Combining Django and Elasticsearch (MelbDjango)

See the Google Docs document for the speaker notes: https://docs.google.com/presentation/d/1g0tNpU0rkTbd6kvq1DUmBVdTk6RNlnPlOWSoKoeo_YI

Code: https://github.com/MarkusH/talk-django-elasticsearch

Transcript

1. Combining Django & Elasticsearch
2. Search is hard
3. What is relevant information
4. Classify information importance
5. Selective search attributes
6. What’s Elasticsearch
7. Based on Lucene
8. Distributed Multitenant Full-text
9. RESTful API
10. JSON serialization
11. Open Source: Apache License 2.0
12. It’s Java
13. Elasticsearch & Python
14. Python Bindings
15. pyelasticsearch
16. elasticsearch elasticsearch_dsl
17. python-requests
18. Use the official libraries $ pip install elasticsearch-dsl
19. >>> from elasticsearch import Elasticsearch >>> from elasticsearch_dsl import Search >>> client = Elasticsearch() >>> s = Search(using=client, index="blog") \ ... .query("match", title="django") \ ... .filter("term", is_public=True) >>> response = s.execute()
20. Elasticsearch & Django
21. Django Bindings
22. Haystack
23. djangoes
24. elasticutils
25. django-simple-elasticsearch
26. Example
27. from elasticsearch_dsl import DocType, String class SearchArticle(DocType): title = String() text = String(analyzer='english') slug = String(index='not_analyzed') url = String(index='not_analyzed') # ... class Meta: index = 'blog' Index Mapping
28. AppConfig from elasticsearch_dsl.connections import connections class BlogConfig(AppConfig): name = 'blog' def ready(self): from . import models, search, signals connections.create_connection( **settings.ELASTICSEARCH_CONNS) search.SearchArticle.init() post_save.connect(signals.post_save_article, sender=models.Article)
29. Signals def post_save_article(sender, instance, created, **kwargs): article = SearchArticle(meta={‘id’: instance.pk}) \ if created else SearchArticle.get(id=instance.pk) article.title = instance.title article.text = instance.text article.url = instance.get_absolute_url() article.save() def post_delete_article(sender, instance, **kwargs): article = SearchArticle.get(id=instance.pk) article.delete()
30. Search View def search(request): q = request.GET.get('q', '') context = {'query': q, 'results': []} if q: search = SearchArticle.search() search = search.query('simple_query_string', query=q, fields=['title', 'text']) if not request.user.is_authenticated(): search = search.filter('term', is_public=True) context['results'] = search.execute() return render(request, 'blog/search.html', context)
31. Problems
32. Indexing happens during request time
33. Solution: Celery Integration
34. Celery Integration @shared_task(bind=True, default_retry_delay=60, max_retries=3) def index_article(self, pk): try: article = Article.objects.get(pk=pk) except Article.ObjectDoesNotExist: self.retry() try: search_article = SearchArticle.get(id=pk) except elasticsearch.NotFoundError: search_article = SearchArticle(meta={‘id’: pk}) search_article.title = instance.title # ... search_article.save()
35. Demo https://github.com/MarkusH/talk-django-elasticsearch
36. Thank you! markusholtermann.eu @m_holtermann github.com/MarkusH