write argument must be string

by Paul Agapow
in programming

tagged python, unicode, wsgi, web-development

When wsgi won't work.

So I was writing a web application using WSGI and - after working for a long time - it started erroring out with:

AssertionError: write() argument must be string

from deep inside the WSGI handler. Now, this usually means that you're returning unicode, which WSGI can't handle. However, it's quite happy to return a bytestream, so you can fix this by encoding your return from the application function:

x = "<html><body><p>dfgdfgfdg</p></body></html>".encode('utf-8')
response_headers = [
        ('Content-Type', 'text/html; charset=UTF-8'),
        ('Content-Length', str (len (x))),
]
start_response ('200 OK', response_headers)
return [x]

Note how the return text is always sent as an item of a list. This is because WSGI insists that you return an iterable, and if you send back a naked string, it will iterate over each character.

But still, I had an error. It was only when I tried to print out the response text that I realised what was going on. I never got to the print out point. In fact, I never got to the return from the application function - because I was accidentally returning from earlier in the function. Despite that return had no headers, WSGI errored out on its type: unicode text. So, easily fixed.