Google App Engine搜索API

Question

When querying a search index in the Python version of the GAE Search API , what is the best practice for searching for items where documents with words match the title are first returned, and then documents where words match the body?

For example given:

body = """This is the body of the document, 
with a set of words"""

my_document = search.Document(
  fields=[
    search.TextField(name='title', value='A Set Of Words'),
    search.TextField(name='body', value=body),
   ])

If it is possible, how might one perform a search on an index of Document s of the above form with results returned in this priority, where the phrase being searched for is in the variable qs :

1. Documents whose title matches the qs ; then
2. Documents whose body match the qs words.

It seems like the correct solution is to use a MatchScorer , but I may be off the mark on this as I have not used this search functionality before. It is not clear from the documentation how to use the MatchScorer , but I presume one subclasses it and overloads some function - but as this is not documented, and I have not delved into the code, I cannot say for sure.

Is there something here that I am missing, or is this the correct strategy? Did I miss where this sort of thing is documented?

---

Just for clarity here is a more elaborate example of the desired outcome:

documents = [
  dict(title="Alpha", body="A"),          # "Alpha"
  dict(title="Beta", body="B Two"),       # "Beta"
  dict(title="Alpha Two", body="A"),      # "Alpha2"
]

for doc in documents: 
  search.Document(
    fields=[
       search.TextField(name="title", value=doc.title),
       search.TextField(name="body", value=doc.body),
    ]
  )
  index.put(doc)  # for some search.Index

# Then when we search, we search the Title and Body.
index.search("Alpha")
# returns [Alpha, Alpha2]

# Results where the search is found in the Title are given higher weight.
index.search("Two")
# returns [Alpha2, Beta]  -- note Alpha2 has 'Two' in the title.

Answer 1

Custom scoring is one of our top priority feature requests. We're hoping to have a good way to do this sort of thing as soon as possible.

In your particular case, you could of course achieve the desired result by doing two separate queries: the first one with field restriction on "title", and the second restricted on "body".