less fields to index
[wolnelektury.git] / apps / search / views.py
index d056a18..fd5883e 100644 (file)
@@ -8,14 +8,14 @@ from django.views.decorators import cache
 from django.http import HttpResponse, HttpResponseRedirect, Http404, HttpResponsePermanentRedirect
 from django.utils.translation import ugettext as _
 
-from catalogue.utils import get_random_hash
+from catalogue.utils import split_tags
 from catalogue.models import Book, Tag, Fragment
-from catalogue.fields import dumps
 from catalogue.views import JSONResponse
 from search import Search, JVM, SearchResult
 from lucene import StringReader
 from suggest.forms import PublishingSuggestForm
-
+from time import sleep
+import re
 import enchant
 
 dictionary = enchant.Dict('pl_PL')
@@ -31,14 +31,15 @@ def match_word_re(word):
 def did_you_mean(query, tokens):
     change = {}
     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:
-                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
 
@@ -51,14 +52,30 @@ def did_you_mean(query, tokens):
     return query
 
 
+JVM.attachCurrentThread()
+_search = None
+
+
+def get_search():
+    global _search
+
+    while _search is False:
+        sleep(1)
+
+    if _search is None:
+        _search = False
+        _search = Search()
+    return _search
+
+
 def hint(request):
     prefix = request.GET.get('term', '')
     if len(prefix) < 2:
         return JSONResponse([])
     JVM.attachCurrentThread()
-    s = Search()
 
-    hint = s.hint()
+    search = get_search()
+    hint = search.hint()
     try:
         tags = request.GET.get('tags', '')
         hint.tags(Tag.get_tag_list(tags))
@@ -70,14 +87,17 @@ 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.
 
-    tags = s.hint_tags(prefix)
-    books = s.hint_books(prefix)
+    tags = search.hint_tags(prefix, pdcounter=True)
+    books = search.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,
-          'category': _(t.category),
+          'category': category_name(t.category),
           'id': t.id,
           'url': t.get_absolute_url()}
           for t in tags] + \
@@ -91,75 +111,123 @@ def hint(request):
 def main(request):
     results = {}
     JVM.attachCurrentThread()  # where to put this?
-    srch = Search()
 
     results = None
     query = None
-    fuzzy = False
+    fuzzy = False #0.8
+
+    query = request.GET.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)
+
+    # hint = search.hint()
+    # try:
+    #     tag_list = Tag.get_tag_list(tags)
+    # except:
+    #     tag_list = []
+
+    if len(query) < 2:
+        return render_to_response('catalogue/search_too_short.html', {'prefix': query},
+                                  context_instance=RequestContext(request))
+
+    search = get_search()
+    # hint.tags(tag_list)
+    # if book:
+    #     hint.books(book)
+    tags = search.hint_tags(query, pdcounter=True, prefix=False, fuzzy=fuzzy)
+    tags = split_tags(tags)
+
+    toks = StringReader(query)
+    tokens_cache = {}
+
+    author_results = search.search_phrase(toks, 'authors', fuzzy=fuzzy, tokens_cache=tokens_cache)
+    title_results = search.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 = search.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(
+        search.search_phrase(toks, 'content', fuzzy=fuzzy, tokens_cache=tokens_cache, snippets=True, book=False, slop=4),
+        search.search_some(toks, ['content'], tokens_cache=tokens_cache, snippets=True, book=False))
+
+    everywhere = search.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'])
 
-    if 'q' in request.GET:
-        tags = request.GET.get('tags', '')
-        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)
+    suggestion = did_you_mean(query, search.get_tokens(toks, field="SIMPLE"))
 
-        hint = srch.hint()
+    def ensure_exists(r):
         try:
-            tag_list = Tag.get_tag_list(tags)
-        except:
-            tag_list = []
-
-        if len(query) < 2:
-            return render_to_response('catalogue/search_too_short.html', {'tags': tag_list, 'prefix': query},
-                                      context_instance=RequestContext(request))
-
-        hint.tags(tag_list)
-        if book:
-            hint.books(book)
-
-        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_book(toks, fuzzy=fuzzy, hint=hint),
-                                         srch.search_perfect_parts(toks, fuzzy=fuzzy, hint=hint),
-                                         srch.search_everywhere(toks, fuzzy=fuzzy, hint=hint))
-
-        for r in results:
-            r.process_hits()
-
-        results.sort(reverse=True)
-
-        for r in results:
-            print "-----"
-            for h in r.hits:
-                print "- %s" % h
-
-                # Did you mean?
-        suggestion = did_you_mean(query, srch.get_tokens(toks, field="SIMPLE"))
-
-        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())
-        elif len(results) == 0:
-            form = PublishingSuggestForm(initial={"books": query + ", "})
-            return render_to_response('catalogue/search_no_hits.html',
-                                      {'tags': tag_list,
-                                       'prefix': query,
-                                       "form": form,
-                                       'did_you_mean': suggestion},
-                context_instance=RequestContext(request))
-
-        return render_to_response('catalogue/search_multiple_hits.html',
-                                  {'tags': tag_list,
+            return r.book
+        except Book.DoesNotExist:
+            return False
+
+    author_results = filter(ensure_exists, author_results)
+    title_results = filter(ensure_exists, title_results)
+    text_phrase = filter(ensure_exists, text_phrase)
+    everywhere = filter(ensure_exists, everywhere)
+
+    results = author_results + title_results + text_phrase + everywhere
+    # ensure books do exists & sort them
+    results.sort(reverse=True)
+
+    if len(results) == 1:
+        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',
+                                  {'tags': tags,
                                    'prefix': query,
-                                   'results': results,
+                                   "form": form,
                                    'did_you_mean': suggestion},
             context_instance=RequestContext(request))
+
+    return render_to_response('catalogue/search_multiple_hits.html',
+                              {'tags': tags,
+                               'prefix': query,
+                               'results': { 'author': author_results,
+                                            'title': title_results,
+                                            'content': text_phrase,
+                                            'other': everywhere},
+                               'did_you_mean': suggestion},
+        context_instance=RequestContext(request))