##// END OF EJS Templates
Added pipes.quote for batch system lauchers. Fixes bug where LSF would not launch engines or controller.
Added pipes.quote for batch system lauchers. Fixes bug where LSF would not launch engines or controller.

File last commit:

r4649:5dc04396
r7845:3b8d8385
Show More
nbxml.py
188 lines | 6.6 KiB | text/x-python | PythonLexer
"""Read and write notebook files as XML.
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
#-----------------------------------------------------------------------------
from base64 import encodestring, decodestring
import warnings
from xml.etree import ElementTree as ET
from .rwbase import NotebookReader, NotebookWriter
from .nbbase import (
new_code_cell, new_text_cell, new_worksheet, new_notebook, new_output,
new_metadata
)
#-----------------------------------------------------------------------------
# Code
#-----------------------------------------------------------------------------
def indent(elem, level=0):
i = "\n" + level*" "
if len(elem):
if not elem.text or not elem.text.strip():
elem.text = i + " "
if not elem.tail or not elem.tail.strip():
elem.tail = i
for elem in elem:
indent(elem, level+1)
if not elem.tail or not elem.tail.strip():
elem.tail = i
else:
if level and (not elem.tail or not elem.tail.strip()):
elem.tail = i
def _get_text(e, tag):
sub_e = e.find(tag)
if sub_e is None:
return None
else:
return sub_e.text
def _set_text(nbnode, attr, parent, tag):
if attr in nbnode:
e = ET.SubElement(parent, tag)
e.text = nbnode[attr]
def _get_int(e, tag):
sub_e = e.find(tag)
if sub_e is None:
return None
else:
return int(sub_e.text)
def _set_int(nbnode, attr, parent, tag):
if attr in nbnode:
e = ET.SubElement(parent, tag)
e.text = unicode(nbnode[attr])
def _get_bool(e, tag):
sub_e = e.find(tag)
if sub_e is None:
return None
else:
return bool(int(sub_e.text))
def _set_bool(nbnode, attr, parent, tag):
if attr in nbnode:
e = ET.SubElement(parent, tag)
if nbnode[attr]:
e.text = u'1'
else:
e.text = u'0'
def _get_binary(e, tag):
sub_e = e.find(tag)
if sub_e is None:
return None
else:
return decodestring(sub_e.text)
def _set_binary(nbnode, attr, parent, tag):
if attr in nbnode:
e = ET.SubElement(parent, tag)
e.text = encodestring(nbnode[attr])
class XMLReader(NotebookReader):
def reads(self, s, **kwargs):
root = ET.fromstring(s)
return self.to_notebook(root, **kwargs)
def to_notebook(self, root, **kwargs):
warnings.warn('The XML notebook format is no longer supported, '
'please convert your notebooks to JSON.', DeprecationWarning)
nbname = _get_text(root,u'name')
nbauthor = _get_text(root,u'author')
nbemail = _get_text(root,u'email')
nblicense = _get_text(root,u'license')
nbcreated = _get_text(root,u'created')
nbsaved = _get_text(root,u'saved')
worksheets = []
for ws_e in root.find(u'worksheets').getiterator(u'worksheet'):
wsname = _get_text(ws_e,u'name')
cells = []
for cell_e in ws_e.find(u'cells').getiterator():
if cell_e.tag == u'codecell':
input = _get_text(cell_e,u'input')
prompt_number = _get_int(cell_e,u'prompt_number')
collapsed = _get_bool(cell_e,u'collapsed')
language = _get_text(cell_e,u'language')
outputs = []
for output_e in cell_e.find(u'outputs').getiterator(u'output'):
output_type = _get_text(output_e,u'output_type')
output_text = _get_text(output_e,u'text')
output_png = _get_binary(output_e,u'png')
output_jpeg = _get_binary(output_e,u'jpeg')
output_svg = _get_text(output_e,u'svg')
output_html = _get_text(output_e,u'html')
output_latex = _get_text(output_e,u'latex')
output_json = _get_text(output_e,u'json')
output_javascript = _get_text(output_e,u'javascript')
out_prompt_number = _get_int(output_e,u'prompt_number')
etype = _get_text(output_e,u'etype')
evalue = _get_text(output_e,u'evalue')
traceback = []
traceback_e = output_e.find(u'traceback')
if traceback_e is not None:
for frame_e in traceback_e.getiterator(u'frame'):
traceback.append(frame_e.text)
if len(traceback) == 0:
traceback = None
output = new_output(output_type=output_type,output_png=output_png,
output_text=output_text, output_svg=output_svg,
output_html=output_html, output_latex=output_latex,
output_json=output_json, output_javascript=output_javascript,
output_jpeg=output_jpeg, prompt_number=out_prompt_number,
etype=etype, evalue=evalue, traceback=traceback
)
outputs.append(output)
cc = new_code_cell(input=input,prompt_number=prompt_number,
language=language,outputs=outputs,collapsed=collapsed)
cells.append(cc)
if cell_e.tag == u'htmlcell':
source = _get_text(cell_e,u'source')
rendered = _get_text(cell_e,u'rendered')
cells.append(new_text_cell(u'html', source=source, rendered=rendered))
if cell_e.tag == u'markdowncell':
source = _get_text(cell_e,u'source')
rendered = _get_text(cell_e,u'rendered')
cells.append(new_text_cell(u'markdown', source=source, rendered=rendered))
ws = new_worksheet(name=wsname,cells=cells)
worksheets.append(ws)
md = new_metadata(name=nbname)
nb = new_notebook(metadata=md,worksheets=worksheets)
return nb
_reader = XMLReader()
reads = _reader.reads
read = _reader.read
to_notebook = _reader.to_notebook