Complex Google App Engine Search -

Some quick questions related to GAE search and datastore:

(1) Why is it that I search Using the service, you can filter inequality on more than one property, but can I filter out the inequality on only one property while checking the datastore? It seems odd that this limit will be present in one service, but not the second.

(2) I want to use Google app engine search for a query of lots of objects (thousands or hundreds of thousands, maybe more). I'm planning to do many inequalities, for example: "time created" before x, less than "y" greater than "y", "rating" less than z, "latitude" between a and b, c And the "longitude" between d etc. looks like a lot of filters and is probably expensive, do app engines find suitable solutions for this?

Thank you very much.

1) Search Services basically gives you an API to fulfill those things You can not use datastore If you can do it on the datastore, you will not actually need the SearchService, however, there is not a very satisfactory answer, too many normal operations that you can do with conventional RDBMS is not possible before the search API is available. Were.

2) It is a bit difficult. Currently search API does not handle failure situations well, generally you will get a SearchServiceException without a meaningful message. It seems that the team is improving in the previous year or more, although improvement in this place is coming very slowly. Failure from tickets taken from me, failure is usually the result of very long lasting questions. It is usually shown as questions which are very complex. You can actually do a lot of inquiries along with the combination of query string, as well as the parameters applied to your search request, the negative side is that its completely black box, I have not seen any guides or devices on optimization queries. When they fail, they just fail.

AppEngine Search AP is designed to solve the problems of your description, even in your case it can be difficult to determine. You can set some sample questions and deploy for a test environment to see if it basically works for your specific data set. I hope this will work fine for the example given by you. I am successfully running a similar search in a mass production environment.