| 1 |
21
|
aaronmk
|
# XML DOM tree manipulation
|
| 2 |
|
|
|
| 3 |
73
|
aaronmk
|
import cgi
|
| 4 |
|
|
from HTMLParser import HTMLParser
|
| 5 |
1809
|
aaronmk
|
import re
|
| 6 |
21
|
aaronmk
|
from xml.dom import Node
|
| 7 |
299
|
aaronmk
|
import xml.dom.minidom as minidom
|
| 8 |
21
|
aaronmk
|
|
| 9 |
73
|
aaronmk
|
import strings
|
| 10 |
331
|
aaronmk
|
import util
|
| 11 |
73
|
aaronmk
|
|
| 12 |
840
|
aaronmk
|
##### Escaping input
|
| 13 |
|
|
|
| 14 |
73
|
aaronmk
|
def escape(str_):
|
| 15 |
|
|
return strings.to_unicode(cgi.escape(str_, True)).encode('ascii',
|
| 16 |
|
|
'xmlcharrefreplace')
|
| 17 |
|
|
|
| 18 |
|
|
def unescape(str_): return HTMLParser().unescape(str_)
|
| 19 |
|
|
|
| 20 |
840
|
aaronmk
|
##### IDs
|
| 21 |
455
|
aaronmk
|
|
| 22 |
21
|
aaronmk
|
def get_id(node): return node.getAttribute('id')
|
| 23 |
|
|
|
| 24 |
|
|
def set_id(node, id_): node.setAttribute('id', id_)
|
| 25 |
|
|
|
| 26 |
840
|
aaronmk
|
##### Element node contents
|
| 27 |
455
|
aaronmk
|
|
| 28 |
840
|
aaronmk
|
def is_elem(node): return node.nodeType == Node.ELEMENT_NODE
|
| 29 |
|
|
|
| 30 |
453
|
aaronmk
|
def is_completely_empty(node): return node.firstChild == None
|
| 31 |
135
|
aaronmk
|
|
| 32 |
301
|
aaronmk
|
def has_one_child(node):
|
| 33 |
|
|
return node.firstChild != None and node.firstChild.nextSibling == None
|
| 34 |
|
|
|
| 35 |
840
|
aaronmk
|
def is_simple(node):
|
| 36 |
|
|
'''Whether every child recursively has no more than one child'''
|
| 37 |
|
|
return (not is_elem(node) or is_completely_empty(node)
|
| 38 |
|
|
or (has_one_child(node) and is_simple(node.firstChild)))
|
| 39 |
|
|
|
| 40 |
305
|
aaronmk
|
class NodeIter:
|
| 41 |
|
|
def __init__(self, node): self.child = node.firstChild
|
| 42 |
|
|
|
| 43 |
|
|
def __iter__(self): return self
|
| 44 |
|
|
|
| 45 |
|
|
def curr(self):
|
| 46 |
|
|
if self.child != None: return self.child
|
| 47 |
|
|
raise StopIteration
|
| 48 |
|
|
|
| 49 |
|
|
def next(self):
|
| 50 |
|
|
child = self.curr()
|
| 51 |
|
|
self.child = self.child.nextSibling
|
| 52 |
|
|
return child
|
| 53 |
|
|
|
| 54 |
840
|
aaronmk
|
##### Comments
|
| 55 |
455
|
aaronmk
|
|
| 56 |
453
|
aaronmk
|
def is_comment(node): return node.nodeType == Node.COMMENT_NODE
|
| 57 |
|
|
|
| 58 |
|
|
def is_empty(node):
|
| 59 |
|
|
for child in NodeIter(node):
|
| 60 |
|
|
if not is_comment(child): return False
|
| 61 |
|
|
return True
|
| 62 |
|
|
|
| 63 |
1809
|
aaronmk
|
def clean_comment(str_):
|
| 64 |
|
|
'''Sanitizes comment node contents. Strips invalid strings.'''
