situation, you understand that it is easy to provide the path to the script (the $0) : it is just the file that the interpreter is loading. I don't know if it is possible at all to get this information in the context of binary execution. And I'm not sure it is a good practice anyway :) -- Paul _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Anton Nikishaev writes:

dokondr writes:

...

Hi, When my program starts it needs to know a complete path to the directory from which it was invoked. In terms of standard shell (sh) I need the Haskell function that will do equivalent to:

#!/bin/sh path=$(dirname $0)

How to get this path in Haskell?

getProgName :: IO String defined System.Environment only returns a file name of the program without its full path.

Thanks!

http://hackage.haskell.org/package/system-argv0-0.1

(which is argv[0], and surely it's not always the “complete path to the directory from which it was invoked”. $0 is not either.) -- lelf xmpp:nik_n@jabber.ru

On 16/12/2011, at 11:55 AM, Brandon Allbery wrote:

Note that "exec -a" is a bash-ism and not portable to POSIX shells

Recent versions of ksh also support this, so it's not just bash. But there are certainly a lot of POSIX shells that don't, including the version of ksh on my main machine.