Allow custom regular expression for query filters

[Prill]

Currently, parse_query_string separates out filter queries (in the form of filter:value) from the rest of the query, using a regular expression. This is hard-coded in the file as filters_regexp. However, I would like to extend this to allow hyphens in the filter values without quotes (e.g. "foo:value-with-hyphens"). This is a trivial change to the regex, but because the value is hard-coded, changing this would require duplicating separate_filters_from_query just to change the expression.

Proposed solution

Change the signature of separate_filters_from_query and parse_query_string to have an optional filter_regex parameter which defaults to the existing expression:

DEFAULT_FILTER_REGEXP = re.compile(r'\b(\w+):(\w+|"[^"]+"|\'[^\']+\')')

def separate_filters_from_query(query_string, filters_regexp=DEFAULT_FILTER_REGEXP):
    ...  # The method body is unchanged.

def parse_query_string(query_string, operator=None, zero_terms=MATCH_NONE, filters_regexp=DEFAULT_FILTER_REGEXP):
    filters, query_string = separate_filters_from_query(query_string, filters_regexp=filters_regexp)
    # Remaining method body unchanged.

This would allow users to easily customize the regular expression for their own needs. Examples would include the one I mentioned above, or perhaps replacing the colon with an equals sign (filter=foo).

If this is something people think could be reasonable, I could try my hand at making a merge request. Wanted to clear with the maintainers first though.

[kaedroho]

Thanks for reporting!

However, I would like to extend this to allow hyphens in the filter values without quotes (e.g. "foo:value-with-hyphens").

To me, it feels like this should be the default behaviour, I'd be happy to accept a PR implementing this!

This would allow users to easily customize the regular expression for their own needs. Examples would include the one I mentioned above, or perhaps replacing the colon with an equals sign (filter=foo).

I'd prefer to keep it hard-coded unless we find a good use case for allowing customisation. AFAICT, the colon is universally used for these kind of filters. Customisation adds some complexity and can make it difficult for us to change how the parsing works if we ever need to. Those who want something different can always copy it.

[Prill]

FWIW, the regex change required to add hyphen is just

r'\b(\w+):([\w-]+|"[^"]+"|\'[^\']+\')'

All that changes is replacing the second \w+ with [\w-]+.

We could also run wild with it and allow any characters that aren't white space or quotes but that's probably excessive:

r'\b(\w+):("[^"]+"|\'[^\']+\'|[^\s"\']+)'