|
| 65 |
|
|
return re.sub(r'-{2,}', r'-', str_) # comments can't contain '--'
|
| 66 |
|
|
|
| 67 |
|
|
def mk_comment(doc, str_): return doc.createComment(clean_comment(str_))
|
| 68 |
|
|
|
| 69 |
840
|
aaronmk
|
##### Child nodes that are elements
|
| 70 |
455
|
aaronmk
|
|
| 71 |
21
|
aaronmk
|
class NodeElemIter:
|
| 72 |
|
|
def __init__(self, node): self.child = node.firstChild
|
| 73 |
|
|
|
| 74 |
|
|
def __iter__(self): return self
|
| 75 |
|
|
|
| 76 |
|
|
def curr(self):
|
| 77 |
|
|
while self.child != None:
|
| 78 |
298
|
aaronmk
|
if is_elem(self.child): return self.child
|
| 79 |
21
|
aaronmk
|
self.child = self.child.nextSibling
|
| 80 |
|
|
raise StopIteration
|
| 81 |
|
|
|
| 82 |
|
|
def next(self):
|
| 83 |
|
|
child = self.curr()
|
| 84 |
|
|
self.child = self.child.nextSibling
|
| 85 |
|
|
return child
|
| 86 |
|
|
|
| 87 |
|
|
def first_elem(node): return NodeElemIter(node).next()
|
| 88 |
|
|
|
| 89 |
450
|
aaronmk
|
def has_elems(node):
|
| 90 |
|
|
try: first_elem(node); return True
|
| 91 |
|
|
except StopIteration: return False
|
| 92 |
|
|
|
| 93 |
21
|
aaronmk
|
class NodeElemReverseIter:
|
| 94 |
|
|
def __init__(self, node): self.child = node.lastChild
|
| 95 |
|
|
|
| 96 |
|
|
def __iter__(self): return self
|
| 97 |
|
|
|
| 98 |
|
|
def curr(self):
|
| 99 |
|
|
while self.child != None:
|
| 100 |
298
|
aaronmk
|
if is_elem(self.child): return self.child
|
| 101 |
21
|
aaronmk
|
self.child = self.child.previousSibling
|
| 102 |
|
|
raise StopIteration
|
| 103 |
|
|
|
| 104 |
|
|
def next(self):
|
| 105 |
|
|
child = self.curr()
|
| 106 |
|
|
self.child = self.child.previousSibling
|
| 107 |
|
|
return child
|
| 108 |
|
|
|
| 109 |
|
|
def last_elem(node): return NodeElemReverseIter(node).next()
|
| 110 |
|
|
|
| 111 |
840
|
aaronmk
|
##### Parent nodes
|
| 112 |
|
|
|
| 113 |
1017
|
aaronmk
|
def parent(node):
|
| 114 |
999
|
aaronmk
|
'''Does not treat the document object as the root node's parent, since it's
|
| 115 |
|
|
not a true element node'''
|
| 116 |
1017
|
aaronmk
|
parent_ = node.parentNode
|
| 117 |
|
|
if parent_ != None and is_elem(parent_): return parent_
|
| 118 |
|
|
else: return None
|
| 119 |
|
|
|
| 120 |
|
|
class NodeParentIter:
|
| 121 |
|
|
'''See parent() for special treatment of root node.
|
| 122 |
|
|
Note that the first element returned is the current node, not its parent.'''
