returns nothing no mater what keyword

[zjyfdu]

Hi, I just modified my code following the instruction of QuickStart and Config,
like this

@course.route("/search")
def w_search():
    keyword = request.args.get('q')
    results = Course.query.msearch(keyword,fields=['title']).first()
    return redirect(url_for('course.classes', id=results.id))

However it returns nothing no mater what keywords
did I miss some other configurations?
actually, I know nothing about the role of Create_index

Thanks a Lot For Your Help!!!

>>> Course.query.msearch('asdf',fields=['title']).all()
[]
>>> Course.query.filter_by(title='asdf').all()
[<Course u'asdf'>]

[honmaple]

You should use search.create_index(Course) if course data is exists before you use flask_msearch,or whoosh can't find any data.

after search.init_app(app),flask_maple will auto update model index and you don't have to create or update index unless set MSEARCH_ENABLE = False

[zjyfdu]

I figured it out! Thanks again!!

[AdamGold]

@honmaple The index doesn't automatically update for me. I have a test which creates a new user and then searches for it:

extra_user = User.create_from_credentials(**details)
db_instance.session.add(extra_user)
db_instance.session.commit()

assert User.query.msearch("name", ["username"]).all()

It doesn't work. When I add search.create_index() before the assertion, it works. MSEARCH_ENABLE is set to True.

[honmaple]

@AdamGold flask-maple use signal models_committed to update index automatically.
Please check if the config SQLALCHEMY_TRACK_MODIFICATIONS is True

[AdamGold]

@honmaple Will give it a shot. Another question - when I delete a row in the table with the flask_login proxy:

db_instance.session.delete(flask_login.current_user)

I get the following error:

  File "/venv/lib/python3.6/site-packages/flask_msearch/backends.py", line 163, in _index_signal
    self.create_one_index(instance, delete=True)
  File "/venv/lib/python3.6/site-packages/flask_msearch/whoosh_backend.py", line 163, in create_one_index
    ix = self._index(table)
  File "/venv/lib/python3.6/site-packages/flask_msearch/whoosh_backend.py", line 142, in _index
    name = model.__table__.name
AttributeError: type object 'LocalProxy' has no attribute '__table__'

[honmaple]

@AdamGold current_user isn't model instance, you should use current_user._get_current_object()

[yomajo]

Not getting results either.
sqlite database contains products with keyword tec:

on route I have:

@search_bp.route('/', methods=['GET', 'POST'])
def search():
    # available at localhost/search
    if request.method == 'POST':
        q = request.form.get('q', None)
        print(f'\nquery: {q}\n')
        
        results_raw = Products.query.msearch(q, fields=['name'], limit=10)
        print(results_raw)

        results = results_raw.all()
        print(results)

        return render_template('results.html', results=results, query=q)
        
    return render_template('search.html')

In terminal this yields:

query: tec

SELECT products.id AS products_id, products.name AS products_name, products.qty AS products_qty, products.brand_name AS products_brand_name 
FROM products 
WHERE 0 = 1
[]

Documentation has filter(...) on query object.
results = Post.query.msearch(keyword,fields=['title'],limit=20).filter(...)

How should I use it?

Branch for reference: yomajo/Flask-Experiments@7b46868

[honmaple]

@yomajo The default analyzer can't get keyword ech

>>> from whoosh.analysis import StemmingAnalyzer
>>> ana = StemmingAnalyzer()
>>> [token.text for token in ana("Ron-tech")]
['ron', 'tech']
>>> [token.text for token in ana("Opentech")]
['opentech']

You should use other analyzer or custom analyzer

from whoosh.analysis import Tokenizer
from whoosh.analysis.acore import Token

class CustomTokenizer(Tokenizer):
    def __call__(self, value, positions=False, start_pos=0, **kwargs):
        token = Token(positions, **kwargs)
        pos = start_pos
        for pos in range(0, len(value) - 2):
            token.text = value[pos:pos + 3]
            if positions:
                token.pos = pos
            yield token


def CustomAnalyzer():
    return CustomTokenizer()


ana = CustomAnalyzer()
print([token.text for token in ana("Ron-tech")])
# ['Ron', 'on-', 'n-t', '-te', 'tec', 'ech']
print([token.text for token in ana("Opentech")])
# ['Ope', 'pen', 'ent', 'nte', 'tec', 'ech']

https://whoosh.readthedocs.io/en/latest/analysis.html

[yomajo]

Thanks for reply @honmaple! I'm new to full-text search space, so please be adaptive on my idiocy.

If I understand correctly, analyzer returns tokens and if token (or their variable sum) is not a whole query, then search fails?

Example in your last example, say query was ch, while non of tokens have exact token:

print([token.text for token in ana("Opentech")])
# ['Ope', 'pen', 'ent', 'nte', 'tec', 'ech']

Then what added benefit is there over using:
results = <Model>.query.filter(<Model>.<column>.ilike(f'%{q}%')).all() - where q is a query.

which is case insensitive search for substrings in Model column. Is it performance? Or am I misunderstanding tokenization on whoosh part?