adiciona o buscador e cria views e templates para ele

This commit is contained in:
root
2025-03-07 15:31:53 +01:00
parent 1cd93f7955
commit 3f5ac79051
18 changed files with 378 additions and 235 deletions

View File

@ -2,5 +2,5 @@ from django.apps import AppConfig
class DiariosConfig(AppConfig):
default_auto_field = 'django.db.models.BigAutoField'
name = 'diarios'
default_auto_field = "django.db.models.BigAutoField"
name = "diarios"

View File

@ -2,67 +2,66 @@ from django_elasticsearch_dsl import Document, fields
from django_elasticsearch_dsl.registries import registry
from .models import PDFDocument
@registry.register_document
class PDFDocumentDocument(Document):
title = fields.TextField()
content = fields.TextField(analyzer='portuguese')
class Index:
name = 'pdf_documents'
settings = {
'number_of_shards': 1,
'number_of_replicas': 0,
'analysis': {
'analyzer': {
'portuguese': {
'type': 'custom',
'tokenizer': 'standard',
'filter': [
'lowercase',
'ascii_folding',
'portuguese_stemmer',
'stop',
'portuguese_synonyms',
]
},
'portuguese_search': {
'type': 'custom',
'tokenizer': 'standard',
'filter': [
'lowercase',
'ascii_folding',
'portuguese_stemmer',
'stop',
'suggest_shingle',
]
}
},
'filter': {
'suggest_shingle': {
'type': 'shingle',
'min_shingle_size': 2,
'max_shingle_size': 3
},
'stop': {
'type': 'stop',
'stopwords': '_portuguese_'
},
'ascii_folding': {
'type': 'asciifolding'
},
'portuguese_stemmer': {
'type': 'stemmer',
'language': 'portuguese'
},
'portuguese_synonyms':{
'type': 'synonym',
'synonyms_path': 'synonyms.txt',
'expand': True
}
}
}
content = fields.TextField(analyzer="portuguese")
pages = fields.NestedField(
properties={
"number": fields.IntegerField(),
"content": fields.TextField(analyzer="portuguese"),
}
)
class Index:
name = "pdf_documents"
settings = {
"number_of_shards": 1,
"number_of_replicas": 0,
"analysis": {
"analyzer": {
"portuguese": {
"type": "custom",
"tokenizer": "standard",
"filter": [
"lowercase",
"ascii_folding",
"portuguese_stemmer",
"stop",
"portuguese_synonyms",
],
},
"portuguese_search": {
"type": "custom",
"tokenizer": "standard",
"filter": [
"lowercase",
"ascii_folding",
"portuguese_stemmer",
"stop",
"suggest_shingle",
],
},
},
"filter": {
"suggest_shingle": {
"type": "shingle",
"min_shingle_size": 2,
"max_shingle_size": 3,
},
"stop": {"type": "stop", "stopwords": "_portuguese_"},
"ascii_folding": {"type": "asciifolding"},
"portuguese_stemmer": {"type": "stemmer", "language": "portuguese"},
"portuguese_synonyms": {
"type": "synonym",
"synonyms_path": "synonyms.txt",
"expand": True,
},
},
},
}
class Django:
model = PDFDocument
fields = ['uploaded_at']
fields = ["uploaded_at", "file"]

View File

@ -7,18 +7,25 @@ class Migration(migrations.Migration):
initial = True
dependencies = [
]
dependencies = []
operations = [
migrations.CreateModel(
name='PDFDocument',
name="PDFDocument",
fields=[
('id', models.BigAutoField(auto_created=True, primary_key=True, serialize=False, verbose_name='ID')),
('title', models.CharField(max_length=255)),
('file', models.FileField(upload_to='pdfs/')),
('content', models.TextField(blank=True)),
('uploaded_at', models.DateTimeField(auto_now_add=True)),
(
"id",
models.BigAutoField(
auto_created=True,
primary_key=True,
serialize=False,
verbose_name="ID",
),
),
("title", models.CharField(max_length=255)),
("file", models.FileField(upload_to="pdfs/")),
("content", models.TextField(blank=True)),
("uploaded_at", models.DateTimeField(auto_now_add=True)),
],
),
]

