Adding custom items to nlp pipeline does not insert them in the correct order

[demongolem]

Using v 2.0 of spaCy with Python 2.7 on Ubuntu 14.03. I attempt to add 2 customer annotators to the pipeline using after. They do not end up in the position I would expect after the complete pipeline has been constructed.

I do as such:

nlp = spacy.load('en')

# Hook in your own deep learning models
print nlp.pipeline
nlp.add_pipe(first_regex, name='regex1', after='ner')
print nlp.pipeline
nlp.add_pipe(custom_model_trivial, name='useless', after='regex1')
print nlp.pipeline

The order goes from

tagger, parser, ner

to

tagger, parser, regex1, ner

to

tagger, parser, useless, regex1, ner

I would expect the order to be

tagger, parser, ner, regex1, useless

[honnibal]

Thanks, I thought we tested this! I think the problem is almost surely when we do after and the referenced item is at the end. I think we probably need to trigger an append here instead of an insert.

[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.