Mailing-List: contact user-help@uima.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@uima.apache.org
Subject: Re: How to correclty implement delta serialization in locally
 deployed CPE pipeline?
To: user@uima.apache.org
References: 
 <CADAZc=r-xSJr9Q_-mz5Zgix-anRG=yLaOqiYybDO4ckYeLpQ1w@mail.gmail.com>
From: Marshall Schor <msa@schor.com>
Message-ID: <560D48D1.4050401@schor.com>
Date: Thu, 1 Oct 2015 10:53:05 -0400
User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:38.0) Gecko/20100101
 Thunderbird/38.2.0
MIME-Version: 1.0
In-Reply-To: 
 <CADAZc=r-xSJr9Q_-mz5Zgix-anRG=yLaOqiYybDO4ckYeLpQ1w@mail.gmail.com>
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8bit

Hi,

A little more detail of what you're doing may help us figure out what's happening.

What API(s) are you using to do the serialization?

-Marshall

On 9/29/2015 2:57 PM, José Tomás Atria wrote:
> Hello all,
>
> I've been trying to wrap my head around this for a while, and I can't seem
> to get it to work. Could someone please explain what is the most
> straightforward way of implementing delta serialization in a local,
> multithreaded CPE pipeline?
>
> So far, I've tried using a collection reader that uses a
> SharedSerializationData that is stored in the current UIMA session, and
> creates a CAS marker that is also stored in a map in the current UIMA
> session under a CAS identifier key, and then using this
> SharedSerializationData oject and the marker retrieved from the UIMA
> session from the CAS identifier to serialize the delta to disk, but this
> procedure causes an OutOfMemory exception if I try to process all of my
> data (Not that much in my opinion, ~2000 CASes).
>
> I assume that I'm missing some basic aspect of the API, but after trying to
> deal with it for a while I just gave up...
>
> A more specific version, as far as I could understand: Delta serialization
> requires a SharedSerializationData object and a CAS marker. What is the
> correct way to create, store and retrieve these in a simple,
> multi-threaded, locally deployed CPE processing pipeline? (i.e. No need to
> support AS or DUCC facilities, etc).
>
> Any help would be greatly appreciated.
> Thanks!
> jta
>