View File

@ -0,0 +1,18 @@
# Generated by Django 5.0.12 on 2025-03-07 13:47
from django.db import migrations, models
class Migration(migrations.Migration):
dependencies = [
("diarios", "0001_initial"),
]
operations = [
migrations.AddField(
model_name="pdfdocument",
name="page_content",
field=models.TextField(blank=True),
),
]

View File

@ -1,12 +1,14 @@
from django.db import models
import PyPDF2
import json
class PDFDocument(models.Model):
title = models.CharField(max_length=255)
file = models.FileField(upload_to='pdfs/')
file = models.FileField(upload_to="pdfs/")
content = models.TextField(blank=True)
uploaded_at = models.DateTimeField(auto_now_add=True)
page_content = models.TextField(blank=True)
def __str__(self):
return self.title
@ -15,9 +17,18 @@ class PDFDocument(models.Model):
if self.file:
pdf = PyPDF2.PdfReader(self.file)
texto = []
for pagina in pdf.pages:
pages_data = []
for i, pagina in enumerate(pdf.pages):
page_text = pagina.extract_text()
pages_data.append(
{
"number": i + 1,
"content": page_text,
}
)
texto.append(pagina.extract_text())
self.content = '\n'.join(texto)
self.content = "\n".join(texto)
self.page_content = json.dumps(pages_data)
super().save(*args, **kwargs)

View File

@ -13,6 +13,5 @@ def extract_text(sender, instance, created, **kwargs):
text = []
for page in pdf.pages:
text.append(page.extract_text())
instance.content = '\n'.join(text)
instance.save(update_fields=['content'])
instance.content = "\n".join(text)
instance.save(update_fields=["content"])

View File

@ -46,7 +46,16 @@
.result-item:last-child {
border-bottom: none;
}
.pdf-link {
color: #e74c3c;
margin-left: 10px;
font-size: 0.8em;
text-decoration: none;
}
.result-title {
display: flex;
align-items: center;
justify-content: space-between;
color: #1a0dab;
font-weight: 500;
margin-bottom: 5px;
@ -218,12 +227,23 @@ document.getElementById('suggestionsBox').addEventListener('click', function(e)
{% endif %}
</div>
<h5 class="result-title">
<a href="#">{{ result.highlighted_title|safe }}</a>
<a href="{{ result.pdf_url }}" target="_blank">{{ result.highlighted_title|safe }}</a>
<a href="{{ result.pdf_url }}" target="_blank" class="pdf-link" title="Abrir PDF completo">
<i class="bi bi-file-pdf"></i>
</a>
</h5>
<div class="result-content">{{ result.highlighted_content|safe }}</div>
<div class="result-meta">
<i class="bi bi-calendar-date"></i> {{ result.uploaded_at|date:"d/m/Y" }}
</div>
<div class="result-meta">
<i class="bi bi-calendar-date"></i> {{ result.uploaded_at|date:"d/m/Y" }}
{% if result.matching_pages %}
<span class="ms-3">
<i class="bi bi-file-earmark-text"></i> Páginas encontradas:
{% for page in result.matching_pages %}
<a href="{{ result.pdf_url }}#page={{ page }}" target="_blank" class="badge bg-light text-dark">{{ page }}</a>
{% endfor %}
</span>
{% endif %}
</div>
</div>
{% endfor %}
</div>

View File

@ -2,6 +2,6 @@ from django.urls import path
from .views import search_view, spellcheck_view
urlpatterns = [
path('pesquisa/', search_view, name='search_view'),
path('spellcheck/', spellcheck_view, name='spellcheck_view'),
path("pesquisa/", search_view, name="search_view"),
path("spellcheck/", spellcheck_view, name="spellcheck_view"),
]

