fnp / wolnelektury.git / blobdiff
? search:
summary | shortlog | log | commit | commitdiff | tree
raw | inline | side by side
infopages fixes
[wolnelektury.git] / apps / search / views.py
diff --git a/apps/search/views.py b/apps/search/views.py
index 6dcaf74..527ca82 100644 (file)
--- a/apps/search/views.py
+++ b/apps/search/views.py
@@ -1,5 +1,6 @@
 # -*- coding: utf-8 -*-
 
 # -*- coding: utf-8 -*-
 
+from django.conf import settings
 from django.shortcuts import render_to_response, get_object_or_404
 from django.template import RequestContext
 from django.contrib.auth.decorators import login_required
 from django.shortcuts import render_to_response, get_object_or_404
 from django.template import RequestContext
 from django.contrib.auth.decorators import login_required
@@ -7,28 +8,38 @@ from django.views.decorators import cache
 from django.http import HttpResponse, HttpResponseRedirect, Http404, HttpResponsePermanentRedirect
 from django.utils.translation import ugettext as _
 
 from django.http import HttpResponse, HttpResponseRedirect, Http404, HttpResponsePermanentRedirect
 from django.utils.translation import ugettext as _
 
-from catalogue.utils import get_random_hash
-from catalogue.models import Book, Tag, Fragment, TAG_CATEGORIES
+from catalogue.utils import split_tags
+from catalogue.models import Book, Tag, Fragment
 from catalogue.fields import dumps
 from catalogue.views import JSONResponse
 from catalogue.fields import dumps
 from catalogue.views import JSONResponse
-from catalogue import forms
-from search import MultiSearch, JVM, SearchResult
+from search import Search, JVM, SearchResult
 from lucene import StringReader
 from suggest.forms import PublishingSuggestForm
 from lucene import StringReader
 from suggest.forms import PublishingSuggestForm
-
+import re
 import enchant
 
 dictionary = enchant.Dict('pl_PL')
 
 
 import enchant
 
 dictionary = enchant.Dict('pl_PL')
 
 
+def match_word_re(word):
+    if 'sqlite' in settings.DATABASES['default']['ENGINE']:
+        return r"\b%s\b" % word
+    elif 'mysql' in settings.DATABASES['default']['ENGINE']:
+        return "[[:<:]]%s[[:>:]]" % word
+
+
 def did_you_mean(query, tokens):
     change = {}
 def did_you_mean(query, tokens):
     change = {}
-    # sprawdzić, czy słowo nie jest aby autorem - proste szukanie termu w author!
     for t in tokens:
     for t in tokens:
-        print("%s ok? %s, sug: %s" %(t, dictionary.check(t), dictionary.suggest(t)))
+        authors = Tag.objects.filter(category='author', name__iregex=match_word_re(t))
+        if len(authors) > 0:
+            continue
+
         if not dictionary.check(t):
             try:
         if not dictionary.check(t):
             try:
-                change[t] = dictionary.suggest(t)[0]
+                change_to = dictionary.suggest(t)[0].lower()
+                if change_to != t.lower():
+                    change[t] = change_to
             except IndexError:
                 pass
 
             except IndexError:
                 pass
 
@@ -46,7 +57,7 @@ def hint(request):
     if len(prefix) < 2:
         return JSONResponse([])
     JVM.attachCurrentThread()
     if len(prefix) < 2:
         return JSONResponse([])
     JVM.attachCurrentThread()
-    s = MultiSearch()
+    s = Search()
 
     hint = s.hint()
     try:
 
     hint = s.hint()
     try:
@@ -60,15 +71,17 @@ def hint(request):
     # jezeli tagi dot tylko ksiazki, to wazne zeby te nowe byly w tej samej ksiazce
     # jesli zas dotycza themes, to wazne, zeby byly w tym samym fragmencie.
 
     # jezeli tagi dot tylko ksiazki, to wazne zeby te nowe byly w tej samej ksiazce
     # jesli zas dotycza themes, to wazne, zeby byly w tym samym fragmencie.
 
