#124: Exceptions when combining regular search and ranking in perform request
search
-----------------------+----------------------------------------------------
Reporter: tbrooks | Owner:
Type: defect | Status: new
Priority: major | Milestone:
Component: WebSearch | Version:
Keywords: |
-----------------------+----------------------------------------------------
caused by searches of the form
recid:240309 soliton
_with_ rm=citation (or word similarity)
I.e. a recid search with a bare extra search term
Joe points out the problem:
The problem appears be in search_engine.py:
{{{
4332 elif rm and p.startswith("recid:"):
4333 ## 3-ter - similarity search or citation search needed
4334 if req and not req.header_only:
4335 page_start(req, of, cc, aas, ln, uid, _("Search Results"),
p=create_page_title_search
4336 if of.startswith("h"):
4337 req.write(create_search_box(cc, colls_to_display, p, f, rg, sf, so,
sp, rm, of, ot, a
4338 p2, f2, m2, op2, p3, f3, m3, sc, pl, d1y, d1m, d1d, d2y,
4339 if record_exists(p[6:]) != 1:
4340 # record does not exist
4341 if of.startswith("h"):
4342 if req.header_only:
4343 raise apache.SERVER_RETURN, apache.HTTP_NOT_FOUND
4344 else:
4345 print_warning(req, _("Requested record does not seem to exist."))
4346 if of == "id":
4347 return []
4348 elif of.startswith("x"):
4349 # Print empty, but valid XML
4350 print_records_prologue(req, of)
4351 print_records_epilogue(req, of)
}}}
record_exists(p[6:]) naturally grabs everything to send along, including
the unexpectedly non-numeric rest of the string. This could just return
an error to the user, or it could use a regex to more correctly parse
out the recid. Or we could ask why the recid isn't already set when
this body of code is called.
Tibor points out that the correct solution is to move to a more general
way of identifying citation or similarity searching:
"recid:XXXXX" -> cites:XXXX
or
similarto:XXXX
then searches of the form:
cites:XXXX soliton
or, more likely
cites:XXXX 980:CORE
year:2007 cites:XXXX
(ie find all records that cite XXXX and were published in 2007...)
This makes more sense to the user, and extends functionality a bit, in
addition to fixing the exceptions raised when someone accidentally tries
the wrong thing.
--
Ticket URL: <http://cdswaredev.cern.ch/invenio/ticket/124>
Invenio <http://invenio-software.org>
