Project

General

Profile

1 21 aaronmk
# XML DOM tree manipulation
2
3 73 aaronmk
import cgi
4
from HTMLParser import HTMLParser
5 3632 aaronmk
import itertools
6 1809 aaronmk
import re
7 21 aaronmk
from xml.dom import Node
8 299 aaronmk
import xml.dom.minidom as minidom
9 21 aaronmk
10 73 aaronmk
import strings
11 331 aaronmk
import util
12 73 aaronmk
13 840 aaronmk
##### Escaping input
14
15 73 aaronmk
def escape(str_):
16
    return strings.to_unicode(cgi.escape(str_, True)).encode('ascii',
17
        'xmlcharrefreplace')
18
19
def unescape(str_): return HTMLParser().unescape(str_)
20
21 1814 aaronmk
##### Names
22
23
def strip_namespace(name):
24
    namespace, sep, base = name.partition(':')
25
    if sep != '': return base
26
    else: return name
27
28 2431 aaronmk
##### Nodes
29
30
def is_node(value): return isinstance(value, Node)
31
32 2008 aaronmk
##### Replacing a node
33
34
def remove(node): node.parentNode.removeChild(node)
35
36
def replace(old, new):
37
    '''@param new Node|None'''
38 3329 aaronmk
    assert old.parentNode != None # not removed from parent tree
39
40 2008 aaronmk
    if new == None: old.parentNode.removeChild(old)
41
    else: old.parentNode.replaceChild(new, old) # note order reversed
42
43
def replace_with_text(node, new):
44 3225 aaronmk
    '''
45
    @return The *new* node
46
    '''
47 4138 aaronmk
    if isinstance(new, bool): new = util.bool2str(new)
48 2008 aaronmk
    if util.is_str(new): new = node.ownerDocument.createTextNode(new)
49
    replace(node, new)
50 3225 aaronmk
    return new
51 2008 aaronmk
52 840 aaronmk
##### Element node contents
53 455 aaronmk
54 840 aaronmk
def is_elem(node): return node.nodeType == Node.ELEMENT_NODE
55
56 453 aaronmk
def is_completely_empty(node): return node.firstChild == None
57 135 aaronmk
58 301 aaronmk
def has_one_child(node):
59
    return node.firstChild != None and node.firstChild.nextSibling == None
60
61 840 aaronmk
def is_simple(node):
62
    '''Whether every child recursively has no more than one child'''
63
    return (not is_elem(node) or is_completely_empty(node)
64
        or (has_one_child(node) and is_simple(node.firstChild)))
65
66 305 aaronmk
class NodeIter:
67
    def __init__(self, node): self.child = node.firstChild
68
69
    def __iter__(self): return self
70
71
    def curr(self):
72
        if self.child != None: return self.child
73
        raise StopIteration
74
75
    def next(self):
76
        child = self.curr()
77
        self.child = self.child.nextSibling
78
        return child
79
80 840 aaronmk
##### Comments
81 455 aaronmk
82 453 aaronmk
def is_comment(node): return node.nodeType == Node.COMMENT_NODE
83
84
def is_empty(node):
85
    for child in NodeIter(node):
86 4038 aaronmk
        if not (is_whitespace(child) or is_comment(child)): return False
87 453 aaronmk
    return True
88
89 1809 aaronmk
def clean_comment(str_):
90
    '''Sanitizes comment node contents. Strips invalid strings.'''
