Re: Keeping schema static as objects are added

On Monday, November 26, 2012 12:19:06 AM UTC+7, Eric wrote:
> On 2012-11-25, sergei.sheinin_at_gmail.com <sergei.sheinin_at_gmail.com> wrote:
>
>
>
> As posted in comp.databases.theory:
>

>
>
>
> Anyway:
>
>
>
> 1) It would be nice to know what problem you are actually trying to
>
> solve here.

As a platform that performs programming logic on application tables it may find use anywhere in database applications. For example it could be programmed to automatically generate html interface to an existing database by scanning its schema, as a tool for generation of reports, forms, etc.

Registry is another potential use for Sprout because of its functions to save/retrieve hierarchical tree structures.

>
>
>
> 2) Most things described as avoiding schema changes turn out to be EAV
>
> which means, for example, that performance will not scale, and that
>
> anything other than very simple queries tend to be very difficult to get
>
> right.
>

I am proposing an EAV schema that accommodates its own metadata entered when application is initialized and contains program code, hierarchical trees, pointers and table schema. All of which are stored in Sprout's normalized database in same physical storage with application data.

Normally EAV table schema requires metadata descriptors (according to Wiki "by a factor of at least three or more") whereas in Sprout the entire data schema contains one dozen of tables and no assumptions of what application metadata will apply.

> performance will not scale

Introduce automatic partitioning of EAV tables. Obviously it is some concern.

> other than very simple queries tend to be very difficult

Sprout has functions that allow querying with regular SQL statements.

>
>
> 3) You seem to have a translator and a run-time interpreter, both of
>
> which are written in an interpreted language. This might be an
>
> unnecessary complication.
>

In tests I did not observe any noticeable slowdown when program code is run with Sprout interpreter. Seems only few new calculations are introduced when using its native syntax vs procedural SQL.

>
>
> 4) You might want to reconsider the name:
>
> http://www.cs.ox.ac.uk/projects/SPROUT/
>
> http://cedb.asce.org/cgi/WWWdisplay.cgi?97328
>
>

It is great I didn't know these were out there. Surely the name can or will change.

>
> Eric
>
> --
>
> ms fnd in a lbry

Sergei Received on Mon Nov 26 2012 - 04:18:55 CET