You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@tinkerpop.apache.org by Marc de Lignie <m....@xs4all.nl> on 2021/05/05 08:32:42 UTC

Re:[DISCUSS] Ongoing IO/io() Strategy

As the subject and audience is much broader than can be dicussed in a 
list like this, it would probably be nice to have some initial 
specification document in a branch on github. Such a document should 
include both user requirements and specifications how these will be 
implemented. This allows both developers and users to insert review 
comments.

E.g. for me it is not clear how the parquet request is linked to user 
requirements. I can think of all kinds of options:

  * save entire graph vs result of a query
  * save vertices and edges in separate files with the label as file name
  * save vertices, properties and adjacencies is a single nested data
    structure
  * write metadata to the parquet file so that it is recognizable as a
    TP file and can be read again (you probably do not want to read
    arbitrary files and build an ETL tool)