The backend is written in python.
This uses flask-login.
This uses JWT for user authentication. The client asks the server for a JWT by giving a username and password. If this succeeds, the token is stored in the browser (localStorage) until it expires (5 minutes is the default). All further requests to the server require knowledge of the JWT.
The client checks if the user is logged in by retrieving a variable from localStorage called user. This variable can be spoofed, of course, but without the proper JWT, no information can be accessed.
What about sessions?
Could we use flask_session with JWT tokens? I guess you could store stuff in the session that is keyed by the JWT token, or the user ID (probably that is a good idea).
What kinds of things do authorized users want to do? Load documents, annotate, save, check personal stats. The next thing one might do is request a dataset, then request a document. Then, they will make changes to the document and save changes to file. There is some tension between annotations "living" in the javascript (in this case, React framework) and in the filesystem. As they are annotated, they will live in the state of page. But when you leave the page, you will save to disk (as in the Java version).
Follow instructions here.
Run:
$ PYTHONPATH=. pytest
Otherwise you will get a module not found error.
If you want to run an individual test:
$ PYTHONPATH=. pytest -s -k "test_annotation_serialization"
To add a new CONLLU dataset (as used in Universal Dependencies), run the following:
$ python -m scripts.conllu_to_mongo --input-file <input_file.conllu> --dataset-name <your_dataset_name> --environment $ENVwhere $ENV should be either dev (for local development), or prod for using the hosted MongoDB.
For example, if you wanted to download the English Universal Dependencies data and put it into the local database, you would do the following:
$ git clone https://github.com/UniversalDependencies/UD_English-EWT.git
$ python -m scripts.conllu_to_mongo --input-file UD_English-EWT/en_ewt-ud-train.conllu --dataset-name en_ewt-ud-train --environment devIf you want to copy annotations from the dev to the prod database, or vice versa, use the following:
$ cd server
$ python -m scripts.copy_annotations --dataset-name en_ewt-ud-train --source dev --target prod --user-id stephen
This may be useful for annotating locally (to find bugs, or to annotate offline), and then copying to the cloud database.
You can use the spacy NER tagger to get starter or ("default") annotations for your dataset.
Run:
$ pip install spacy
$ python -m spacy download en_core_web_md
$ python -m scripts.merge_default_annotations --dataset-name <your_dataset_name> --environment $ENVwhere $ENV should be either dev (for local development), or prod for using the hosted MongoDB.