View File

@ -1,9 +1,11 @@
import json
import debugpy
from django.shortcuts import render
from elasticsearch_dsl import Search, Q
from elasticsearch_dsl.connections import connections
from django.conf import settings
import re
from .documents import PDFDocument
from django.http import JsonResponse
@ -12,224 +14,295 @@ connections.create_connection(hosts=[settings.ELASTICSEARCH_HOSTS])
def spellcheck_view(request):
query = request.GET.get('q', '')
query = request.GET.get("q", "")
suggestions = []
if query:
s = Search(index='pdf_documents')
s = s.suggest('auto_correct', query,
phrase={
'field': 'suggest',
'size': 3,
'gram_size': 3,
'confidence': 2.0,
'direct_generator': [{
'field': 'suggest',
'suggest_mode': 'popular'
}]
})
s = Search(index="pdf_documents")
s = s.suggest(
"auto_correct",
query,
phrase={
"field": "suggest",
"size": 3,
"gram_size": 3,
"confidence": 2.0,
"direct_generator": [{"field": "suggest", "suggest_mode": "popular"}],
},
)
response = s.execute()
if hasattr(response.suggest, 'auto_correct'):
if hasattr(response.suggest, "auto_correct"):
for option in response.suggest.auto_correct[0].options:
suggestions.append(option.text)
return JsonResponse({'suggestions': suggestions})
return JsonResponse({"suggestions": suggestions})
def search_view(request):
query = request.GET.get('q', '') # Obtém o termo de pesquisa da URL
page = int(request.GET.get('page', 1))
query = request.GET.get("q", "") # Obtém o termo de pesquisa da URL
page = request.GET.get("page", 1) # Obtém o valor de "page" (padrão: 1)
# Converte page para int
try:
page = int(page)
except ValueError:
page = 1 # Valor padrão em caso de erro
results = []
suggestions = []
spelling_correction = None
total_hits = 0
per_page = 10
if query:
# Processamento especial para termos entre aspas
exact_phrases = re.findall(r'"([^"]*)"', query)
# Remove os termos entre aspas da consulta principal
cleaned_query = query
for phrase in exact_phrases:
cleaned_query = cleaned_query.replace(f'"{phrase}"', '')
cleaned_query = cleaned_query.replace(f'"{phrase}"', "")
# Remove espaços extras e pontuação desnecessária
cleaned_query = re.sub(r'\s+', ' ', cleaned_query).strip()
cleaned_query = re.sub(r"\s+", " ", cleaned_query).strip()
# Cria uma consulta no Elasticsearch
search = Search(index='pdf_documents')
search = Search(index="pdf_documents")
# Lista para armazenar todas as consultas
queries = []
# Adiciona consulta para termos gerais (com fuzziness para tolerância a erros)
if cleaned_query:
queries.append(
Q('multi_match',
query=cleaned_query,
fields=['title^3', 'content^2', 'synonyms^1'],
fuzziness='AUTO',
boost=2,)
)
queries.append(
Q('match',
synonyms={
'query': cleaned_query,
'boost': 0.5
})
)
Q(
"multi_match",
query=cleaned_query,
fields=["title^3", "content^2", "synonyms^1"],
fuzziness="AUTO",
boost=2,
)
)
queries.append(Q("match", synonyms={"query": cleaned_query, "boost": 0.5}))
# Adiciona consultas exatas para frases entre aspas (sem fuzziness)
for phrase in exact_phrases:
if phrase.strip():
# Consulta de frase exata para o título com peso alto
queries.append(
Q('match_phrase',
title={
'query': phrase,
'boost': 3,
'slop': 0 # Sem flexibilidade na ordem das palavras
})
Q(
"match_phrase",
title={
"query": phrase,
"boost": 3,
"slop": 0, # Sem flexibilidade na ordem das palavras
},
)
)
# Consulta de frase exata para o conteúdo com peso médio
queries.append(
Q('match_phrase',
content={
'query': phrase,
'boost': 2,
'slop': 0 # Sem flexibilidade na ordem das palavras
})
Q(
"match_phrase",
content={
"query": phrase,
"boost": 2,
"slop": 0, # Sem flexibilidade na ordem das palavras
},
)
)
# Combina as consultas com OR (se houver alguma)
if queries:
search = search.query(
Q('bool', should=queries, minimum_should_match=1)
)
