less required ADQL support [TAP-1.1]

gerard.lemson at gmail.com gerard.lemson at gmail.com
Tue Sep 29 21:07:28 CEST 2015


Hi 
> >
> > I'm not very familiar with NoSQL query syntax, havig only hacked
> > around with ElasticSearch in the context of log analysis... but the
> > query language wouldn't be in TAP and it would surely take some time
> > to make it a standard rather than custom lang...
> 
> What I am wondering is if there is a simple mapping from this NoSQL query
> syntax to ADQL.  My impression is that the NoSQL syntax can be expressed
as a
> subset of ADQL.  In that case, you are not looking so much for extension
as for a
> way to subset.  I already have to subset ADQL for my services.  When I
gave a
> talk about this at Sesto, the consensus was to let unsupported queries
fail.  So I 
> am not sure that anything needs to be done now.
> 

What is this "NoSQL query language"? Does it exist at all? 
Or does it depend on the particular product that you use as a NoSQL
database?
This article seems to indicate that CouchBase is developing N1QL, a dialect
of SQL++, as a kind of "standard" nosql query language:
http://www.informationweek.com/big-data/big-data-analytics/couchbase-bets-on
-standard-nosql-query-language/d/d-id/1320681 also
http://www.calit2.net/newsroom/release.php?id=2541 . 
Some argue that the whole point of NoSQL is that data is too diverse to be
handled by a single, standard No-SQL query language. 
In fact it seems that a lot of work in this direction is to allow one to
massage one's data in such a form that it can be queried by a SQL-like
language.

But in any case I don't really understand why we think we need to complicate
a TABLE Access Protocol with dealing with things that are not tables (but
aggregates of JSON strings, or BLOBs, or ...).
Even though it may not cover everything, at least we have a standard that
can be used from tools like TOPCAT to mimic interoperability.
In fact I think the main issue for lack of true interoperability between TAP
implementations is NOT that we do not also cover nosql databases.
I think the main issue is that we do not have (a sufficient number of)
standard data models mapped to a TAP compliant representation so that we can
use *a single ADQL string* to query all implementations of that model.

Cheers

Gerard

> Cheers,
> Walter Landry



More information about the dal mailing list