91
    return re.sub(r'-{2,}', r'-', str_) # comments can't contain '--'
92
93
def mk_comment(doc, str_): return doc.createComment(clean_comment(str_))
94
95 840 aaronmk
##### Child nodes that are elements
96 455 aaronmk
97 21 aaronmk
class NodeElemIter:
98
    def __init__(self, node): self.child = node.firstChild
99
100
    def __iter__(self): return self
101
102
    def curr(self):
103
        while self.child != None:
104 298 aaronmk
            if is_elem(self.child): return self.child
105 21 aaronmk
            self.child = self.child.nextSibling
106
        raise StopIteration
107
108
    def next(self):
109
        child = self.curr()
110
        self.child = self.child.nextSibling
111
        return child
112
113
def first_elem(node): return NodeElemIter(node).next()
114
115 450 aaronmk
def has_elems(node):
116
    try: first_elem(node); return True
117
    except StopIteration: return False
118
119 21 aaronmk
class NodeElemReverseIter:
120
    def __init__(self, node): self.child = node.lastChild
121
122
    def __iter__(self): return self
123
124
    def curr(self):
125
        while self.child != None:
126 298 aaronmk
            if is_elem(self.child): return self.child
127 21 aaronmk
            self.child = self.child.previousSibling
128
        raise StopIteration
129
130
    def next(self):
131
        child = self.curr()
132
        self.child = self.child.previousSibling
133
        return child
134
135
def last_elem(node): return NodeElemReverseIter(node).next()
136
137 840 aaronmk
##### Parent nodes
138
139 1017 aaronmk
def parent(node):
140 999 aaronmk
    '''Does not treat the document object as the root node's parent, since it's
141
    not a true element node'''
142 1017 aaronmk
    parent_ = node.parentNode
143
    if parent_ != None and is_elem(parent_): return parent_
144
    else: return None
145
146
class NodeParentIter:
147
    '''See parent() for special treatment of root node.
148
    Note that the first element returned is the current node, not its parent.'''
149 21 aaronmk
    def __init__(self, node): self.node = node
150
151
    def __iter__(self): return self
152
153
    def curr(self):
154 1017 aaronmk
        if self.node != None: return self.node
155
        else: raise StopIteration
156 21 aaronmk
157
    def next(self):
158
        node = self.curr()
159 1017 aaronmk
        self.node = parent(self.node)
160 21 aaronmk
        return node
161
162 840 aaronmk
##### Element nodes containing text
163 661 aaronmk
164 298 aaronmk
def is_text_node(node): return node.nodeType == Node.TEXT_NODE
165
166 301 aaronmk
def is_text(node): return has_one_child(node) and is_text_node(node.firstChild)
167 21 aaronmk
168 2008 aaronmk
def value_node(node):
169 1832 aaronmk
    if is_elem(node):
170
        iter_ = NodeIter(node)
171
        util.skip(iter_, is_comment)
172 2008 aaronmk
        try: return iter_.next()
173 1832 aaronmk
        except StopIteration: return None
174 2008 aaronmk
    else: return node
175 21 aaronmk
176 2008 aaronmk
def value(node):
177 4029 aaronmk
    value_ = util.do_ignore_none(lambda n: n.nodeValue, value_node(node))
178 4034 aaronmk
    if value_ == strings.isspace_none_str: value_ = None # None equiv
179 4029 aaronmk
    return value_
180 2008 aaronmk
181 4031 aaronmk
def is_whitespace(node):
182
    return is_text_node(node) and (node.nodeValue == ''
183
        or node.nodeValue.isspace())
184 1720 aaronmk
185 143 aaronmk
def set_value(node, value):
186 2008 aaronmk
    value_node_ = value_node(node)
187
    if value != None:
188
        if value_node_ != None:
189
            value_node_.nodeValue = value
190
        else:
191
            assert is_elem(node)
192
            node.appendChild(node.ownerDocument.createTextNode(value))
193 2009 aaronmk
    elif value_node_ != None:
194 4029 aaronmk
        if is_elem(node): remove(value_node_)
195
        else:
196 4033 aaronmk
            if is_text_node(node): value = strings.isspace_none_str # None equiv
197 4029 aaronmk
            node.nodeValue = value
198 22 aaronmk
199 86 aaronmk
class NodeTextEntryIter:
200
    def __init__(self, node): self.iter_ = NodeElemIter(node)
201
202
    def __iter__(self): return self
203
204 1176 aaronmk
    def next(self):
205
        entry = self.iter_.next()
206 3632 aaronmk
        if is_empty(entry): value_ = None
207
        elif is_text(entry): value_ = value(entry)
208 839 aaronmk
        else:
209
            assert has_one_child(entry) # TODO: convert to an exception
210
            value_ = entry.firstChild
211
        return (entry.tagName, value_)
212 86 aaronmk
213 792 aaronmk
def is_text_node_entry(val): return util.is_str(val[1])
214
215 3632 aaronmk
def non_empty(iterable):
216
    return itertools.ifilter(lambda i: i[1] != None, iterable)
217
218 792 aaronmk
class TextEntryOnlyIter(util.CheckedIter):
219
    def __init__(self, iterable):
220 3632 aaronmk
        util.CheckedIter.__init__(self, is_text_node_entry, non_empty(iterable))
221 792 aaronmk
222 1836 aaronmk
##### IDs
223
224
def get_id(node):
225
    '''If the node doesn't have an ID, assumes the node itself is the ID.
