##// END OF EJS Templates
Ensure handler patterns are str, not unicode...
Ensure handler patterns are str, not unicode Python < 2.6.5 doesn't accept unicode keys in f(**kwargs), and base_project_url will always be unicode, which will in turn make the patterns unicode, and ultimately result in unicode keys in kwargs to handler._execute(**kwargs) in tornado. This enforces that base_project_url be ascii in that situation. Note that the URLs these patterns check against are escaped, and thus guaranteed to be ASCII: 'héllo' is really 'h%C3%A9llo'. If you actually use u'héllo' in your regex, it will not match the URLs you think it should.

File last commit:

r5877:38534e1a
r6067:24a6c071
Show More
notebookmanager.py
274 lines | 10.4 KiB | text/x-python | PythonLexer
"""A notebook manager that uses the local file system for storage.
Authors:
* Brian Granger
"""
#-----------------------------------------------------------------------------
# Copyright (C) 2008-2011 The IPython Development Team
#
# Distributed under the terms of the BSD License. The full license is in
# the file COPYING, distributed as part of this software.
#-----------------------------------------------------------------------------
#-----------------------------------------------------------------------------
# Imports
#-----------------------------------------------------------------------------
import datetime
import os
import uuid
import glob
from tornado import web
from IPython.config.configurable import LoggingConfigurable
from IPython.nbformat import current
from IPython.utils.traitlets import Unicode, List, Dict, Bool
#-----------------------------------------------------------------------------
# Classes
#-----------------------------------------------------------------------------
class NotebookManager(LoggingConfigurable):
notebook_dir = Unicode(os.getcwd(), config=True, help="""
The directory to use for notebooks.
""")
save_script = Bool(False, config=True,
help="""Automatically create a Python script when saving the notebook.
For easier use of import, %run and %loadpy across notebooks, a
<notebook-name>.py script will be created next to any
<notebook-name>.ipynb on each save. This can also be set with the
short `--script` flag.
"""
)
filename_ext = Unicode(u'.ipynb')
allowed_formats = List([u'json',u'py'])
# Map notebook_ids to notebook names
mapping = Dict()
# Map notebook names to notebook_ids
rev_mapping = Dict()
def list_notebooks(self):
"""List all notebooks in the notebook dir.
This returns a list of dicts of the form::
dict(notebook_id=notebook,name=name)
"""
names = glob.glob(os.path.join(self.notebook_dir,
'*' + self.filename_ext))
names = [os.path.splitext(os.path.basename(name))[0]
for name in names]
data = []
for name in names:
if name not in self.rev_mapping:
notebook_id = self.new_notebook_id(name)
else:
notebook_id = self.rev_mapping[name]
data.append(dict(notebook_id=notebook_id,name=name))
data = sorted(data, key=lambda item: item['name'])
return data
def new_notebook_id(self, name):
"""Generate a new notebook_id for a name and store its mappings."""
# TODO: the following will give stable urls for notebooks, but unless
# the notebooks are immediately redirected to their new urls when their
# filemname changes, nasty inconsistencies result. So for now it's
# disabled and instead we use a random uuid4() call. But we leave the
# logic here so that we can later reactivate it, whhen the necessary
# url redirection code is written.
#notebook_id = unicode(uuid.uuid5(uuid.NAMESPACE_URL,
# 'file://'+self.get_path_by_name(name).encode('utf-8')))
notebook_id = unicode(uuid.uuid4())
self.mapping[notebook_id] = name
self.rev_mapping[name] = notebook_id
return notebook_id
def delete_notebook_id(self, notebook_id):
"""Delete a notebook's id only. This doesn't delete the actual notebook."""
name = self.mapping[notebook_id]
del self.mapping[notebook_id]
del self.rev_mapping[name]
def notebook_exists(self, notebook_id):
"""Does a notebook exist?"""
if notebook_id not in self.mapping:
return False
path = self.get_path_by_name(self.mapping[notebook_id])
return os.path.isfile(path)
def find_path(self, notebook_id):
"""Return a full path to a notebook given its notebook_id."""
try:
name = self.mapping[notebook_id]
except KeyError:
raise web.HTTPError(404, u'Notebook does not exist: %s' % notebook_id)
return self.get_path_by_name(name)
def get_path_by_name(self, name):
"""Return a full path to a notebook given its name."""
filename = name + self.filename_ext
path = os.path.join(self.notebook_dir, filename)
return path
def get_notebook(self, notebook_id, format=u'json'):
"""Get the representation of a notebook in format by notebook_id."""
