Skip to content

Latest commit

 

History

History
106 lines (64 loc) · 2.9 KB

README.md

File metadata and controls

106 lines (64 loc) · 2.9 KB
Raw

Flask for ElasticSearch

I have built a nice search-engine template using Python Flask in the backend and providing faceted navigation in the frontend.

The frontend was built using:

http://materializecss.com/

This project might be interesting for you, if you want to build a lightweight searchengine with Flask on top of an already existing ElasticSearch index.

Demo

alt tag

Dependencies

You need Python 2.7+, the packages elasticsearch and flask

$ pip install elasticsearch
$ pip install flask

And you should have a running instance of ElasticSearch Version 1.x.

Setup

You need to edit 2 files accordingly to run your project.

  1. fantasticsearch/views.py:
#TODO: Configuration
host = "http://localhost:9200"
indexName = "myIndex"
aggregationFields = ["field1", "field2"]

Basic queries with filters are already setup, but you can configure them according to your needs. Currently it is implemented to support the most generic and powerful query_string_query:

http://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-query-string-query.html#query-string-syntax

  1. fantasticsearch/templates/container.html

Inside the first row, please edit FIELD1 and FIELD2 according to your needs. And change the fieldnames inside the properties id and for to match your exact fieldnames as provided in the mapping.

    <input type="checkbox" id="field1-{{bucket.key}}">
    <label for="field1-{{bucket.key}}">{{bucket.key}} ({{bucket.doc_count}})</label>

Feel free to add more aggregations.

And then just start the flask-development server:

$ python run.py

Mapping considerations

Usually you don't need to create a mapping for an index with ElasticSearch, as it's schemaless, or better, creates a schema on the fly based on the first document. BUT if you want to provide faceted navigation, you should create an explicit mapping. Per default each field is analyzed. A terms facet of an analyzed field, will show you the analyzed tokens.

Which looks like this:

alt tag

What you want instead, is usually this:

alt tag

And for this you need to configure your mapping accordingly, before creating the index:

Usually you want the field both, intuitively searchable and "facetable", so you should use multifields like this:

"author" : {
      "type" : "string",
      "analyzer" : "english", #or standard, or any other language. For proper names, it's hard to find a "right way" 
      "fields": {
              "raw":   { "type": "string", "index": "not_analyzed" }
      }
}

http://www.elastic.co/guide/en/elasticsearch/reference/current/_multi_fields.html