search = search.query(Q("bool", should=queries, minimum_should_match=1))
# Configuração do highlight para mostrar mais contexto
search = search.highlight('content', fragment_size=300, number_of_fragments=2, pre_tags=['<mark>'], post_tags=['</mark>'])
search = search.highlight('title', fragment_size=300, number_of_fragments=1, pre_tags=['<mark>'], post_tags=['</mark>'])
search = search.highlight(
"content",
fragment_size=300,
number_of_fragments=2,
pre_tags=["<mark>"],
post_tags=["</mark>"],
)
search = search.highlight(
"title",
fragment_size=300,
number_of_fragments=1,
pre_tags=["<mark>"],
post_tags=["</mark>"],
)
# Paginação
search = search[(page-1)*per_page:page*per_page]
search = search[(page - 1) * per_page : page * per_page]
# Executa a consulta
response = search.execute()
total_hits = response.hits.total.value
# Processa os resultados
for hit in response:
# Obter o objeto PDFDocument correspondente
try:
pdf_doc = PDFDocument.objects.get(id=hit.meta.id)
pdf_url = pdf_doc.file.url # URL do PDF
matching_pages = []
if hasattr(hit.meta, 'highlight') and hasattr(hit.meta.highlight, 'pages.content'):
for highlight in hit.meta.highlight['pages.content']:
page_matches = re.findall(r'page_(\d+)', highlight)
if page_matches:
matching_pages.append(int(page_matches[0]))
if not matching_pages and query:
if pdf_doc.page_content:
try:
page_data = json.loads(pdf_doc.page_content)
for page_d in page_data:
if query.lower() in page_d['content'].lower():
matching_pages.append(page_d['number'])
except json.JSONDecodeError as e:
logger.error(f"Erro ao decodificar JSON para o documento {pdf_doc.id}: {e}")
page_data = []
else:
page_data = []
matching_pages = sorted(list(set(matching_pages)))
except PDFDocument.DoesNotExist:
pdf_url = ""
matching_pages = []
# Extrai o conteúdo destacado ou usa o original
if hasattr(hit.meta, 'highlight') and hasattr(hit.meta.highlight, 'content'):
highlighted_content = ' ... '.join(hit.meta.highlight.content)
if hasattr(hit.meta, "highlight") and hasattr(
hit.meta.highlight, "content"
):
highlighted_content = " ... ".join(hit.meta.highlight.content)
else:
# Se não houver highlight, pegue os primeiros 300 caracteres
highlighted_content = hit.content[:300] + '...' if len(hit.content) > 300 else hit.content
highlighted_content = (
hit.content[:300] + "..."
if len(hit.content) > 300
else hit.content
)
# Extrai o título destacado ou usa o original
if hasattr(hit.meta, 'highlight') and hasattr(hit.meta.highlight, 'title'):
if hasattr(hit.meta, "highlight") and hasattr(
hit.meta.highlight, "title"
):
highlighted_title = hit.meta.highlight.title[0]
else:
highlighted_title = hit.title
# Verifica se o resultado corresponde a uma frase exata
is_exact_match = any(phrase.lower() in hit.content.lower() or
phrase.lower() in hit.title.lower()
for phrase in exact_phrases)
results.append({
'id': hit.meta.id,
'title': hit.title,
'highlighted_title': highlighted_title,
'highlighted_content': highlighted_content,
'uploaded_at': hit.uploaded_at,
'score': hit.meta.score,
'is_exact_match': is_exact_match
})
# Sugestões "Você quis dizer" (apenas para termos fora de aspas)
if total_hits < 5 and cleaned_query:
suggestion_search = Search(index='pdf_documents')
suggestion_search = suggestion_search.suggest(
'term_suggestion',
cleaned_query,
term={
'field': 'content',
'suggest_mode': 'popular',
'size': 5
is_exact_match = any(
phrase.lower() in hit.content.lower()
or phrase.lower() in hit.title.lower()
for phrase in exact_phrases
)
results.append(
{
"id": hit.meta.id,
"title": hit.title,
"highlighted_title": highlighted_title,
"highlighted_content": highlighted_content,
"uploaded_at": hit.uploaded_at,
"score": hit.meta.score,
"is_exact_match": is_exact_match,
"pdf_url": pdf_url,
"matching_pages": matching_pages,
}
)
