argparse.FileType does not expand tilde "~". 

This would be useful to take file parameters beginning with "~" and use os.path.expanduser to expand this.

Add patch to add an option to argparse.FileType to enable expanduser.

Update patch to include docs.

Why not take this a step further and make it take a callable that's expected to take the command line argument and coerce it into a path

Normally the unix shell takes care of this expansion.  If I call a simple script that prints sys.argv and runs a simple parser, I get:

    2135:~/mypy$ python2.7 filetypetest.py ~/mypy/test.txt
    ['filetypetest.py', '/home/paul/mypy/test.txt']
    Namespace(file=<open file '/home/paul/mypy/test.txt', mode 'r'...)

I have to quote the argument '~/mypy/test.blog' to bypass this expansion.

This is under Linux (bash).  Is Windows or Mac different?

My impression is that `FileType` has been provided as an example of a custom `type`, as opposed to a comprehensive file handler.  But I can't quote a developer on this.  

http://bugs.python.org/issue13824 is a bug report that worries that 'argparse.FileType opens a file and never closes it'.  There are some comments there about whether it is appropriate to expand on the capabilities of FileType.

I think this change is not appropriate for Unix, because it is redundant and would conflict with normal shell expansion. I’m not familiar with Windows, but I suspect it is not appropriate there either (neither Windows command shells nor argparse do wildcard * and ? glob expansion, right?). Pending response from Garth or Julian.

I'm with Martin.

My support (or really, asking for a more generic callable) was to enable other use cases, not this one specifically -- e.g., allowing for constructing a pathlib Path or a twisted.python.filepath.FilePath without needing to write one's own argparse action.

For ~ specifically -- it's true if you're going to go via a shell it'd be expanded first, but I'm not sure that'd count as "conflicting" with the shell's expansion, it just won't be needed. Consider that twisted.python.filepath.FilePath for example, expands ~ in its constructor -- people often construct paths in other places and occasionally expansion is convenient. Like in test cases.

I certainly don't feel strongly about the patch though, especially as-is.

I’m not super familiar with argparse, but to get a Path or FilePath object, can’t you use that directly instead of argparse’s FileType class?

The proposal would conflict with escaping a literal tilde (~) in the shell. It would make it harder for the end user to specify a real file name that starts with a tilde.

messages: + msg262807

messages: + msg262569

nosy: + martin.panter
messages: + msg262546

resolution: rejected

messages: + msg205954