|
| 123 |
21
|
aaronmk
|
def __init__(self, node): self.node = node
|
| 124 |
|
|
|
| 125 |
|
|
def __iter__(self): return self
|
| 126 |
|
|
|
| 127 |
|
|
def curr(self):
|
| 128 |
1017
|
aaronmk
|
if self.node != None: return self.node
|
| 129 |
|
|
else: raise StopIteration
|
| 130 |
21
|
aaronmk
|
|
| 131 |
|
|
def next(self):
|
| 132 |
|
|
node = self.curr()
|
| 133 |
1017
|
aaronmk
|
self.node = parent(self.node)
|
| 134 |
21
|
aaronmk
|
return node
|
| 135 |
|
|
|
| 136 |
840
|
aaronmk
|
##### Element nodes containing text
|
| 137 |
661
|
aaronmk
|
|
| 138 |
298
|
aaronmk
|
def is_text_node(node): return node.nodeType == Node.TEXT_NODE
|
| 139 |
|
|
|
| 140 |
301
|
aaronmk
|
def is_text(node): return has_one_child(node) and is_text_node(node.firstChild)
|
| 141 |
21
|
aaronmk
|
|
| 142 |
|
|
def value(node):
|
| 143 |
29
|
aaronmk
|
if node.firstChild != None: return node.firstChild.nodeValue
|
| 144 |
21
|
aaronmk
|
else: return node.nodeValue
|
| 145 |
|
|
|
| 146 |
1720
|
aaronmk
|
def is_whitespace(node): return is_text_node(node) and value(node).isspace()
|
| 147 |
|
|
|
| 148 |
143
|
aaronmk
|
def set_value(node, value):
|
| 149 |
298
|
aaronmk
|
if is_elem(node): node.appendChild(node.ownerDocument.createTextNode(value))
|
| 150 |
22
|
aaronmk
|
else: node.nodeValue = value
|
| 151 |
|
|
|
| 152 |
86
|
aaronmk
|
class NodeTextEntryIter:
|
| 153 |
|
|
def __init__(self, node): self.iter_ = NodeElemIter(node)
|
| 154 |
|
|
|
| 155 |
|
|
def __iter__(self): return self
|
| 156 |
|
|
|
| 157 |
1176
|
aaronmk
|
def next(self):
|
| 158 |
|
|
util.skip(self.iter_, is_empty)
|
| 159 |
|
|
entry = self.iter_.next()
|
| 160 |
839
|
aaronmk
|
if is_text(entry): value_ = value(entry)
|
| 161 |
|
|
else:
|
| 162 |
|
|
assert has_one_child(entry) # TODO: convert to an exception
|
| 163 |
|
|
value_ = entry.firstChild
|
| 164 |
|
|
return (entry.tagName, value_)
|
| 165 |
86
|
aaronmk
|
|
| 166 |
792
|
aaronmk
|
def is_text_node_entry(val): return util.is_str(val[1])
|
| 167 |
|
|
|
| 168 |
|
|
class TextEntryOnlyIter(util.CheckedIter):
|
| 169 |
|
|
def __init__(self, iterable):
|
| 170 |
|
|
util.CheckedIter.__init__(self, is_text_node_entry, iterable)
|
| 171 |
|
|
|
| 172 |
840
|
aaronmk
|
##### Modifying/replacing a node
|
| 173 |
455
|
aaronmk
|
|
| 174 |
135
|
aaronmk
|
def set_child(node, name, value):
|
| 175 |
|
|
'''Note: does not remove any existing child of the same name'''
|
| 176 |
|
|
child = node.ownerDocument.createElement(name)
|
| 177 |
143
|
aaronmk
|
set_value(child, value)
|
| 178 |
135
|
aaronmk
|
node.appendChild(child)
|
| 179 |
|
|
|
| 180 |
435
|
aaronmk
|
def remove(node): node.parentNode.removeChild(node)
|
| 181 |
|
|
|
| 182 |
963
|
aaronmk
|
def replace(old, new):
|
| 183 |
|
|
