Currently, we don't have a comprehensive document for building hyracks app. But the process of building an application won't be too hard.
The process is that:
1. you build your runtime operators (we already have a number of sample operators in hyracks/hyracks-dataflow-std, e.g., hybrid hash join, preclustered group-by, external sort, and so on)
2. you build your jobspec (build a DAG of operators with connectors, eg., one-to-one connector, hash-partitioning connector, hash-merging-partitioning connector, and so on)
3. execute the job spec (connect to the hyracks cc and submit the job)
We already have a number of examples, in hyracks/hyracks-examples, and you can learn from the code directly. Also I sent you the entry java file of our pregelix driver, which could also be an learning example for you.
If you want to build similarity search, you probably can use inverted list: hyracks/hyracks-storage-am-inverted-index.
Best regards,
Yingyi
On Thu, Oct 25, 2012 at 2:31 AM, Jiaying wang
<wangjia...@yahoo.com.cn> wrote:
Hi everyone:
I want to build an app to do similarity search. Where do I find the documentation for building Hyracks app?