Settings, mappings and bulk indexing

Settings and mapping

Before creating our index in Elasticsearch, we can specify some settings and mappings as follows

index_body = {
    "settings": {
        "analysis": {
            "filter": {
                "dutch_stop": {
                "type": "stop",
                "stopwords": "_dutch_" 
                },
                "dutch_stemmer": {
                    "type": "stemmer",
                    "language": "dutch"
                }
            },
            "analyzer": {
                "dutch": {
                    "tokenizer": "standard",
                    "filter": [
                        "lowercase",
                        "dutch_stop",
                        "dutch_stemmer"
                    ]
                }
            }
        }
    },
    "mappings": {
        "article": {
            "properties": {
                "title":   { "type": "string", "analyzer": "dutch" },
                "body":    { "type": "string", "analyzer": "dutch" },
                "source":  { "type": "string"},
                "subject": { "type": "string"},
                "date":    { "type": "string"},
                "id":      { "type": "string"}
            }
        }
    }
}

Bulk indexing

After we have specified our settings and mapping, we can finally start indexing our previously created JSON files. To do this, we simply create our index and then use a generator expression to create a series of dictionaries containing our data. We can then bulk index this series of dictionaries, which is much faster than indexing them separately!

def bulk_index():
    es = Elasticsearch()
    es.indices.delete(index="telegraaf", ignore=[400, 404])
    es.indices.create("telegraaf", body=index_body, request_timeout=300)

    for infile in glob.glob(os.path.join("JSON", "*.json")):
        print infile
        with open(infile, "r") as f:
            all_articles = json.load(f)

        # Our generator
        k = ({"_type": "article", "_index": "telegraaf", "_source": article}
            for article in all_articles)

        helpers.bulk(es, k)