[
https://issues.apache.org/jira/browse/CALCITE-471?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14221131#comment-14221131
]
Vladimir Sitnikov commented on CALCITE-471:
-------------------------------------------
It might be better to split adapter in different classes/packages/modules so
each feature is easily observable:
1) "Demonstration of CSV full scan"
2) "Demonstration of CSV table that returns just required columns"
3) "Demonstration of CSV filterable table"
4) "Demonstration of table that returns _MAP"
etc.
Currently lots of features are mixed in a single implementation, thus it is a
bit tricky to figure out how CsvSmart table differs from CsvDump table. Earlier
I somehow managed to get over it, however that was very confusing.
> Create a "file" adapter, copying the demo CSV adapter
> ------------------------------------------------------
>
> Key: CALCITE-471
> URL: https://issues.apache.org/jira/browse/CALCITE-471
> Project: Calcite
> Issue Type: Bug
> Reporter: Julian Hyde
> Assignee: Julian Hyde
>
> The CSV adapter was intended as a demo. We should create a real,
> product-quality file adapter, by copying the CSV adapter, then letting people
> add features.
> [~lmccay] already suggested adding HDFS capabilities.
> Then the CSV adapter can remain simple enough to serve as a demo. We might
> even remove from it some recently added features, such as handling JSON and
> GZ files.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
