__file__ is not defined when file end with .ipy

[steverweber]

I named my IPython file.ipy
I propose .ipy ? objections ?
I feel ipython should not be confused with plain python!

echo "#!/usr/bin/env ipython" > test.ipy
echo "print __file__" >> test.ipy

cat test.ipy
#!/usr/bin/env ipython
print __file__

chmod +x test.ipy
./test.ipy
> NameError: name '__file__' is not defined

ipython test.ipy
> NameError: name '__file__' is not defined

## run with plain python
python test.ipy
> ./test.ipy
SUCCESS:

## rename file to be .py seems to solve the issue
mv ./test.ipy ./test.py
ipython test.ipy
> ./test.py
SUCCESS:

./test.py
> ./test.py
SUCCESS:

Should we not trust files with different endings... I have a feeling files not ending with .py are blacklisted to prevent issues with cgi or something... not to sure.. Perhaps it has something to do with the she-bang notation..?

debian kubuntu 12.10

ipython -v
0.12.1

Thanks

[bfroehle]

Yes, see IPython.core.shellapp.InteractiveShellApp._exec_file:

if full_filename.endswith('.ipy'):
    self.log.info("Running file in user namespace: %s" %
                  full_filename)
    self.shell.safe_execfile_ipy(full_filename)
else:
    # default to python, even without extension
    self.log.info("Running file in user namespace: %s" %
                  full_filename)
    # Ensure that __file__ is always defined to match Python behavior
    self.shell.user_ns['__file__'] = fname
    try:
        self.shell.safe_execfile(full_filename, self.shell.user_ns)
    finally:
        del self.shell.user_ns['__file__']

Running a file ending with *.ipy current is exactly the same as opening up an ipython terminal and pasting the entire file into one cell. (And in that way, __file__ is not set).

Should this be changed?

[takluyver]

I think it's a reasonable idea, but not very important - .ipy scripts are a convenience, and we don't want people relying on them for anything complex.

[steverweber]

Without this patch.
How should I get the file string.. without renaming the file to .py?

I don't see the harm with having file set.

People often need to know the dynamic location of the file and its name in shell scripts.

part of old .sh scripts that I would like to move to IPython.

#!/bin/sh
frl="$(readlink -f $0)"
fdn="$(dirname $frl)"
fn="$(basename $frl)"
fnb="${fn%.*}"
fnbb="${fn%%.*}"
fnxx="${fn##*.}"

# ... many shell calls

Anyways ill just rename to .py

[bfroehle]

Well, you can rename to anything except .ipy.

[takluyver]

If you can just call it foo.py, absolutely do so. We don't want to encourage people to write code for IPython when standard Python code would do the job. .ipy scripts exist so you can use IPython syntax, like %magic functions and !shell commands.

But I agree with you that there should be a way to find the running filename from a .ipy script. We're quite busy at the moment, but if you want to dig into the code and make a pull request, that's great. Otherwise, we'll leave this issue open, and someone will get round to it at some point.

[steverweber]

this patch fix the root issue.
sorry about the multiple commits.. darn tabs/spaces :(

test.ipy

#! /usr/bin/env ipython
print('in:'+__file__)
import os
test2 = os.path.dirname(__file__) + '/test2.ipy'
print('run:'+test2)
%run $test2 
print('in:'+__file__)

test2.ipy

#! /usr/bin/env ipython
print('in:'+__file__)

output

in:/home/s8weber/svc/ipython/test.ipy
run:/home/s8weber/svc/ipython/test2.ipy
in:/home/s8weber/svc/ipython/test2.ipy
in:/home/s8weber/svc/ipython/test.ipy

running: ipython ./test.ipy
as well as %run test.ipy
with recursion test.ipy having a nested %run test2.ipy return correct file

seems to work well now.

[Carreau]

Reopening as #1831 has been reverted.

[steverweber]

Darn now my scripts wont work again :(

I wish a fix was created instead of the revert.

[bfroehle]

I'll post something soon for you to try.

[bfroehle]

@steverweber Can you check out #2460 ?

[steverweber]

@bfroehle I confirmed your commit bfcd64e will solve this issue.

[bfroehle]

Closed by #2460.

[TheChymera]

I am still experiencing the issue with ipython 0.13.2 :-/

---> 13 print file
NameError: name 'file' is not defined

[minrk]

@TheChymera that's correct, it is fixed in 1.0 (note the milestone on the Issue), which will be released this month.

[Carreau]

(note the milestone on the Issue)

Do the non-contributor see the milestone ?

[minrk]

Do the non-contributor see the milestone ?

I believe so, yes.

[jakobgager]

I think I'm a non-contributor an I see the milestone tag at the issue heading.

[TheChymera]

Yeah, sry, I saw that, just didn't know how to interpret it.

[minrk]

No worries - in general, issues are closed when they are fixed in master. All closed issues should have a milestone (this will be None if there is no related code to change), which is generally the first release with the change (some bugfixes are backported, but not feature changes).