Google App Engine (python) : Search API : String Search

Question

i am using the Google App Engine Search API (https://developers.google.com/appengine/docs/python/search/). I have indexed all of the entities and the search is working fine. but only if i search for the exact matches else it returns 0 results. For Example:

from google.appengine.api import search

_INDEX_NAME = 'searchall'


query_string ="United Kingdom"
query = search.Query(query_string=query_string)
index = search.Index(name=_INDEX_NAME)

print index.search(query)

if i run the following script i do get results as follows :

search.SearchResults(results='[search.ScoredDocument(doc_id='c475fd24-34ba-42bd-a3b5-d9a48d880012', fields='[search.TextField(name='name', value='United Kingdom')]', language='en', order_id='45395666'), search.ScoredDocument(doc_id='5fa757d1-05bf-4012-93ff-79dd4b77a878', fields='[search.TextField(name='name', value='United Kingdom')]', language='en', order_id='45395201')]', number_found='2')

but if i change the query_string to "United Kin" or "United" it return 0 results as follows:

search.SearchResults(number_found='0')

I want to use this API for both normal search and AutoSuggest. What would be the best way to achieve this ?

Horwath

Answer 1

App Engine's full text search API does not support substring matching.

However, I needed this behavior myself to support search suggestions as the user types. Here is my solution for that:

""" Takes a sentence and returns the set of all possible prefixes for each word.
    For instance "hello world" becomes "h he hel hell hello w wo wor worl world" """
def build_suggestions(str):
    suggestions = []
    for word in str.split():
        prefix = ""
        for letter in word:
            prefix += letter
            suggestions.append(prefix)
    return ' '.join(suggestions)

# Example use
document = search.Document(
    fields=[search.TextField(name='name', value=object_name),
            search.TextField(name='suggest', value=build_suggestions(object_name))])

The basic idea is to manually generate separate keywords for every possible substring. This is only practical for short sentences, but it works great for my purposes.

Northeasterly

Thanks for your response nick, but i have quiet a large datastore, if i do that for each entity in the datastore would that not effect the size of the datastore ? i have like 4 different searchable models and each of them have more than 100,000 records on an average .. – Horwath

Yes, you'd have to store these auto-generated keywords alongside the rest of your data. – Northeasterly