Old-style classes interfere with returning strings

[Slater-Victoroff]

I have no idea how to solve this one, but when trying to return strings from a function decorated with an @threads, rather than executing, it simply errors. Code below to reproduce:

def download_nm(delimiter, source, dump):
    images = [line.split(delimiter)[10] for line in open(source)][1:]  # avoid header row
    with open(dump, 'a') as sink:
        for i, image in enumerate(images):
            results = return_image_json(image)  # type(results) == <type 'instance'> 
            sink.write(results)  # Error, expected string or buffer

@threads(16)
def return_image_json(image_link):
    response = requests.get(image_link)
    encoded = "data:%s;base64,%s" % (response.headers['Content-Type'], base64.b64encode(response.content))
    return json.dumps({image_link: encoded}) + '\n'

It looks like moving to new-style classes resolves this issue, but since you're relying on some of the syntax-hacks of old-style classes I'm not sure if this is solvable.

[Slater-Victoroff]

Did more digging into this error. Seems that tomorrow breaks if any type-checking whatsoever is done on the resulting object. So passing the results into most any library function should break.

[madisonmay]

Super valid complaint. It's actually unrelated to the kind of class being used -- it's simply that calls to isinstance and type(obj) don't actually trigger a __getattr__ call.

You can fix this by doing literally anything to the result object before calling a function that checks the results type. I would recommend calling tomorrow._wait() explicitly in these circumstances. Simplified example below:

import base64
import json

from tomorrow import threads
import requests

def save_image_json(images, filename):
    with open(filename, 'a') as sink:
        for i, image in enumerate(images):
            results = return_image_json(image)
            results._wait()
            sink.write(results) 

@threads(4)
def return_image_json(image_link):
    response = requests.get(image_link)
    encoded = "data:%s;base64,%s" % (response.headers['Content-Type'], base64.b64encode(response.content))
    return json.dumps({image_link: encoded}) + '\n'

if __name__ == "__main__":
    # koala images
    images = [
        'http://i.imgur.com/1p5nKyZ.jpg',
        'http://i.imgur.com/lDt3nJ8.jpg', 
        'http://i.imgur.com/hkUeCBT.jpg', 
        'http://i.imgur.com/ZglSElj.jpg'
    ]
    save_image_json(images, filename='koalas.json')

I'm going to look into forcing isinstance and type to trigger resolution of the future, but I'm not entirely certain this is possible. If that doesn't work I'll probably just end up updating the README to make sure people are aware of this edge case.

[980202006]

if I have the error after run code，what I get is all the Tomorrow class. Is there any method to deal with it？