simserver.SimServer.buffer¶

SimServer.buffer(*args, **kwargs)[source]¶

Add a sequence of documents to be processed (indexed or trained on).

Here, the documents are simply collected; real processing is done later, during the self.index or self.train calls.

buffer can be called repeatedly; the result is the same as if it was called once, with a concatenation of all the partial document batches. The point is to save memory when sending large corpora over network: the entire documents must be serialized into RAM. See utils.upload_chunked().

A call to flush() clears this documents-to-be-processed buffer (flush is also implicitly called when you call index() and train()).