Mercurial > traipse_dev
view upmana/xmltramp.py @ 151:06f10429eedc alpha
Traipse Alpha 'OpenRPG' {091125-00}
Traipse is a distribution of OpenRPG that is designed to be easy to setup and go.
Traipse also makes it easy for developers to work on code without fear of
sacrifice. 'Ornery-Orc' continues the trend of 'Grumpy' and adds fixes to the
code. 'Ornery-Orc's main goal is to offer more advanced features and enhance the
productivity of the user.
Update Summary (Cleaning up for Beta)
Added Bookmarks
Fix to Remote Admin Commands
Minor fix to text based Server
Fix to Pretty Print, from Core
Fix to Splitter Nodes not being created
Fix to massive amounts of images loading, from Core
Added 'boot' command to remote admin
Added confirmation window for sent nodes
Minor changes to allow for portability to an OpenSUSE linux OS
Miniatures Layer pop up box allows users to turn off Mini labels, from FlexiRPG
Zoom Mouse plugin added
Images added to Plugin UI
Switching to Element Tree
Map efficiency, from FlexiRPG
Added Status Bar to Update Manager
default_manifest.xml renamed to default_upmana.xml
Cleaner clode for saved repositories
New TrueDebug Class in orpg_log (See documentation for usage)
Mercurial's hgweb folder is ported to upmana
Pretty important update that can help remove thousands of dead children from your
gametree.
Children, <forms />, <group_atts />, <horizontal />, <cols />, <rows />, <height
/>, etc... are all tags now. Check your gametree and look for dead children!!
New Gametree Recursion method, mapping, and context sensitivity. !Infinite Loops
return error instead of freezing the software!
New Syntax added for custom PC sheets
Tip of the Day added, from Core and community
Fixed Whiteboard ID to prevent random line or text deleting. Modified ID's to
prevent non updated clients from ruining the fix.
author | sirebral |
---|---|
date | Wed, 25 Nov 2009 06:15:49 -0600 |
parents | 3687514e0218 |
children |
line wrap: on
line source
"""xmltramp: Make XML documents easily accessible.""" __version__ = "2.16" __author__ = "Aaron Swartz" __credits__ = "Many thanks to pjz, bitsko, and DanC." __copyright__ = "(C) 2003 Aaron Swartz. GNU GPL 2." if not hasattr(__builtins__, 'True'): True, False = 1, 0 def isstr(f): return isinstance(f, type('')) or isinstance(f, type(u'')) def islst(f): return isinstance(f, type(())) or isinstance(f, type([])) empty = {'http://www.w3.org/1999/xhtml': ['img', 'br', 'hr', 'meta', 'link', 'base', 'param', 'input', 'col', 'area']} def quote(x, elt=True): if elt and '<' in x and len(x) > 24 and x.find(']]>') == -1: return "<![CDATA["+x+"]]>" else: x = x.replace('&', '&').replace('<', '<').replace(']]>', ']]>') if not elt: x = x.replace('"', '"') return x class Element: def __init__(self, name, attrs=None, children=None, prefixes=None): if islst(name) and name[0] == None: name = name[1] if attrs: na = {} for k in attrs.keys(): if islst(k) and k[0] == None: na[k[1]] = attrs[k] else: na[k] = attrs[k] attrs = na self._name = name self._attrs = attrs or {} self._dir = children or [] prefixes = prefixes or {} self._prefixes = dict(zip(prefixes.values(), prefixes.keys())) if prefixes: self._dNS = prefixes.get(None, None) else: self._dNS = None def __repr__(self, recursive=0, multiline=0, inprefixes=None): def qname(name, inprefixes): if islst(name): if inprefixes[name[0]] is not None: return inprefixes[name[0]]+':'+name[1] else: return name[1] else: return name def arep(a, inprefixes, addns=1): out = '' for p in self._prefixes.keys(): if not p in inprefixes.keys(): if addns: out += ' xmlns' if addns and self._prefixes[p]: out += ':'+self._prefixes[p] if addns: out += '="'+quote(p, False)+'"' inprefixes[p] = self._prefixes[p] for k in a.keys(): out += ' ' + qname(k, inprefixes)+ '="' + quote(a[k], False) + '"' return out inprefixes = inprefixes or {u'http://www.w3.org/XML/1998/namespace':'xml'} # need to call first to set inprefixes: attributes = arep(self._attrs, inprefixes, recursive) out = '<' + qname(self._name, inprefixes) + attributes if not self._dir and (self._name[0] in empty.keys() and self._name[1] in empty[self._name[0]]): out += ' />' return out out += '>' if recursive: content = 0 for x in self._dir: if isinstance(x, Element): content = 1 pad = '\n' + ('\t' * recursive) for x in self._dir: if multiline and content: out += pad if isstr(x): out += quote(x) elif isinstance(x, Element): out += x.