'''@param new Node|None'''
|
| 184 |
|
|
if new == None: old.parentNode.removeChild(old)
|
| 185 |
|
|
else: old.parentNode.replaceChild(new, old) # note order reversed
|
| 186 |
86
|
aaronmk
|
|
| 187 |
757
|
aaronmk
|
def replace_with_text(node, new):
|
| 188 |
792
|
aaronmk
|
if util.is_str(new): new = node.ownerDocument.createTextNode(new)
|
| 189 |
757
|
aaronmk
|
replace(node, new)
|
| 190 |
86
|
aaronmk
|
|
| 191 |
840
|
aaronmk
|
##### Searching child nodes
|
| 192 |
455
|
aaronmk
|
|
| 193 |
22
|
aaronmk
|
def by_tag_name(node, name, last_only=False):
|
| 194 |
135
|
aaronmk
|
'''last_only optimization returns last matching node'''
|
| 195 |
22
|
aaronmk
|
children = []
|
| 196 |
888
|
aaronmk
|
if last_only: iter_ = NodeElemReverseIter(node)
|
| 197 |
|
|
else: iter_ = NodeElemIter(node)
|
| 198 |
|
|
for child in iter_:
|
| 199 |
22
|
aaronmk
|
if child.tagName == name:
|
| 200 |
|
|
children.append(child)
|
| 201 |
|
|
if last_only: break
|
| 202 |
|
|
return children
|
| 203 |
28
|
aaronmk
|
|
| 204 |
840
|
aaronmk
|
##### XML documents
|
| 205 |
455
|
aaronmk
|
|
| 206 |
133
|
aaronmk
|
def create_doc(root='_'):
|
| 207 |
303
|
aaronmk
|
return minidom.getDOMImplementation().createDocument(None, root, None)
|
| 208 |
133
|
aaronmk
|
|
| 209 |
840
|
aaronmk
|
##### Printing XML
|
| 210 |
455
|
aaronmk
|
|
| 211 |
304
|
aaronmk
|
prettyxml_config = dict(addindent=' ', newl='\n')
|
| 212 |
331
|
aaronmk
|
toprettyxml_config = prettyxml_config.copy()
|
| 213 |
|
|
util.rename_key(toprettyxml_config, 'addindent', 'indent')
|
| 214 |
304
|
aaronmk
|
|
| 215 |
840
|
aaronmk
|
##### minidom modifications
|
| 216 |
455
|
aaronmk
|
|
| 217 |
1720
|
aaronmk
|
#### Module
|
| 218 |
|
|
|
| 219 |
299
|
aaronmk
|
minidom._write_data = lambda writer, data: writer.write(escape(data))
|
| 220 |
73
|
aaronmk
|
|
| 221 |
305
|
aaronmk
|
minidom.Node.__iter__ = lambda self: NodeIter(self)
|
| 222 |
|
|
|
| 223 |
857
|
aaronmk
|
def __Node_str(self):
|
| 224 |
860
|
aaronmk
|
return strings.remove_extra_newl(self.toprettyxml(**toprettyxml_config))
|
| 225 |
796
|
aaronmk
|
minidom.Node.__str__ = __Node_str
|
| 226 |
|
|
minidom.Node.__repr__ = __Node_str
|
| 227 |
|
|
minidom.Element.__repr__ = __Node_str
|
| 228 |
301
|
aaronmk
|
|
| 229 |
1720
|
aaronmk
|
#### Node
|
| 230 |
888
|
aaronmk
|
|
| 231 |
315
|
aaronmk
|
minidom.Node.pop = lambda self: self.removeChild(self.lastChild)
|
| 232 |
|
|
|
| 233 |
|
|
def __Node_clear(self):
|
| 234 |
|
|
while not is_empty(self): self.pop()
|
| 235 |
|
|
minidom.Node.clear = __Node_clear
|
| 236 |
|
|
|
| 237 |
1720
|
aaronmk
|
#### Text
|
| 238 |
|
|
|
| 239 |
|
|
__Text_writexml_orig = minidom.Text.writexml
|
| 240 |
|
|
def __Text_writexml(self, *args, **kw_args):
|
| 241 |
|
