The behaviour in grouping by MVA is wrong now. For example, we have some query like this:

q = Sphinxit('index').cluster('mva_attr')

where 'mva_attr' is an ordinary MVA attribute. I expect to receive counted distinct values, not unique groups of them.

Конфиг Sphinxit:

class SphinxitConfig(object):
    DEBUG = True
    WITH_META = True
    WITH_STATUS = False
    POOL_SIZE = 5
    SEARCHD_CONNECTION = {
        'host': '127.0.0.1',
        'port': 9306,
    }

Для воспроизведения бага не требуется создавать какие-либо индексы вообще.
При выполнении следующего кода

Snippet(index='nonexistent_index', config=SphinxitConfig)
    .for_query("nonexistent_query") 
    .from_data("'", 'test')
    .ask()

Sphinx не сможет выполнить запрос, так как в нем содержится неэкранированная одинарная кавычка.

sphinxql: syntax error, unexpected $end, expecting ')' or ',' near ...

When i try to fulltext searching with unicode keyword

from sphinxit.core.processor import Search
search_query = Search(**myconfig)
search_query = search.match(u'молоко')
search_query.ask() # returns empty result items, but in fact there are exists

Solution is:

lex = search_query.lex().encode('utf8')
It can be used in Search->ask() for example

Also Search class could be inherited like this

class UnicodeSearch(Search):
def lex(self):
return super(UnicodeSearch, self).lex().encode('utf8')

Sphinxit saves connection after invalid query (index is not exists, for example). It`s bad for some reasons, we have to destroy connector first, to see chages after fix.

Sphinxit.Snippets('index', data=['only good news'], query='Good News').process() sucks. I think it's possible to make more clean interface for that.

It would be easier to debug invalid query chain.

Hello,
when i use it, it always show error: (1040, 'Too many connections', None),
i think maybe you don't close Mysql connection after query.
would you please check it again ?
thank you for this great project!

Could you please create tutorial about how to integrate sphinxit with django application?

For example, limit() method. Proper values are integers but I don't want to make redundant conversion from strings if they are values from GET, for example. limit('10', '100') -> limit(10, 100), automatically. I want the same behaviour in another methods too.

Sphinxit is very sensitive for proper values and raises exceptions if something is wrong. It's good for debug but annoys in production. We have to check values and their types manually in our views and that sucks. It would be nice to have some 'strict' or 'debug' mode to raise or not to raise exceptions (on/off). If values are improper - just ignore them and go on.

Does the implementation allow rankers in expression mode?

I see, there is check of 'expr' ranker in in https://github.com/semirook/sphinxit/blob/master/sphinxit/core/convertors.py#L364 , but I can't find any statement to pass an expression to the ranker. How to do that?

I could pass my expression to my ranker using a hack of the line https://github.com/semirook/sphinxit/blob/master/sphinxit/core/convertors.py#L364 from
if not ranker in valid_rankers:
to
if not (ranker in valid_rankers or ranker[:4]=='expr'):
and the client code:

searcher = Search(indexes=[index], config=SphinxitConfig)
options = {
        'ranker': "expr('sum(lcs*user_weight)*1000+bm25')",
        'max_matches': 1,
}
search_query = searcher.match(query).options(**options)

The code works, but I suggest there is a direct way to do that. Isn't it?

I encountered several bugs with Q objects in Sphinxit.

• In complex expressions there are no parentheses around groups, and also some parts just get lost. For example:

.filter((Q(a__eq=1) | Q(b__eq=2)) & (Q(c__eq=3) | (Q(d__eq=4))))

The generated query:

(a=1) OR (b=2) AND (d=4) AS cnd

The c attribute is gone.

• Negation doesn't work at all.

Expressions .filter(~Q(a__eq=1)) and

.filter(Q(a__eq=1))

have the same result:

(a=1) AS cnd

Your tests are wrong too, as far as I can see:
https://github.com/semirook/sphinxit/blob/master/sphinxit/tests/test_core.py#L200, https://github.com/semirook/sphinxit/blob/master/sphinxit/tests/test_core.py#L224

Not working for me on my CentOS system with Sphinx 2.19. I suspect that Sphinx 2.19 might have had some changes that broke it.

>>> from sphinxit.core.helpers import BaseSearchConfig
>>> from sphinxit.core.processor import Search
>>> search_query = Search(indexes=['sourcecode'], config=BaseSearchConfig)
>>> search_query.ask()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "sphinxit/core/processor.py", line 287, in ask
    return self.connector.execute(query_batch)
  File "sphinxit/core/connector.py", line 149, in execute
    raise SphinxQLDriverException(e)
sphinxit.core.exceptions.SphinxQLDriverException: u'Variable_name'

Hi, i have a perfomance trouble. I used python sphinxapi before, and sometimes i had strange problems with huge batch queries (unpack exception). Anyway, it worked pretty fast, about 2s. Now i use sphinxit, and the same queries executes for 10s. I have only one question: is it really batch?

for sub_ql_pair in sxql_batch:
        subresult = {}
        sub_ql, sub_alias = sub_ql_pair
        cursor_exec(sub_ql)
        subresult['items'] = [r for r in cursor]

Profiler said some interesting:

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
      670    8.266    0.012    8.266    0.012 {method 'execute' of 'oursql.Cursor' objects}

It's hard to isolate faults with improper .filter() usage in some cases.

If I create query with conditions:

query = Search().match(term + '*').order_by('somefield2', 'desc').order_by('somefield', 'asc')

I want to get this ordering: "ORDER BY somefield2 DESC, somefield ASC"

But I always have this: "ORDER BY somefield ASC, somefield2 DESC"

This occurs because type of query._nodes.OrderBy.orderings is "set".

I temporarily solved this problem by set a "OrderBy.ordering" manifestly as list. You can fix this?

Example query:

search_query = Search(indexes=['test'], config=SphinxitConfig)
search_query = search_query.update(broken = 1).filter(id__in=[1, 2])
search_query.ask()

ends with:

sphinxit.core.exceptions.SphinxQLDriverException: no results available

It's failing on line 117 in connector.py

subresult['items'] = [r for r in cursor]

And to be exact the real exception is:

  File "/usr/lib/python2.6/site-packages/sphinxit/core/connector.py", line 118, in _execute_batch
    subresult['items'] = [r for r in cursor]
  File "cursor.pyx", line 183, in oursql.Cursor.fetchone (oursqlx/oursql.c:17532)
  File "cursor.pyx", line 161, in oursql.Cursor._check_statements (oursqlx/oursql.c:17369)
ProgrammingError: (None, 'no results available', None)

Hello,

Could you advise how to use SphinxIt to search via sql_attr_multi attributes? I need to implement search through the text with specified tags and rank the search results based on the number of matched tags as well. Could you provide me with an example how this can be implemented.

Thank you.