Explaining NullPointerException

[John Kristian <jk...@linkedin.com>]

When one writes an erroneous null datum, it would be helpful for the stack
trace to say more about what happened; for example:

java.lang.NullPointerException: MyRecord.myField
    at ...
Caused by: java.lang.NullPointerException: string
    at ...

Is someone working to implement something like this?

One can make a subclass of SpecificDatumWriter to do this, but one must
copy-n-paste GenericDatumWriter.writeRecord.  It could be done without
copy-n-pasting, if GenericDatumWriter.writeRecord were split into two
methods:

  protected void writeRecord(Schema schema, Object datum, Encoder out)
    throws IOException {
    for (Field field : schema.getFields()) {
      writeField(schema, datum, field, out);
    }
  }

  protected void writeField(Schema schema, Object datum, Field field,
Encoder out)
    throws IOException {
    write(field.schema(), getField(datum, field.name(), field.pos()), out);
  }

Then the subclass could override two methods:

  @Override
  protected void writeField(Schema schema, Object datum, Field field,
Encoder out)
    throws IOException {
    try {
      super.writeField(schema, datum, field, out);
    } catch (NullPointerException e)  {
      NullPointerException e2 =
        new NullPointerException(schema.getName() + "." + field.name());
      e2.initCause(e);
      throw e2;
    }
  }

  @Override
  protected void write(Schema schema, Object datum, Encoder out)
    throws IOException {
    switch (schema.getType()) {
    case UNION:
    case NULL:
      break;
    default:
      if (datum == null)
        throw new NullPointerException(schema.getName());
    }
    super.write(schema, datum, out);
  }

- John Kristian

Re: Explaining NullPointerException

[Jeff Hammerbacher <ha...@cloudera.com>]

Hey John

Please file a JIRA at https://issues.apache.org/jira/browse/AVRO requesting
that this error message be clarified, and we'll get to it right away (see
https://issues.apache.org/jira/browse/AVRO-583, for example, for similar
such clarifications requested and implemented).

Thanks,
Jeff

On Fri, Aug 13, 2010 at 5:39 PM, John Kristian <jk...@linkedin.com>wrote:

> When one writes an erroneous null datum, it would be helpful for the stack
> trace to say more about what happened; for example:
>
> java.lang.NullPointerException: MyRecord.myField
>    at ...
> Caused by: java.lang.NullPointerException: string
>    at ...
>
> Is someone working to implement something like this?
>
> One can make a subclass of SpecificDatumWriter to do this, but one must
> copy-n-paste GenericDatumWriter.writeRecord.  It could be done without
> copy-n-pasting, if GenericDatumWriter.writeRecord were split into two
> methods:
>
>  protected void writeRecord(Schema schema, Object datum, Encoder out)
>    throws IOException {
>    for (Field field : schema.getFields()) {
>      writeField(schema, datum, field, out);
>    }
>  }
>
>  protected void writeField(Schema schema, Object datum, Field field,
> Encoder out)
>    throws IOException {
>    write(field.schema(), getField(datum, field.name(), field.pos()), out);
>  }
>
> Then the subclass could override two methods:
>
>  @Override
>  protected void writeField(Schema schema, Object datum, Field field,
> Encoder out)
>    throws IOException {
>    try {
>      super.writeField(schema, datum, field, out);
>    } catch (NullPointerException e)  {
>      NullPointerException e2 =
>        new NullPointerException(schema.getName() + "." + field.name());
>      e2.initCause(e);
>      throw e2;
>    }
>  }
>
>  @Override
>  protected void write(Schema schema, Object datum, Encoder out)
>    throws IOException {
>    switch (schema.getType()) {
>    case UNION:
>    case NULL:
>      break;
>    default:
>      if (datum == null)
>        throw new NullPointerException(schema.getName());
>    }
>    super.write(schema, datum, out);
>  }
>
> - John Kristian
>
>