__repr__(recursive+1, multiline, inprefixes.copy()) else: raise TypeError, "I wasn't expecting "+`x`+"." if multiline and content: out += '\n' + ('\t' * (recursive-1)) else: if self._dir: out += '...' out += '</'+qname(self._name, inprefixes)+'>' return out def __unicode__(self): text = '' for x in self._dir: text += unicode(x) return ' '.join(text.split()) def __str__(self): return self.__unicode__().encode('utf-8') def __getattr__(self, n): if n[0] == '_': raise AttributeError, "Use foo['"+n+"'] to access the child element." if self._dNS: n = (self._dNS, n) for x in self._dir: if isinstance(x, Element) and x._name == n: return x raise AttributeError, 'No child element named \''+n+"'" def __hasattr__(self, n): for x in self._dir: if isinstance(x, Element) and x._name == n: return True return False def __setattr__(self, n, v): if n[0] == '_': self.__dict__[n] = v else: self[n] = v def __getitem__(self, n): if isinstance(n, type(0)): # d[1] == d._dir[1] return self._dir[n] elif isinstance(n, slice(0).__class__): # numerical slices if isinstance(n.start, type(0)): return self._dir[n.start:n.stop] # d['foo':] == all <foo>s n = n.start if self._dNS and not islst(n): n = (self._dNS, n) out = [] for x in self._dir: if isinstance(x, Element) and x._name == n: out.append(x) return out else: # d['foo'] == first <foo> if self._dNS and not islst(n): n = (self._dNS, n) for x in self._dir: if isinstance(x, Element) and x._name == n: return x raise KeyError def __setitem__(self, n, v): if isinstance(n, type(0)): # d[1] self._dir[n] = v elif isinstance(n, slice(0).__class__): # d['foo':] adds a new foo n = n.start if self._dNS and not islst(n): n = (self._dNS, n) nv = Element(n) self._dir.append(nv) else: # d["foo"] replaces first <foo> and dels rest if self._dNS and not islst(n): n = (self._dNS, n) nv = Element(n); nv._dir.append(v) replaced = False todel = [] for i in range(len(self)): if self[i]._name == n: if replaced: todel.append(i) else: self[i] = nv replaced = True if not replaced: self._dir.append(nv) for i in todel: del self[i] def __delitem__(self, n): if isinstance(n, type(0)): del self._dir[n] elif isinstance(n, slice(0).__class__): # delete all <foo>s n = n.start if self._dNS and not islst(n): n = (self._dNS, n) for i in range(len(self)): if self[i]._name == n: del self[i] else: # delete first foo for i in range(len(self)): if self[i]._name == n: del self[i] break def __call__(self, *_pos, **_set): if _set: for k in _set.keys(): self._attrs[k] = _set[k] if len(_pos) > 1: for i in range(0, len(_pos), 2): self._attrs[_pos[i]] = _pos[i+1] if len(_pos) == 1 is not None: return self._attrs[_pos[0]] if len(_pos) == 0: return self._attrs def __len__(self): return len(self._dir) class Namespace: def __init__(self, uri): self.__uri = uri def __getattr__(self, n): return (self.__uri, n) def __getitem__(self, n): return (self.__uri, n) from xml.sax.handler import EntityResolver, DTDHandler, ContentHandler, ErrorHandler class Seeder(EntityResolver, DTDHandler, ContentHandler, ErrorHandler): def __init__(self): self.stack = [] self.ch = '' self.prefixes = {} ContentHandler.__init__(self) def startPrefixMapping(self, prefix, uri): if not self.prefixes.has_key(prefix): self.prefixes[prefix] = [] self.prefixes[prefix].append(uri) def endPrefixMapping(self, prefix): self.prefixes[prefix].pop() def startElementNS(self, name, qname, attrs): ch = self.ch; self.ch = '' if ch and not ch.isspace(): self.stack[-1]._dir.append(ch) attrs = dict(attrs) newprefixes = {} for k in self.prefixes.keys(): newprefixes[k] = self.prefixes[k][-1] self.stack.append(Element(name, attrs, prefixes=newprefixes.copy())) def characters(self, ch): self.ch += ch def endElementNS(self, name, qname): ch = self.ch; self.ch = '' if ch and not ch.isspace(): self.stack[-1]._dir.append(ch) element = self.stack.pop() if self.stack: self.stack[-1]._dir.append(element) else: self.result = element from xml.sax import make_parser from xml.sax.handler import feature_namespaces def seed(fileobj): seeder = Seeder() parser = make_parser() parser.setFeature(feature_namespaces, 1) parser.setContentHandler(seeder) parser.parse(fileobj) return seeder.result def parse(text): from StringIO import StringIO return seed(StringIO(text)) def load(url): import urllib return seed(urllib.urlopen(url)) def unittest(): parse('<doc>a<baz>f<b>o</b>ob<b>a</b>r</baz>a</doc>').