226
    @return None if the node doesn't have an ID or a value
227
    '''
228
    id_ = node.getAttribute('id')
229
    if id_ != '': return id_
230
    else: return value(node) # assume the node itself is the ID
231
232
def set_id(node, id_): node.setAttribute('id', id_)
233
234 2008 aaronmk
##### Child nodes
235 455 aaronmk
236 135 aaronmk
def set_child(node, name, value):
237
    '''Note: does not remove any existing child of the same name'''
238
    child = node.ownerDocument.createElement(name)
239 143 aaronmk
    set_value(child, value)
240 135 aaronmk
    node.appendChild(child)
241
242 1814 aaronmk
def by_tag_name(node, name, last_only=False, ignore_namespace=False):
243
    '''last_only optimization returns only the last matching node'''
244
    if ignore_namespace: filter_name = strip_namespace
245
    else: filter_name = lambda name: name
246
    name = filter_name(name)
247
248 22 aaronmk
    children = []
249 888 aaronmk
    if last_only: iter_ = NodeElemReverseIter(node)
250
    else: iter_ = NodeElemIter(node)
251
    for child in iter_:
252 1814 aaronmk
        if filter_name(child.tagName) == name:
253 22 aaronmk
            children.append(child)
254
            if last_only: break
255
    return children
256 28 aaronmk
257 3226 aaronmk
def merge(from_, into):
258 3331 aaronmk
    '''Merges two nodes of the same tag name and their newly-adjacent children.
259
    @post The into node is saved; the from_ node is deleted.
260
    '''
261
    if from_ == None or into == None: return # base case
262
    if from_.tagName != into.tagName: return # not mergeable
263
264
    from_first = from_.firstChild # save before merge
265 3226 aaronmk
    for child in NodeIter(from_): into.appendChild(child)
266
    remove(from_)
267
268 3331 aaronmk
    # Recurse
269
    merge(from_first, from_first.previousSibling) # = into.lastChild
270 3226 aaronmk
271 3331 aaronmk
def merge_by_name(root, name):
272
    '''Merges siblings in root with the given name'''
273
    children = by_tag_name(root, name)
274
    child0 = children.pop(0)
275
    for child in children: merge(child, child0)
276
277 840 aaronmk
##### XML documents
278 455 aaronmk
279 133 aaronmk
def create_doc(root='_'):
280 303 aaronmk
    return minidom.getDOMImplementation().createDocument(None, root, None)
281 133 aaronmk
282 840 aaronmk
##### Printing XML
283 455 aaronmk
284 304 aaronmk
prettyxml_config = dict(addindent='    ', newl='\n')
285 331 aaronmk
toprettyxml_config = prettyxml_config.copy()
286
util.rename_key(toprettyxml_config, 'addindent', 'indent')
287 304 aaronmk
288 840 aaronmk
##### minidom modifications
289 455 aaronmk
290 1720 aaronmk
#### Module
291
292 299 aaronmk
minidom._write_data = lambda writer, data: writer.write(escape(data))
293 73 aaronmk
294 305 aaronmk
minidom.Node.__iter__ = lambda self: NodeIter(self)
295
296 857 aaronmk
def __Node_str(self):
297 860 aaronmk
    return strings.remove_extra_newl(self.toprettyxml(**toprettyxml_config))
