devicemap-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Reza Naghibi <reza.nagh...@yahoo.com.INVALID>
Subject Re: Draft of 2.0 pattern spec
Date Sat, 10 Jan 2015 18:11:14 GMT
Ya, I left the 2 regex sections, input transformers and attribute parsing, as a TODO since
I wasnt exactly sure what best. Regex can be problematic and we would have to define what
subset or syntax would have to be supported.

Another option would be just to define a set of predefined transforms and parsers. Like lowercase(),
replaceAll(), tokenAfter() etc. This would be safer, easier, and leave the implementation
to the classifier, but obviously not as powerful.

So definitely something to think about.

      From: Werner Keil <werner.keil@gmail.com>
 To: dev@devicemap.apache.org 
 Sent: Saturday, January 10, 2015 11:03 AM
 Subject: Re: Draft of 2.0 pattern spec
   
Looks good at first sight. Must dig a bit deeper over the weekend or next
week (after flying to SF)

Especially based on current usage of RegEx I'd like to think about that
part in the proposal.
JSON is a valid idea for a format, processing it becomes increasingly easy
for many languages, though long term persistence (one big or many small
.JSON files?) other than in popular NoSQL databases like MongoDB (if we did
we'd certainly have to think of the right DB) can be a little tricky and
requires more thoughts in that direction.
Validation and Schema, thanks to XSL a very mature aspect of XML is also
still making the first "baby steps" with JSON, no real JSON Schema defined
yet although there are ideas and plans in that direction.

Werner



On Fri, Jan 9, 2015 at 11:47 PM, Reza Naghibi <
reza.naghibi@yahoo.com.invalid> wrote:

> Here is a draft:
>
> http://wiki.apache.org/devicemap/Patterns2
>
>
> So this is what I envision as the meat of 2.0. Think of it as high level
> ideas. Nothing written in stone, just brainstorming.
>
> Note how there is no mention of devices, browsers, or os. This is done on
> purpose. The classifiers simply implement this spec. They can focus on
> speed, efficiency, correctness. The data then uses the spec to classify.
> Data shouldn't have to worry about how it gets implemented and on what
> platform or language. So this spec allows for a really fast classification
> implementation and very concise and expressive data domains. So both sides
> should be served well. I hope :)
>
> So thats really it. Please let me know your thoughts. Im going to dive
> back into trawling JIRA for new devices and start getting the 1.0.2 data
> release ready.
>
>


   
Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message