-    
-    tags = s.hint_tags(prefix)
+    tags = s.hint_tags(prefix, pdcounter=True)
     books = s.hint_books(prefix)
 
     books = s.hint_books(prefix)
 
-    # TODO DODAC TU HINTY
+    def category_name(c):
+        if c.startswith('pd_'):
+            c=c[len('pd_'):]
+        return _(c)
 
     return JSONResponse(
         [{'label': t.name,
 
     return JSONResponse(
         [{'label': t.name,
-          'category': _(t.category),
+          'category': category_name(t.category),
           'id': t.id,
           'url': t.get_absolute_url()}
           for t in tags] + \
           'id': t.id,
           'url': t.get_absolute_url()}
           for t in tags] + \
@@ -79,82 +92,118 @@ def hint(request):
             for b in books])
 
 
             for b in books])
 
 
-def foo(s, q, tag_list=None):
-    hint = s.hint()
-    try:
-        tag_list = Tag.get_tag_list(tag_list)
-        hint.tags(tag_list)
-    except:
-        tag_list = None
-
-    q = StringReader(q)
-    return (q, hint)
-
-
 def main(request):
     results = {}
     JVM.attachCurrentThread()  # where to put this?
 def main(request):
     results = {}
     JVM.attachCurrentThread()  # where to put this?
-    srch = MultiSearch()
+    srch = Search()
 
     results = None
     query = None
 
     results = None
     query = None
-    fuzzy = False
+    fuzzy = False #0.8
 
     if 'q' in request.GET:
 
     if 'q' in request.GET:
-        tags = request.GET.get('tags', '')
+        # tags = request.GET.get('tags', '')
         query = request.GET['q']
         query = request.GET['q']
-        book_id = request.GET.get('book', None)
-        book = None
-        if book_id is not None:
-            book = get_object_or_404(Book, id=book_id)
+        # book_id = request.GET.get('book', None)
+        # book = None
+        # if book_id is not None:
+        #     book = get_object_or_404(Book, id=book_id)
 
 
-        hint = srch.hint()
-        try:
-            tag_list = Tag.get_tag_list(tags)
-        except:
-            tag_list = []
+        # hint = srch.hint()
+        # try:
+        #     tag_list = Tag.get_tag_list(tags)
+        # except:
+        #     tag_list = []
 
         if len(query) < 2:
 
         if len(query) < 2:
-            return render_to_response('catalogue/search_too_short.html', {'tags': tag_list, 'prefix': query},
+            return render_to_response('catalogue/search_too_short.html', {'prefix': query},
                                       context_instance=RequestContext(request))
 
                                       context_instance=RequestContext(request))
 
-        hint.tags(tag_list)
-        if book:
-            hint.books(book)
+        # hint.tags(tag_list)
+        # if book:
+        #     hint.books(book)
+        tags = srch.hint_tags(query, pdcounter=True, prefix=False, fuzzy=fuzzy)
+        tags = split_tags(tags)
 
         toks = StringReader(query)
 
