Translation not working - always returns empty string as response

[AdrianLC]

Hello,

Google must have changed something.

In [1]: import textblob

In [2]: blob = textblob.TextBlob("bonjour")

In [3]: blob.translate()
> /home/alopez/.virtualenvs/proj/lib/python2.7/site-packages/textblob/translate.py(42)translate()
     41         import ipdb; ipdb.set_trace()
---> 42         json5 = self._get_json5(self.url, host=host, type_=type_, data=data)
     43         if self._translation_successful(json5):

ipdb> s
--Call--
> /home/alopez/.virtualenvs/proj/lib/python2.7/site-packages/textblob/translate.py(84)_get_json5()
     83 
---> 84     def _get_json5(self, url, host=None, type_=None, data=None):
     85         encoded_data = urlencode(data).encode('utf-8')

ipdb> n
> /home/alopez/.virtualenvs/proj/lib/python2.7/site-packages/textblob/translate.py(85)_get_json5()
     84     def _get_json5(self, url, host=None, type_=None, data=None):
---> 85         encoded_data = urlencode(data).encode('utf-8')
     86         req = request.Request(url=url, headers=self.headers, data=encoded_data)

ipdb> 
> /home/alopez/.virtualenvs/proj/lib/python2.7/site-packages/textblob/translate.py(86)_get_json5()
     85         encoded_data = urlencode(data).encode('utf-8')
---> 86         req = request.Request(url=url, headers=self.headers, data=encoded_data)
     87         if host or type_:

ipdb> 
> /home/alopez/.virtualenvs/proj/lib/python2.7/site-packages/textblob/translate.py(87)_get_json5()
     86         req = request.Request(url=url, headers=self.headers, data=encoded_data)
---> 87         if host or type_:
     88             req.set_proxy(host=host, type=type_)

ipdb> 
> /home/alopez/.virtualenvs/proj/lib/python2.7/site-packages/textblob/translate.py(89)_get_json5()
     88             req.set_proxy(host=host, type=type_)
---> 89         resp = request.urlopen(req)
     90         content = resp.read()

ipdb> 
> /home/alopez/.virtualenvs/proj/lib/python2.7/site-packages/textblob/translate.py(90)_get_json5()
     89         resp = request.urlopen(req)
---> 90         content = resp.read()
     91         return content.decode('utf-8')

ipdb> 
> /home/alopez/.virtualenvs/proj/lib/python2.7/site-packages/textblob/translate.py(91)_get_json5()
     90         content = resp.read()
---> 91         return content.decode('utf-8')
     92 

ipdb> content
'""'
ipdb> resp.code
200
ipdb> resp.msg
'OK'

I will post an update if I find out something...

[AdrianLC]

@jschnurr Hi. I see you did the change to the json format. May I ask, where/how did you find out about the client=p parameter and its options?

[jschnurr]

I don't recall. I do know that the documentation was limited / non-existent.

[Gijs-Koot]

Google did indeed change the format. The request still works, but the response body is no longer a json encoded object, but a simple string. I was investigating the same problem, see http://stackoverflow.com/questions/35420602/python-textblob-not-translating/35421116#35421116.

[AdrianLC]

They also added a new code parameter to validate requests from their apps exclusively.
There's more info and a solution in soimort/translate-shell#94.
I will have a PR ready soon.

[jschnurr]

I provided PR #118 last night, but it doesn't deal with this extra parameter.

[AdrianLC]

Yeah. We've been working in the same :(
Sorry, I would have merged your PR into mine if I didn't have the work already done.

[jschnurr]

The solutions are similar. It seems there are two factors that can cause "empty string" and/or 403 results from the Google Translate API:

1. Headers. Once I added Connection & Accept headers, I started getting results back.

    headers = {'Connection': 'keep-alive',
               'Accept': '*/*',
               'User-Agent': ('Mozilla/5.0 (Macintosh; Intel Mac OS X 10_6_8) '
                              'AppleWebKit/535.19 (KHTML, like Gecko) Chrome/18'
                              '.0.1025.168 Safari/535.19')

2. tk parameter. Your solution doesn't have the headers, but the addition of the tk parameter appears to have the same effect.

Perhaps we should go with a 'belt and suspenders' approach here, and merge them together.

What do you think?

[AdrianLC]

It can't hurt I suppose. And I see in the inspector that Google is using them so why not.
Already added. Thanks @jschnurr