Re: GSOC 2018 ideas

2018-03-05 1:42 GMT-08:00 Aleksander Alekseev <a(dot)alekseev(at)postgrespro(dot)ru>:

> Hello Charles,
>
> > Went through the documents listed by you, and they are helpful!
> > It seems the main purpose of extension pg_protobuf is to parse
> > a protobuf struct and return the decoded field. May I ask how these kinds
> > of extensions are used in postgreSQL (or in other words, the scenarios to
> > use these plugins)?
>
> There are a few ideas behind all of this.
>
> 1) Sometimes people are not quite happy with strict relational schema by
> various reasons and prefer something more agile, like XML or JSON. These
> formats are indeed more convenient under certain circumstances, for
> instance in terms of ease of changing and migrating the schema.
>
> 2) One drawback of JSON is redundancy. For instance, you have to store
> the names of all document fields. These names don't carry much
> information but consume disk space and RAM thus affecting the overall
> performance. ZSON extension [1] partially solved this issue. However I
> wouldn't call it particularly convenient and the whole approach of
> compressing JSON seems to me more like a dirty hack, not a solution. The
> problem appeared because of using the wrong data format in the first
> place.
>
> 3) Unlike JSON, formats like Protobuf or Thrift are binary formats and
> most importantly don't store any field names. Thus they don't create a
> problem described above. However, PostgreSQL is not capable to access
> Protobuf fields out-of-the-box, for instance to index these fields. This
> is what pg_protobuf is for.
>
> The idea of using flexible schema and build index on top of them is
awesome!
Will definitely submit a proposal and focus on this if get selected.
Thanks for answering my questions.

> Hopefully this answers you question. If you have other questions please
> don't hesitate to ask!
>
> [1]: https://github.com/postgrespro/zson
>
>
> --
> Best regards,
> Aleksander Alekseev
>