writing to DynamoDB

[evdokim]

I'm going to clarify things that I we were talking during the last meeting.

• Actual writing should be separated from every model classes. The should produce just maps: Map[String, AttributeValue].
• Then other thing let's call it DynamoDBWriter will take that maps and do real uploading of data. It can be implemented in a different ways: straightforward synchronous writing, local blocking queue + working threads, SQS queue + worker instances, ....

[alberskib]

@evdokim Ok, in other words: For each dataset like GO, uniprot etc there shoul be parser that from files with data create abstract model (Map[String, AtributeValue]). Next there will be single Writer(the same for each dataset) that takes abstract model and save it to the database Thanks to it: hadling additional dataset require only writing specific parser. Correct?

[evdokim]

yes something like this. I think that writer can even don't know about dataset, it should have only maps with items together with tables names

[alberskib]

@evdokim Exactly it should operate on map(which is something like intermediate representation). Really similar concept could be found inside architecture of compilers.

[evdokim]

yes quite similar to IL