[jira] [Commented] (MRQL-93) Support query evaluation tracing and how-provenance

["Leonidas Fegaras (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MRQL-93?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15406179#comment-15406179 ] 

Leonidas Fegaras commented on MRQL-93:
--------------------------------------

This patch will support why-provenance only but it includes all the necessary code to support how-provenance soon. Basically, for each output data (such as a record), it will display all input data that contributed to the output data. This is done using the MRQL syntax: {{lineage e}}, where e is an MRQL expression. It lifts an expression e of type bag(t) to an expression of type bag((t,provenance)), which propagates (and extends) provenance information from the input to the output. The query processor will evaluate the expression e and, for each output value, it will print all the input values (from data sources) that contributed to the computation of this output value. The same patch provides also support for fine-grained provenance to explain how the output value was created (ie, what operations were used for constructing this value), but currently the is no GUI to explore this information. This will basically be the MRQL debugger to be completed soon. 

> Support query evaluation tracing  and how-provenance
> ----------------------------------------------------
>
>                 Key: MRQL-93
>                 URL: https://issues.apache.org/jira/browse/MRQL-93
>             Project: MRQL
>          Issue Type: New Feature
>          Components: Evaluation, Query Translation
>    Affects Versions: 0.9.8
>            Reporter: Leonidas Fegaras
>            Assignee: Leonidas Fegaras
>            Priority: Critical
>
> With this new feature, one can trace any piece of a query result to the input data that contributed to the result as well as the detailed operations used to derive the result. This is called why- and how-provenance in data analysis. A traced query will be lifted so that its output will contain the provenance information. There will be a GUI (based on JTree) to lazily navigate through the provenance data.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)