__repr__(1,1) == \ '<doc>\n\ta<baz>\n\t\tf<b>o</b>ob<b>a</b>r\n\t</baz>a\n</doc>' assert str(parse("<doc />")) == "" assert str(parse("<doc>I <b>love</b> you.</doc>")) == "I love you." assert parse("<doc>\nmom\nwow\n</doc>")[0].strip() == "mom\nwow" assert str(parse('<bing> <bang> <bong>center</bong> </bang> </bing>')) == "center" assert str(parse('<doc>\xcf\x80</doc>')) == '\xcf\x80' d = Element('foo', attrs={'foo':'bar'}, children=['hit with a', Element('bar'), Element('bar')]) try: d._doesnotexist raise "ExpectedError", "but found success. Damn." except AttributeError: pass assert d.bar._name == 'bar' try: d.doesnotexist raise "ExpectedError", "but found success. Damn." except AttributeError: pass assert hasattr(d, 'bar') == True assert d('foo') == 'bar' d(silly='yes') assert d('silly') == 'yes' assert d() == d._attrs assert d[0] == 'hit with a' d[0] = 'ice cream' assert d[0] == 'ice cream' del d[0] assert d[0]._name == "bar" assert len(d[:]) == len(d._dir) assert len(d[1:]) == len(d._dir) - 1 assert len(d['bar':]) == 2 d['bar':] = 'baz' assert len(d['bar':]) == 3 assert d['bar']._name == 'bar' d = Element('foo') doc = Namespace("http://example.org/bar") bbc = Namespace("http://example.org/bbc") dc = Namespace("http://purl.org/dc/elements/1.1/") d = parse("""<doc version="2.7182818284590451" xmlns="http://example.org/bar" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:bbc="http://example.org/bbc"> <author>John Polk and John Palfrey</author> <dc:creator>John Polk</dc:creator> <dc:creator>John Palfrey</dc:creator> <bbc:show bbc:station="4">Buffy</bbc:show> </doc>""") assert repr(d) == '<doc version="2.7182818284590451">...</doc>' assert d.__repr__(1) == '<doc xmlns:bbc="http://example.org/bbc" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns="http://example.org/bar" version="2.7182818284590451"><author>John Polk and John Palfrey</author><dc:creator>John Polk</dc:creator><dc:creator>John Palfrey</dc:creator><bbc:show bbc:station="4">Buffy</bbc:show></doc>' assert d.__repr__(1,1) == '<doc xmlns:bbc="http://example.org/bbc" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns="http://example.org/bar" version="2.7182818284590451">\n\t<author>John Polk and John Palfrey</author>\n\t<dc:creator>John Polk</dc:creator>\n\t<dc:creator>John Palfrey</dc:creator>\n\t<bbc:show bbc:station="4">Buffy</bbc:show>\n</doc>' assert repr(parse("<doc xml:lang='en' />")) == '<doc xml:lang="en"></doc>' assert str(d.author) == str(d['author']) == "John Polk and John Palfrey" assert d.author._name == doc.author assert str(d[dc.creator]) == "John Polk" assert d[dc.creator]._name == dc.creator assert str(d[dc.creator:][1]) == "John Palfrey" d[dc.creator] = "Me!!!" assert str(d[dc.creator]) == "Me!!!" assert len(d[dc.creator:]) == 1 d[dc.creator:] = "You!!!" assert len(d[dc.creator:]) == 2 assert d[bbc.show](bbc.station) == "4" d[bbc.show](bbc.station, "5") assert d[bbc.show](bbc.station) == "5" e = Element('e') e.c = '<img src="foo">' assert e.__repr__(1) == '<e><c><img src="foo"></c></e>' e.c = '2 > 4' assert e.__repr__(1) == '<e><c>2 > 4</c></e>' e.c = 'CDATA sections are <em>closed</em> with ]]>.' assert e.__repr__(1) == '<e><c>CDATA sections are <em>closed</em> with ]]>.</c></e>' e.c = parse('<div xmlns="http://www.w3.org/1999/xhtml">i<br /><span></span>love<br />you</div>') assert e.__repr__(1) == '<e><c><div xmlns="http://www.w3.org/1999/xhtml">i<br /><span></span>love<br />you</div></c></e>' e = Element('e') e('c', 'that "sucks"') assert e.__repr__(1) == '<e c="that "sucks""></e>' assert quote("]]>") == "]]>" assert quote('< dkdkdsd dkd sksdksdfsd fsdfdsf]]> kfdfkg >') == '< dkdkdsd dkd sksdksdfsd fsdfdsf]]> kfdfkg >' assert parse('<x a="<"></x>').__repr__(1) == '<x a="<"></x>' assert parse('<a xmlns="http://a"><b xmlns="http://b"/></a>').__repr__(1) == '<a xmlns="http://a"><b xmlns="http://b"></b></a>' if __name__ == '__main__': unittest()