translate.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-

"""
Google Translate API code
Based on http://code.activestate.com/recipes/576890/ but heavily modified.
"""

__author__ = 'David F. Gleich'
__version__ = 1

import re
import sys
import urllib
import urllib2
import simplejson
import time
import pprint

baseTranslateUrl = "http://ajax.googleapis.com/ajax/services/language/translate"
""" The URL for the Google translate API. """

baseUserAgent = "Python-Google-Translate-%i"%(__version__)
""" The default User-Agent string for the browser. """

retryDelay = 10
""" The default retry delay when a AJAX call fails. """

htmlCodes = [
    ('&', '&'),
    ('<', '&lt;'),
    ('>', '&gt;'),
    ('"', '&quot;'),
    ("'", '&#39;'),
]
""" A base set of html characters to encode and decode. """

def htmlDecode(s, codes=htmlCodes):
    """ Returns the ASCII decoded version of the given HTML string. This does
        NOT remove normal HTML tags like <p>. It is the inverse of htmlEncode().
        Ripped from somewhere on the internet.)
    """
    for code in codes:
        s = s.replace(code[1], code[0])
    return s


def translateRaw(text,format='text',langpair='|en',
        key=None,v='1.0',curReferrer=None,retry=5):
    """ Perform a raw call to the Google translate API.
    
    This function requires all attributes to be listed and only provides
    a simple retry option as a convinence.
    
    
    """
    params = {'q':text.encode('utf-8'),'format':format,'langpair':langpair,'v':v}
    if key is not None:
        params['key']=key
    
    if curReferrer is None:
        try:
            curReferrer = referrer # load referrer from outer namespace
        except:
            print "invalid referrer"
            return None
    
    if len(params['q'])>4500:
        return None
        
    request = urllib2.Request(baseTranslateUrl,data=urllib.urlencode(params))
    request.add_header("Referer", curReferrer)
    request.add_header("Accept-encoding", "gzip")
    request.add_header("User-Agent", baseUserAgent)
    
    resp = simplejson.load(urllib2.urlopen(request))
    
    try:
        retText = resp['responseData']['translatedText']
    except:
        if retry<=0:
            print "**************** Error! ****************"
            pprint.pprint(resp)
            retText = ""
        else:
            print "Warning (retry=%i): %s"%(retry,resp['responseDetails'])
            retText = translateRaw(text,format,langpair,key,v,curReferrer,
                retry-1)
        
    return retText
    #return params['q']
        
def translateLines(lines,src='',to='en',delay=1,callback=None,
    sentinal='___UNABLE_TO_TRANSLATE___'):
    """
    Translate multiple lines of independent text.
    
    This function takes a set of lines of text (non-html) and encodes them
    into a single translate call using the maximum possible size.
    If translation of any group of lines fail, they are translated with
    a sentinal value.  
    """
    
    output = []
    langpair = '%s|%s'%(src,to)
    oldlen = 0
    curlen = 7 # 7 extra chars on everything
    curlines = []
    for i,line in enumerate(lines):
        linelen = len(line.encode('utf-8'))+7
        if curlen + linelen > 4500:
            # translate the current group
            q = '<p>'+'</p><p>'.join(curlines)+'</p>'
            resp = translateRaw(q,format='html',langpair=langpair)
            if len(resp) == 0:
                # this is an error, so append a sentinal
                for line in curlines:
                    output.append(sentinal)
            else:
                for p in resp.split('</p>'):
                    output.append(htmlDecode(p[4:-1]))
                output.pop() # we always get an extra
            if callback:
                callback(i,len(lines))
            time.sleep(delay)
            curlines = [line]
            curlen = linelen+7 # first line has 7 extra chars
        else:
            curlines.append(line)
            curlen += linelen
    q = '<p>'+'</p><p>'.join(curlines)+'</p>'
    resp = translateRaw(q,format='html',langpair=langpair)
    if len(resp) == 0:
        # this is an error, so append a sentinal
        for line in curlines:
            output.append(sentinal)
    else:
        for p in resp.split('</p>'):
            output.append(htmlDecode(p[4:-1]))
        output.pop() # we always get an extra
    
    return output