         toks = StringReader(query)
-        fuzzy = 'fuzzy' in request.GET
-        if fuzzy:
-            fuzzy = 0.7
-
-        results = SearchResult.aggregate(srch.search_perfect_book(toks, fuzzy=fuzzy, hint=hint),
-                                         srch.search_perfect_parts(toks, fuzzy=fuzzy, hint=hint),
-                                         srch.search_everywhere(toks, fuzzy=fuzzy, hint=hint))
+        tokens_cache = {}
+
+        author_results = srch.search_phrase(toks, 'authors', fuzzy=fuzzy, tokens_cache=tokens_cache)
+        title_results = srch.search_phrase(toks, 'title', fuzzy=fuzzy, tokens_cache=tokens_cache)
+
+        # Boost main author/title results with mixed search, and save some of its results for end of list.
+        # boost author, title results
+        author_title_mixed = srch.search_some(toks, ['authors', 'title', 'tags'], fuzzy=fuzzy, tokens_cache=tokens_cache)
+        author_title_rest = []
+        for b in author_title_mixed:
+            bks = filter(lambda ba: ba.book_id == b.book_id, author_results + title_results)
+            for b2 in bks:
+                b2.boost *= 1.1
+            if bks is []:
+                author_title_rest.append(b)
+
+        # Do a phrase search but a term search as well - this can give us better snippets then search_everywhere,
+        # Because the query is using only one field.
+        text_phrase = SearchResult.aggregate(
+            srch.search_phrase(toks, 'content', fuzzy=fuzzy, tokens_cache=tokens_cache, snippets=True, book=False, slop=4),
+            srch.search_some(toks, ['content'], tokens_cache=tokens_cache, snippets=True, book=False))
+
+        everywhere = srch.search_everywhere(toks, fuzzy=fuzzy, tokens_cache=tokens_cache)
+
+        def already_found(results):
+            def f(e):
+                for r in results:
+                    if e.book_id == r.book_id:
+                        e.boost = 0.9
+                        results.append(e)
+                        return True
+                return False
+            return f
+        f = already_found(author_results + title_results + text_phrase)
+        everywhere = filter(lambda x: not f(x), everywhere)
+
+        author_results = SearchResult.aggregate(author_results)
+        title_results = SearchResult.aggregate(title_results)
+
+        everywhere = SearchResult.aggregate(everywhere, author_title_rest)
+
+        for res in [author_results, title_results, text_phrase, everywhere]:
+            res.sort(reverse=True)
+            for r in res:
+                for h in r.hits:
+                    h['snippets'] = map(lambda s:
+                                        re.subn(r"(^[ \t\n]+|[ \t\n]+$)", u"",
+                                                re.subn(r"[ \t\n]*\n[ \t\n]*", u"\n", s)[0])[0], h['snippets'])
+
+        suggestion = did_you_mean(query, srch.get_tokens(toks, field="SIMPLE"))
+        print "dym? %s" % repr(suggestion).encode('utf-8')
+
+        results = author_results + title_results + text_phrase + everywhere
         results.sort(reverse=True)
 
         results.sort(reverse=True)
 
-        for r in results:
-            print r.hits
-
         if len(results) == 1:
         if len(results) == 1:
-            if len(results[0].hits) == 0:
-                return HttpResponseRedirect(results[0].book.get_absolute_url())
-            elif len(results[0].hits) == 1 and results[0].hits[0] is not None:
-                frag = Fragment.objects.get(anchor=results[0].hits[0])
-                return HttpResponseRedirect(frag.get_absolute_url())
+            fragment_hits = filter(lambda h: 'fragment' in h, results[0].hits)
+            if len(fragment_hits) == 1:
+                #anchor = fragment_hits[0]['fragment']
+                #frag = Fragment.objects.get(anchor=anchor)
+                return HttpResponseRedirect(fragment_hits[0]['fragment'].get_absolute_url())
+            return HttpResponseRedirect(results[0].book.get_absolute_url())
         elif len(results) == 0:
             form = PublishingSuggestForm(initial={"books": query + ", "})
             return render_to_response('catalogue/search_no_hits.html',
         elif len(results) == 0:
             form = PublishingSuggestForm(initial={"books": query + ", "})
             return render_to_response('catalogue/search_no_hits.html',
-                                      {'tags': tag_list, 'prefix': query, "pubsuggest_form": form,
-                                       'form': forms.SearchForm()},
+                                      {'tags': tags,
+                                       'prefix': query,
+                                       "form": form,
+                                       'did_you_mean': suggestion},
                 context_instance=RequestContext(request))
 
                 context_instance=RequestContext(request))
 
+        print "TAGS: %s" % tags
         return render_to_response('catalogue/search_multiple_hits.html',
         return render_to_response('catalogue/search_multiple_hits.html',
-                                  {'tags': tag_list, 'prefix': query,
-                                   'results': results, 'from': forms.SearchForm()},
+                                  {'tags': tags,
+                                   'prefix': query,
+                                   'results': { 'author': author_results,
+                                                'title': title_results,
+                                                'content': text_phrase,
+                                                'other': everywhere},
+                                   'did_you_mean': suggestion},
             context_instance=RequestContext(request))
             context_instance=RequestContext(request))
-
-    # return render_to_response('newsearch/search.html', {'results': results,
-    #                                                     'did_you_mean': (query is not None) and
-    #                                                     did_you_mean(query, srch.get_tokens(query, field='SIMPLE')),
-    #                                                     'fuzzy': fuzzy},
-    #                           context_instance=RequestContext(request))