[jira] [Updated] (SYSTEMDS-2565) FrameBlock read and write

["Kevin Innerebner (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/SYSTEMDS-2565?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Kevin Innerebner updated SYSTEMDS-2565:
---------------------------------------
    Description: 
FrameBlocks fail (null pointer access) to be serialized in cases where their _coldata is null. This is reproducable:
{code:java}
// ... code
FrameBlock fb = new FrameBlock(data.getNumColumns(), Types.ValueType.STRING);
fb.write(do);
{code}
Note that this is an internal error, so I am not sure if this can happen while executing a dml script.

By calling FrameBlock.ensureAllocatedColumns(rows) before writing this can be stopped from happening and I am not aware of instances where this lead to problems at the moment.

The problem arises in the FrameBlock.write() method:
{code:java}
// ... code
//write columns (value type, data)
for( int j=0; j<getNumColumns(); j++ ) {
   out.writeByte(_schema[j].ordinal());
   if( !isDefaultMeta ) {
      out.writeUTF(getColumnName(j));
      out.writeLong(_colmeta[j].getNumDistinct());
      out.writeUTF( (_colmeta[j].getMvValue()!=null) ? 
            _colmeta[j].getMvValue() : "" );
   }
   _coldata[j].write(out);
}
{code}
The problem is the line:
{code:java}
   _coldata[j].write(out);
{code}
Although we have columns _coldata is still null.

  was:
FrameBlocks fail (null pointer access) to be serialized in cases where their _coldata is null. This is reproducable:
{code:java}
// ... code
FrameBlock fb = new FrameBlock(data.getNumColumns(), Types.ValueType.STRING);
fb.write(do);
{code}
Note that this is an internal error, so I am not sure if this can happen while executing a dml script.

By calling FrameBlock.ensureAllocatedColumns(rows) before writing this can be stopped from happening and I am not aware of instances where this lead to problems at the moment.
{code:java}
// ... code
//write columns (value type, data)
for( int j=0; j<getNumColumns(); j++ ) {
   out.writeByte(_schema[j].ordinal());
   if( !isDefaultMeta ) {
      out.writeUTF(getColumnName(j));
      out.writeLong(_colmeta[j].getNumDistinct());
      out.writeUTF( (_colmeta[j].getMvValue()!=null) ? 
            _colmeta[j].getMvValue() : "" );
   }
   _coldata[j].write(out);
}
{code}
The problem is the line:
{code:java}
   _coldata[j].write(out);
{code}
Although we have columns _coldata is still null.


> FrameBlock read and write
> -------------------------
>
>                 Key: SYSTEMDS-2565
>                 URL: https://issues.apache.org/jira/browse/SYSTEMDS-2565
>             Project: SystemDS
>          Issue Type: Bug
>            Reporter: Kevin Innerebner
>            Priority: Minor
>   Original Estimate: 2h
>  Remaining Estimate: 2h
>
> FrameBlocks fail (null pointer access) to be serialized in cases where their _coldata is null. This is reproducable:
> {code:java}
> // ... code
> FrameBlock fb = new FrameBlock(data.getNumColumns(), Types.ValueType.STRING);
> fb.write(do);
> {code}
> Note that this is an internal error, so I am not sure if this can happen while executing a dml script.
> By calling FrameBlock.ensureAllocatedColumns(rows) before writing this can be stopped from happening and I am not aware of instances where this lead to problems at the moment.
> The problem arises in the FrameBlock.write() method:
> {code:java}
> // ... code
> //write columns (value type, data)
> for( int j=0; j<getNumColumns(); j++ ) {
>    out.writeByte(_schema[j].ordinal());
>    if( !isDefaultMeta ) {
>       out.writeUTF(getColumnName(j));
>       out.writeLong(_colmeta[j].getNumDistinct());
>       out.writeUTF( (_colmeta[j].getMvValue()!=null) ? 
>             _colmeta[j].getMvValue() : "" );
>    }
>    _coldata[j].write(out);
> }
> {code}
> The problem is the line:
> {code:java}
>    _coldata[j].write(out);
> {code}
> Although we have columns _coldata is still null.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)