format = unicode(format)
if format not in self.allowed_formats:
raise web.HTTPError(415, u'Invalid notebook format: %s' % format)
last_modified, nb = self.get_notebook_object(notebook_id)
kwargs = {}
if format == 'json':
# don't split lines for sending over the wire, because it
# should match the Python in-memory format.
kwargs['split_lines'] = False
data = current.writes(nb, format, **kwargs)
name = nb.get('name','notebook')
return last_modified, name, data
def get_notebook_object(self, notebook_id):
"""Get the NotebookNode representation of a notebook by notebook_id."""
path = self.find_path(notebook_id)
if not os.path.isfile(path):
raise web.HTTPError(404, u'Notebook does not exist: %s' % notebook_id)
info = os.stat(path)
last_modified = datetime.datetime.utcfromtimestamp(info.st_mtime)
with open(path,'r') as f:
s = f.read()
try:
# v1 and v2 and json in the .ipynb files.
nb = current.reads(s, u'json')
except:
raise web.HTTPError(500, u'Unreadable JSON notebook.')
if 'name' not in nb:
nb.name = os.path.split(path)[-1].split(u'.')[0]
return last_modified, nb
def save_new_notebook(self, data, name=None, format=u'json'):
"""Save a new notebook and return its notebook_id.
If a name is passed in, it overrides any values in the notebook data
and the value in the data is updated to use that value.
"""
if format not in self.allowed_formats:
raise web.HTTPError(415, u'Invalid notebook format: %s' % format)
try:
nb = current.reads(data.decode('utf-8'), format)
except:
raise web.HTTPError(400, u'Invalid JSON data')
if name is None:
try:
name = nb.metadata.name
except AttributeError:
raise web.HTTPError(400, u'Missing notebook name')
nb.metadata.name = name
notebook_id = self.new_notebook_id(name)
self.save_notebook_object(notebook_id, nb)
return notebook_id
def save_notebook(self, notebook_id, data, name=None, format=u'json'):
"""Save an existing notebook by notebook_id."""
if format not in self.allowed_formats:
raise web.HTTPError(415, u'Invalid notebook format: %s' % format)
try:
nb = current.reads(data.decode('utf-8'), format)
except:
raise web.HTTPError(400, u'Invalid JSON data')
if name is not None:
nb.metadata.name = name
self.save_notebook_object(notebook_id, nb)
def save_notebook_object(self, notebook_id, nb):
"""Save an existing notebook object by notebook_id."""
if notebook_id not in self.mapping:
raise web.HTTPError(404, u'Notebook does not exist: %s' % notebook_id)
old_name = self.mapping[notebook_id]
try:
new_name = nb.metadata.name
except AttributeError:
raise web.HTTPError(400, u'Missing notebook name')
path = self.get_path_by_name(new_name)
try:
with open(path,'w') as f:
current.write(nb, f, u'json')
except Exception as e:
raise web.HTTPError(400, u'Unexpected error while saving notebook: %s' % e)
# save .py script as well
if self.save_script:
pypath = os.path.splitext(path)[0] + '.py'
try:
with open(pypath,'w') as f:
current.write(nb, f, u'py')
except Exception as e:
raise web.HTTPError(400, u'Unexpected error while saving notebook as script: %s' % e)
if old_name != new_name:
old_path = self.get_path_by_name(old_name)
if os.path.isfile(old_path):
os.unlink(old_path)
if self.save_script:
old_pypath = os.path.splitext(old_path)[0] + '.py'
if os.path.isfile(old_pypath):
os.unlink(old_pypath)
self.mapping[notebook_id] = new_name
self.rev_mapping[new_name] = notebook_id
def delete_notebook(self, notebook_id):
"""Delete notebook by notebook_id."""
path = self.find_path(notebook_id)
if not os.path.isfile(path):
raise web.HTTPError(404, u'Notebook does not exist: %s' % notebook_id)
os.unlink(path)
self.delete_notebook_id(notebook_id)
def increment_filename(self, basename):
"""Return a non-used filename of the form basename<int>.
This searches through the filenames (basename0, basename1, ...)
until is find one that is not already being used. It is used to
create Untitled and Copy names that are unique.
"""
i = 0
while True:
name = u'%s%i' % (basename,i)
path = self.get_path_by_name(name)
if not os.path.isfile(path):
break
else:
i = i+1
return path, name
def new_notebook(self):
"""Create a new notebook and return its notebook_id."""
path, name = self.increment_filename('Untitled')
notebook_id = self.new_notebook_id(name)
metadata = current.new_metadata(name=name)
nb = current.new_notebook(metadata=metadata)
with open(path,'w') as f:
current.write(nb, f, u'json')
return notebook_id
def copy_notebook(self, notebook_id):
"""Copy an existing notebook and return its notebook_id."""
last_mod, nb = self.get_notebook_object(notebook_id)
name = nb.metadata.name + '-Copy'
path, name = self.increment_filename(name)
nb.metadata.name = name
notebook_id = self.new_notebook_id(name)
self.save_notebook_object(notebook_id, nb)
return notebook_id