Re: [xsl] Data science, data analytics using XSLT streaming

Subject: Re: [xsl] Data science, data analytics using XSLT streaming
From: Wendell Piez <wapiez@xxxxxxxxxxxxxxx>
Date: Tue, 5 Nov 2013 11:35:18 -0500

I agree with Andrew. These projects are fun, and XSLT pipelines are
well suited to them, because they are capable of exposing the semantic
issues and keeping syntax out of the way. This is even true when the
first step is a rendering of a non-XML format into an XML
representation, and the last step is serializing the data in a form
optimized for something else (such as your query engine of choice).

And no, streaming is not necessary, although it can help.

Plus, the biggest problem isn't scale anyway: it's the semantic
integrity of the data or (more likely) the lack thereof. This can be
compounded by non-technical issues such as data owners not seeing the
information they actually have because they are blinded by their
expectations of what it is "supposed" to be.

Cheers, Wendell

Wendell Piez |
XML | XSLT | electronic publishing
Eat Your Vegetables

On Tue, Nov 5, 2013 at 5:26 AM, Andrew Welch <andrew.j.welch@xxxxxxxxx> wrote:
>> XSLT streaming is all about processing large amounts of (XML-formatted) data.
>> So XSLT streaming should fit in the "data science" and "data analytics" categories.
>> Broad Question: Would you provide a scenario/example of doing data science/data analytics using XSLT streaming please?
> Typically the data is held in multiple files rather than 1 big one, so
> you don't necessarily need streaming, just a set of steps that process
> directories of xml into various intermediate formats, then into the
> final presentation view (such as a table with the data grouped,
> sorted, with counts)
> I've done this sort of thing a few times now and I always enjoy it.
> --
> Andrew Welch

Current Thread