# Sugestões "Você quis dizer" (apenas para termos fora de aspas)
if total_hits < 5 and cleaned_query:
suggestion_search = Search(index="pdf_documents")
suggestion_search = suggestion_search.suggest(
"term_suggestion",
cleaned_query,
term={"field": "content", "suggest_mode": "popular", "size": 5},
)
suggestion_response = suggestion_search.execute()
if hasattr(suggestion_response, 'suggest') and hasattr(suggestion_response.suggest, 'term_suggestion'):
if hasattr(suggestion_response, "suggest") and hasattr(
suggestion_response.suggest, "term_suggestion"
):
for suggestion in suggestion_response.suggest.term_suggestion:
for option in suggestion.options:
suggestions.append(option.text)
# Cria uma correção ortográfica se necessário
if suggestions and total_hits == 0:
corrected_query = cleaned_query
for suggestion_term in suggestion_response.suggest.term_suggestion:
for (
suggestion_term
) in suggestion_response.suggest.term_suggestion:
if suggestion_term.options:
# Substitui palavras incorretas por sugestões
word_to_replace = suggestion_term.text
corrected_word = suggestion_term.options[0].text
corrected_query = re.sub(r'\b' + re.escape(word_to_replace) + r'\b',
corrected_word,
corrected_query,
flags=re.IGNORECASE)
corrected_query = re.sub(
r"\b" + re.escape(word_to_replace) + r"\b",
corrected_word,
corrected_query,
flags=re.IGNORECASE,
)
# Reconstrói a consulta original mantendo as frases entre aspas
if corrected_query != cleaned_query:
spelling_correction = corrected_query
for phrase in exact_phrases:
spelling_correction += f' "{phrase}"'
spelling_correction = spelling_correction.strip()
# Busca por termos relacionados (apenas se houver poucos resultados)
if total_hits < 3 and cleaned_query:
related_terms = Search(index='pdf_documents')
related_terms = Search(index="pdf_documents")
related_terms = related_terms.query(
'more_like_this',
fields=['content', 'title'],
"more_like_this",
fields=["content", "title"],
like=cleaned_query,
min_term_freq=1,
max_query_terms=10,
min_doc_freq=1
min_doc_freq=1,
)
related_terms = related_terms[:5]
related_response = related_terms.execute()
for hit in related_response:
# Verifica se este documento já está nos resultados
if not any(r.get('id') == hit.meta.id for r in results):
results.append({
'id': hit.meta.id,
'title': hit.title,
'highlighted_title': hit.title,
'highlighted_content': hit.content[:300] + '...' if len(hit.content) > 300 else hit.content,
'uploaded_at': hit.uploaded_at,
'score': hit.meta.score,
'is_related': True
})
if not any(r.get("id") == hit.meta.id for r in results):
results.append(
{
"id": hit.meta.id,
"title": hit.title,
"highlighted_title": hit.title,
"highlighted_content": (
hit.content[:300] + "..."
if len(hit.content) > 300
else hit.content
),
"uploaded_at": hit.uploaded_at,
"score": hit.meta.score,
"is_related": True,
"pdf_url": pdf_url,
}
)
# Calcula a paginação
total_pages = (total_hits + per_page - 1) // per_page if total_hits > 0 else 0
# Renderiza o template com os resultados
return render(request, 'diarios/search_results.html', {
'query': query,
'results': results,
'suggestions': suggestions[:5], # Limita a 5 sugestões
'spelling_correction': spelling_correction,
'total_hits': total_hits,
'page': page,
'total_pages': total_pages,
'page_range': range(max(1, page-2), min(total_pages+1, page+3)),
'has_exact_phrases': bool(exact_phrases)
})
# Renderiza o template com os resultados
return render(
request,
"diarios/search_results.html",
{
"query": query,
"results": results,
"suggestions": suggestions[:5], # Limita a 5 sugestões
"spelling_correction": spelling_correction,
"total_hits": total_hits,
"page": page,
"total_pages": total_pages,
"page_range": range(max(1, page - 2), min(total_pages + 1, page + 3)),
"has_exact_phrases": bool(exact_phrases),
},
)