drill-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jacques Nadeau (JIRA)" <j...@apache.org>
Subject [jira] [Comment Edited] (DRILL-3572) Provide a simple interface to append metadata to files and directories (.drill)
Date Tue, 27 Oct 2015 15:43:27 GMT

    [ https://issues.apache.org/jira/browse/DRILL-3572?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14976576#comment-14976576
] 

Jacques Nadeau edited comment on DRILL-3572 at 10/27/15 3:42 PM:
-----------------------------------------------------------------

For some formats (for example text), it would be important to be able to potentially provide
field types and field names. This seems to be a property of the format rather than a general
concept that must be available for all formats.  For example 

{code}
{
  // Drill version identifier
  version: "dd1"  
  
  // Format Plugin Configuration
  format: {  
    type: "text", 
    fieldDelimiter: ",",
    lineDelimiter: "\n",
    recordStructure: [
      {name: "name", type: "VARCHAR(255)"},
      {name: "email", type: "VARCHAR(255)"},
      {name: "age", type: "INT"},
      {name: "registrationDate", type: "TIMESTAMP"},
    ]
  },

  // Option that describes what we should do if we fail to read a record
  // options include: warn, fail, record
  recordFailureMode: "warn",
  
  // Option that describes what to do if we fail to parse a file 
  // options include: warn, fail, record
  fileFailureMode: "warn"
}
{code}


was (Author: jnadeau):
For some formats (for example text), it would be important to be able to potentially provide
field types and field names. This seems to be a property of the format rather than a general
concept that must be available for all formats.  For example 

{code}
{
  // Drill version identifier
  version: "dd1"  
  
  // Format Plugin Configuration
  format: {  
    type: "text", 
    fieldDelimiter: ",",
    lineDelimiter: "\n",
    recordStructure: [
      {name: "name", type: "VARCHAR(255)"},
      {name: "email", type: "VARCHAR(255)"},
      {name: "age", type: "INT"},
      {name: "registrationDate", type: "TIMESTAMP"},
    ]
  }
}
{code}

> Provide a simple interface to append metadata to files and directories (.drill)
> -------------------------------------------------------------------------------
>
>                 Key: DRILL-3572
>                 URL: https://issues.apache.org/jira/browse/DRILL-3572
>             Project: Apache Drill
>          Issue Type: Bug
>          Components: Storage - Other
>            Reporter: Jacques Nadeau
>             Fix For: Future
>
>
> We need a way to store small amounts of metadata about a file or a collection of files.
 The current thinking was a way to have a "dot drill file" that ascribes metadata to a particular
asset. 
> Initial example file might be something that includes the following:
> {code}
> {
>   // Drill version identifier
>   version: "dd1"  
>   
>   // Format Plugin Configuration
>   format: {  
>     type: "httpd", 
>     format: "%h %l %u %t \"%r\" %>s %b \"%{Referer}i\" \"%{User-Agent}i\" \"%{Cookie}i\""}
>   },
>   
>   // Traits of underlying data (a.k.a physical properties)
>   traits: [ // traits of the underlying data
>     {type: "sort_nulls_first", columns: ["request.uri", "client.host"]}
>     {type: "unique", columns ["abc"]}
>     {type: "unqiue", columns ["xy", "zz"]}
>   ],
>   
>   // Mappings between directory names and exposed columns
>   dirs: [
>     {skip: true}, // don't include this directory name in the directory path.
>     {name: "year", type: "integer"},
>     {name: "month", type: "integer"},
>     {name: "day", type: "integer"}
>   ],
>   // whether or not a user can add new columns to the table through insert
>   rigid_table: true
>   
> }
> {code}
> We also need to support adding more machine-generated/managed data such as statistics.
 That should be done using a separate file from the one that is human description.
> A user should be able to ascribe this metadata directly through the file system as well
as through sql commands such as 
> {code}
> ALTER TABLE ADD METADATA ...
> {code}



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

Mime
View raw message