FIx IPython 0.13 issues

[jason]

Can we keep this ticket open to track all the remaining IPython issues left over from #12719?

Still outstanding:

From Kannappan Sampath:

1. I am unable to exit by typing "exit", but Ctrl+D however works!
2. The following (from #12719)

   jec@fermat%./sage
   ----------------------------------------------------------------------
   | Sage Version 5.7.beta1, Release Date: 2013-01-26                   |
   | Type "notebook()" for the browser-based notebook interface.        |
   | Type "help()" for help.                                            |
   ----------------------------------------------------------------------
   **********************************************************************
   *                                                                    *
   * Warning: this is a prerelease version, and it may be unstable.     *
   *                                                                    *
   **********************************************************************
   ---------------------------------------------------------------------------
   NameError                                 Traceback (most recent call last)
   /home/jec/sage-5.7.beta1/local/lib/python2.7/site-packages/IPython/utils/py3compat.pyc
   in execfile(fname, *where)
       176             else:
       177                 filename = fname
   --> 178             __builtin__.execfile(filename, *where)
   
   /home/jec/.sage/init.sage in <module>()
   ----> 1 load /home/jec/sage/ecdb.py
   
   NameError: name 'load' is not defined
   

apply:

• trac_14066-quitexit.patch​
• trac_14066-startup.patch​

[jason]

Another

[jason]

Regarding (3), loading a sage file (i.e., %runfile init.sage) certainly works with sage syntax like load. So the problem appears to be that the %runfile code isn't being used to run init.sage.

[jdemeyer]

(2) is certainly not related to IPython.

[jason]

For (1): In devel/sage/sage/all.py, we find this code at the very top:

# Error message that matches the Sage/IPython defaults
quit = "Use Ctrl-D (i.e. EOF), %Exit, or %Quit to exit without confirmation."
exit = quit

Who put that there? Deleting those lines restores exit and quit to act just as expected.

[jason]

It looks like William put it there in 2006, so apparently it is a leftover from the old system. I wonder how exit and quit worked before; clearly they didn't just print the string!

[jhpalmieri]

So issue (1) seems to have a solution. Here's an idea for (2):

sage/misc/sage_extension.py

@@ -453 +453 @@
         """
         startup_file = os.environ.get('SAGE_STARTUP_FILE', '')
         if os.path.exists(startup_file):
-            self.shell.run_cell('%%run %r'%startup_file)
+            with open(startup_file, 'r') as f:
+                self.shell.run_cell(f.read())
 
     def init_inspector(self):
         # Ideally, these would just be methods of the Inspector class

(That is, call run_cell on the contents of the init file, instead of calling run_cell(%%run ...) on the init file itself.)

[jdemeyer]

I don't know much about this IPython stuff, but it would be good to resolve these issues as this is the only remaining blocker for sage-5.7.

[jason]

John, that fix seems reasonable. Looking at the IPython source, the default for the %run magic is to run using the python execfile function, so that explains why sage-specific things didn't work. run_cell should apply the sage-specific things.

But I would change the call to run_cell(f.read(), silent=True) (and then check to make sure tracebacks still work; if not, then use run_cell(f.read(), store_history=False))

[jason]

To be consistent with shell.runfile_ipy, I just use store_history=False...

[chrisjamesberg]

I've noticed that attach no longer works. It says it's been deprecated and to use %attach, but then this appears to have a bug.

If I have a file "test.py" which contains one line

print "hello"

then when I attach:

%attach 'test.py'

then it prints hello. But if I rewrite the line to:

print "hey"

and then hit enter on sage, it doesn't print anything.

[jhpalmieri]

I think that %attach sort of works, maybe even mostly works, but printing does not. If my file contains the line a=3:

sage: %attach test.py
sage: a
3
sage: # now I edit test.py to say a=12.0
sage: a
12.0

Function definitions get updated just fine, but as you point out, print statements don't seem to work right. What else needs to be tested to know if print statements are the only issue?

Note that if I attach a file with print statements, then change it, then attach it again, the print statements appear twice. So I think the change is being detected, but the output isn't showing up when it should.

[jdemeyer]

Replying to chrisjamesberg:

and then hit enter on sage, it doesn't print anything.

I don't quite understand what the problem is. Is it a problem that simply hitting ENTER doesn't reload the file, that the file is only reloaded before executing an actual command?

Because other than that, I find that %attach works fine. So, I'm willing to give this positive_review.

[jason]

Regarding attach, I suppose that is a difference between the old and new. In the old, the attach expressions were run after every enter press (i.e., before the prompt was displayed). Now they are run just before any code is executed. I guess the issue is that blank lines in some sense "don't count" because of a shortcut. Let me see if I can add an option for attached files to run there as well.

[jason]

It's going to be a bit tricky to execute attached files on empty input (i.e., just pressing enter) because, by default, IPython ignores attempts to execute code on empty strings. Do we really need this?

[vbraun]

I don't think its necessary. I'm sure it'll break somebody's workflow (​http://xkcd.com/1172/) but there are better solutions to "reload when saved" that don't involve sending non-commands.

[vbraun]

Since nobody complained I set this to positive review ;-)

[jhpalmieri]

See #14144 for another one.