mkdic-apertium-es-ca.py

import xml.dom.minidom, json

class DictError(Exception):
	pass

def innerText(node):
	if node.nodeType == node.TEXT_NODE:
		return node.nodeValue
	return u''.join(innerText(child) for child in node.childNodes)

def getPos(rnode):
	snodes = rnode.getElementsByTagName('s')
	if not snodes: return u''
	return snodes[0].getAttribute('n') or u''

dictSets = {} # es -> [[ca options]]	
doc = xml.dom.minidom.parse('apertium-es-ca.es-ca.dix')
for p in doc.getElementsByTagName('p'):
	lnodes = p.getElementsByTagName('l')
	rnodes = p.getElementsByTagName('r')
	if len(lnodes) > 1: raise DictError('Multiple l nodes: ' + p.toxml())
	if len(rnodes) > 1: raise DictError('Multiple r nodes: ' + p.toxml())
	if not lnodes or not rnodes: continue
	lnode = lnodes[0]
	rnode = rnodes[0]
	ltext = innerText(lnode)
	rtext = innerText(rnode)
	rpos = getPos(rnode)
	dictSets.setdefault(ltext, set()).add((rtext, rpos))

dictData = dict((k, sorted(v)) for k, v in dictSets.items())
print '{'
for i, (k, v) in enumerate(sorted(dictSets.items())):
	maybeComma = ',' if i < len(dictSets) else ''
	print '\t' + k.encode('UTF-8') + ': ' + json.dumps(sorted(v), ensure_ascii=False).encode('UTF-8') + maybeComma
print '}'