Cannot process a txt file with spacy

[ghost]

Hi all,

I am trying to load an english text file but failing to do so, I am following spacy website 101 tutorial:

ff = io.open('moby_dick.txt', 'r', encoding='utf-8')
nlp(ff)
ff.close()

TypeError                                 Traceback (most recent call last)
<ipython-input-20-f474742159a7> in <module>()
      1 ff = io.open(files[7], 'r', encoding='utf-8')
----> 2 nlp(ff)
      3 ff.close()

~/.local/lib/python3.5/site-packages/spacy/language.py in __call__(self, text, disable)
    327             ('An', 'NN')
    328         """
--> 329         doc = self.make_doc(text)
    330         for name, proc in self.pipeline:
    331             if name in disable:

~/.local/lib/python3.5/site-packages/spacy/language.py in make_doc(self, text)
    355 
    356     def make_doc(self, text):
--> 357         return self.tokenizer(text)
    358 
    359     def update(self, docs, golds, drop=0., sgd=None, losses=None):

TypeError: Argument 'string' has incorrect type (expected str, got _io.TextIOWrapper)

Your Environment

• spaCy version: 2.0.5
• Platform: Linux-4.13.0-26-generic-x86_64-with-LinuxMint-18.3-sylvia
• Models: en
• Python version: 3.5.2

[bwj-GitHub]

I wasn't aware that nlp could take a file; you should read the file first: nlp(ff.read()).

[ghost]

Oh. I see. I took the idea from the example in the serialization section.

https://spacy.io/usage/spacy-101#serialization

[ines]

Sorry – this is a mistake in the example, it's indeed missing the .read(). Fixing! We might also update the example to not use Moby Dick and a made-up filename instead (like customer_feedback_627.txt, which we use in the lightning tour example on the homepage). Loading and parsing a large document like this in one is actually not the best strategy, especially in spaCy v2.0. For better performance, you'd usually want to split up the text and use nlp.pipe, which returns a generator.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.