|
if is_whitespace(self): pass # we add our own whitespace
|
| 242 |
|
|
else: __Text_writexml_orig(self, *args, **kw_args)
|
| 243 |
|
|
minidom.Text.writexml = __Text_writexml
|
| 244 |
|
|
|
| 245 |
|
|
#### Attr
|
| 246 |
|
|
|
| 247 |
|
|
def __Attr_str(self): return escape(self.name)+'="'+escape(self.value)+'"'
|
| 248 |
|
|
minidom.Attr.__str__ = __Attr_str
|
| 249 |
|
|
minidom.Attr.__repr__ = __Attr_str
|
| 250 |
|
|
|
| 251 |
|
|
#### Element
|
| 252 |
|
|
|
| 253 |
301
|
aaronmk
|
def __Element_write_opening(self, writer, indent='', addindent='', newl=''):
|
| 254 |
1563
|
aaronmk
|
writer.write(indent+'<'+escape(self.tagName))
|
| 255 |
298
|
aaronmk
|
for attr_idx in xrange(self.attributes.length):
|
| 256 |
891
|
aaronmk
|
writer.write(' '+str(self.attributes.item(attr_idx)))
|
| 257 |
298
|
aaronmk
|
writer.write('>'+newl)
|
| 258 |
301
|
aaronmk
|
minidom.Element.write_opening = __Element_write_opening
|
| 259 |
73
|
aaronmk
|
|
| 260 |
301
|
aaronmk
|
def __Element_write_closing(self, writer, indent='', addindent='', newl=''):
|
| 261 |
1563
|
aaronmk
|
writer.write('</'+escape(self.tagName)+'>'+newl)
|
| 262 |
301
|
aaronmk
|
minidom.Element.write_closing = __Element_write_closing
|
| 263 |
298
|
aaronmk
|
|
| 264 |
1720
|
aaronmk
|
__Element_writexml_orig = minidom.Element.writexml
|
| 265 |
301
|
aaronmk
|
def __Element_writexml(self, writer, indent='', addindent='', newl=''):
|
| 266 |
306
|
aaronmk
|
if isinstance(indent, int): indent = addindent*indent
|
| 267 |
774
|
aaronmk
|
if is_simple(self):
|
| 268 |
|
|
writer.write(indent)
|
| 269 |
1720
|
aaronmk
|
__Element_writexml_orig(self, writer)
|
| 270 |
774
|
aaronmk
|
writer.write(newl)
|
| 271 |
1720
|
aaronmk
|
else: __Element_writexml_orig(self, writer, indent, addindent, newl)
|
| 272 |
301
|
aaronmk
|
minidom.Element.writexml = __Element_writexml
|
| 273 |
28
|
aaronmk
|
|
| 274 |
1720
|
aaronmk
|
#### Document
|
| 275 |
|
|
|
| 276 |
301
|
aaronmk
|
def __Document_write_opening(self, writer, indent='', addindent='', newl='',
|
| 277 |
|
|
encoding=None):
|
| 278 |
|
|
xmlDecl = '<?xml version="1.0" '
|
| 279 |
|
|
if encoding != None: xmlDecl += 'encoding="'+escape(encoding)+'"'
|
| 280 |
|
|
xmlDecl += '?>'+newl
|
| 281 |
|
|
writer.write(xmlDecl)
|
| 282 |
|
|
assert has_one_child(self)
|
| 283 |
|
|
assert is_elem(self.firstChild)
|
| 284 |
|
|
self.firstChild.write_opening(writer, indent, addindent, newl)
|
| 285 |
|
|
minidom.Document.write_opening = __Document_write_opening
|
| 286 |
|
|
|
| 287 |
|
|
def __Document_write_closing(self, writer, indent='', addindent='', newl=''):
|
| 288 |
|
|
self.firstChild.write_closing(writer, indent, addindent, newl)
|
| 289 |
|
|
minidom.Document.write_closing = __Document_write_closing
|
