Atlas Data Structure

#1

Post by kcl » Fri, Sep 02, 2016, 11:23 AM UTC

Atlas Data Structure
getkerf.wordpress.com

Re: Atlas Data Structure

#2

Post by dalke » Fri, Sep 02, 2016, 12:21 PM UTC

Is this like a data table like in the pandas package for Python, or in R?

Re: Atlas Data Structure

#3

Post by coderzach » Fri, Sep 02, 2016, 1:09 PM UTC

dalke wrote: post #2

Is this like a data table like in the pandas package for Python, or in R?

I don't think those data tables are schemaless. It seems like the value prop here is that it efficiently handles heterogenous data shapes.

Re: Atlas Data Structure

#4

Post by martinaparicio » Fri, Sep 02, 2016, 2:03 PM UTC

I would love to see the implementation of this. In the article doesn't say anything about the implementation details or the space tradeoffs.

Re: Atlas Data Structure

#5

Post by threeseed » Fri, Sep 02, 2016, 2:13 PM UTC

> interest in NoSQL-style computation seemed to be on the wane

Well that's just ridiculous. With machine/deep learning and big data analytics unquestionably on the rise there is actually more demand for a heterogenous architecture. And since in these fields you are often doing lazy-schema application having tools that support this paradigm are important.

In fact what does seem to be on the rise is having SQL on top of NoSQL stores e.g. Phoenix/HBase, Spark SQL. And they support all of SQL 2003 which by the looks of it this doesn't.

Re: Atlas Data Structure

#6

Post by threeseed » Fri, Sep 02, 2016, 2:20 PM UTC

martinaparicio wrote: post #4

I would love to see the implementation of this. In the article doesn't say anything about the implementation details or the space tradeoffs.

https://github.com/kevinlawler/kerf

It has no benchmarks so that's not a good sign. But there is another time series database that also allows you to use basic SQL: InfluxDB.

Re: Atlas Data Structure

#7

Post by martinaparicio » Fri, Sep 02, 2016, 5:19 PM UTC

threeseed wrote: post #6

martinaparicio wrote: post #4

I would love to see the implementation of this. In the article doesn't say anything about the implementation details or the space tradeoffs.

https://github.com/kevinlawler/kerf It has no benchmarks so that's not a good sign. But there is another time series database that also allows you to use basic SQL: InfluxDB.

I know about InfluxDB but I was interested in the implementation of the data structure. I'll check the git, thanks

-- edit --

Oh, that's shitty! The git is not the code of Kerf, is some documentation and examples :(