298 796 aaronmk
minidom.Node.__str__ = __Node_str
299
minidom.Node.__repr__ = __Node_str
300
minidom.Element.__repr__ = __Node_str
301 301 aaronmk
302 1720 aaronmk
#### Node
303 888 aaronmk
304 315 aaronmk
minidom.Node.pop = lambda self: self.removeChild(self.lastChild)
305
306
def __Node_clear(self):
307
    while not is_empty(self): self.pop()
308
minidom.Node.clear = __Node_clear
309
310 1720 aaronmk
#### Text
311
312
__Text_writexml_orig = minidom.Text.writexml
313
def __Text_writexml(self, *args, **kw_args):
314
    if is_whitespace(self): pass # we add our own whitespace
315
    else: __Text_writexml_orig(self, *args, **kw_args)
316
minidom.Text.writexml = __Text_writexml
317
318
#### Attr
319
320
def __Attr_str(self): return escape(self.name)+'="'+escape(self.value)+'"'
321
minidom.Attr.__str__ = __Attr_str
322
minidom.Attr.__repr__ = __Attr_str
323
324
#### Element
325
326 301 aaronmk
def __Element_write_opening(self, writer, indent='', addindent='', newl=''):
327 1563 aaronmk
    writer.write(indent+'<'+escape(self.tagName))
328 298 aaronmk
    for attr_idx in xrange(self.attributes.length):
329 891 aaronmk
        writer.write(' '+str(self.attributes.item(attr_idx)))
330 298 aaronmk
    writer.write('>'+newl)
331 301 aaronmk
minidom.Element.write_opening = __Element_write_opening
332 73 aaronmk
333 301 aaronmk
def __Element_write_closing(self, writer, indent='', addindent='', newl=''):
334 1563 aaronmk
    writer.write('</'+escape(self.tagName)+'>'+newl)
335 301 aaronmk
minidom.Element.write_closing = __Element_write_closing
336 298 aaronmk
337 1720 aaronmk
__Element_writexml_orig = minidom.Element.writexml
338 301 aaronmk
def __Element_writexml(self, writer, indent='', addindent='', newl=''):
339 306 aaronmk
    if isinstance(indent, int): indent = addindent*indent
340 774 aaronmk
    if is_simple(self):
341
        writer.write(indent)
342 1720 aaronmk
        __Element_writexml_orig(self, writer)
343 774 aaronmk
        writer.write(newl)
344 1720 aaronmk
    else: __Element_writexml_orig(self, writer, indent, addindent, newl)
345 301 aaronmk
minidom.Element.writexml = __Element_writexml
346 28 aaronmk
347 1720 aaronmk
#### Document
348
349 301 aaronmk
def __Document_write_opening(self, writer, indent='', addindent='', newl='',
350
    encoding=None):
351
    xmlDecl = '<?xml version="1.0" '
352
    if encoding != None: xmlDecl += 'encoding="'+escape(encoding)+'"'
353
    xmlDecl += '?>'+newl
354
    writer.write(xmlDecl)
355
    assert has_one_child(self)
356
    assert is_elem(self.firstChild)
357
    self.firstChild.write_opening(writer, indent, addindent, newl)
358
minidom.Document.write_opening = __Document_write_opening
359
360
def __Document_write_closing(self, writer, indent='', addindent='', newl=''):
361
    self.firstChild.write_closing(writer, indent, addindent, newl)
362
minidom.Document.write_closing = __Document_write_closing