cell magics do not work with empty lines after #2447

[juliantaylor]

a file cell magic looking like this:

%%file /tmp/test.py
line1

line3

will give an error with the current 1.0dev git head but it worked fine with 0.13.2

NameError: line3 does not exist

removing the empty line avoids the issue, but that is not really acceptable for readability

Also the docstring says for the notebook -f is implied but it does not say what -f does.

[takluyver]

Hmm, that sounds like a consequence of #2447. I'll look into it.

-f is force, i.e. don't prompt before overwriting the file. I agree that it should be documented.

[minrk]

-f was removed, but not from the docstring. That part should be fixed by #3182.

[grsr]

This seems to be a general problem with cell magics, the perl magic also doesn't work with newlines in the cell:

%%perl

my $foo = 'bar';
print $foo;

produces:

File "<ipython-input-18-2874bd4e8290>", line 2
    my $foo = 'bar';
       ^
SyntaxError: invalid syntax

while a version without a newline works as expected:

%%perl
my $foo = 'bar';
print $foo;

produces:

bar

[takluyver]

Yep, it will affect all cell magics. If you do %hist -t after hitting it, you should see why - it only catches the cell magic code until a blank line, then treats the rest of the cell as regular Python code. I know the general shape of how to fix this, I'm just thinking about how it should fit into our framework.

[ellisonbg]

But I thought that algorithm for detecting cell magics was dead simple. If The cell begins with double percent the rest of the cel is handled as a cell magic, right?

Sent from my iPhone

On Apr 18, 2013, at 1:43 PM, Thomas Kluyver notifications@github.com wrote:

Yep, it will affect all cell magics. If you do %hist -t after hitting it, you should see why - it only catches the cell magic code until a blank line, then treats the rest of the cell as regular Python code. I know the general shape of how to fix this, I'm just thinking about how it should fit into our framework.

—
Reply to this email directly or view it on GitHub.

[takluyver]

Yep, but for the line-based frontends, we decide when a cell magic is complete and ready to execute by looking for a blank line. That also lets you use cell magics in a .ipy script. The code path for 'is this block of input complete' needs to be slightly different from the code path for 'run this complete cell'.

At present, my thinking is to have an attribute of input transformers that indicates whether they're used for completion testing, cell transformation, or both (the default, and what most will use). Then the rule will be that if your input transformer doesn't do both, it must be one of a pair that do both together. Once I've worked out the right API for this, I'll put together a pull request for it.

[takluyver]

See PR #3203