drill-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From paul-rogers <...@git.apache.org>
Subject [GitHub] drill pull request #805: Drill-4139: Exception while trying to prune partiti...
Date Tue, 11 Jul 2017 00:17:11 GMT
Github user paul-rogers commented on a diff in the pull request:

    https://github.com/apache/drill/pull/805#discussion_r126567484
  
    --- Diff: exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/ParquetGroupScan.java
---
    @@ -548,23 +567,57 @@ public void populatePruningVector(ValueVector v, int index, SchemaPath
column, S
             NullableVarCharVector varCharVector = (NullableVarCharVector) v;
             Object s = partitionValueMap.get(f).get(column);
             byte[] bytes;
    -        if (s instanceof String) { // if the metadata was read from a JSON cache file
it maybe a string type
    -          bytes = ((String) s).getBytes();
    -        } else if (s instanceof Binary) {
    -          bytes = ((Binary) s).getBytes();
    -        } else if (s instanceof byte[]) {
    -          bytes = (byte[]) s;
    +        if (s == null) {
    +          varCharVector.getMutator().setNull(index);
    +          return;
             } else {
    -          throw new UnsupportedOperationException("Unable to create column data for type:
" + type);
    +          bytes = getBytes(type, s);
             }
             varCharVector.getMutator().setSafe(index, bytes, 0, bytes.length);
             return;
           }
    +      case INTERVAL: {
    +        NullableIntervalVector intervalVector = (NullableIntervalVector) v;
    +        Object s = partitionValueMap.get(f).get(column);
    +        byte[] bytes;
    +        if (s == null) {
    +          intervalVector.getMutator().setNull(index);
    +          return;
    +        } else {
    +          bytes = getBytes(type, s);
    +        }
    +        intervalVector.getMutator().setSafe(index, 1,
    +          ParquetReaderUtility.getIntFromLEBytes(bytes, 0),
    +          ParquetReaderUtility.getIntFromLEBytes(bytes, 4),
    +          ParquetReaderUtility.getIntFromLEBytes(bytes, 8));
    +        return;
    +      }
           default:
             throw new UnsupportedOperationException("Unsupported type: " + type);
         }
       }
     
    +  /**
    +   * Returns the sequence of bytes received from {@code Object source}.
    +   *
    +   * @param type   the column type
    +   * @param source the source of the bytes sequence
    +   * @return bytes sequence obtained from {@code Object source}
    +   */
    +  private byte[] getBytes(MinorType type, Object source) {
    +    byte[] bytes;
    +    if (source instanceof String) { // if the metadata was read from a JSON cache file
it maybe a string type
    +      bytes = Base64.decodeBase64(((String) source).getBytes());
    --- End diff --
    
    The data here should be assumed to be binary. Binary can have any characters, even JSON
symbols. For example:
    
    "value:" : "look a quote",\n"bogus" : { 10, "foo" }"
    
    The above will make a mash of any JSON structure. Better to encode binary data as Base
64, which is designed to fit within a text format.
    
    Note, however, that the casting is unfortunate. The caller knows the type. Better to have:
    
    byte[] getBytes(String value) ...
    byte[] getBytes(byte[] value) ...
    byte[] getBytes(Binary value) ...
    {code}
    
    Seems we'd also need a to/from Base64 method so that, on read, we do:
    
    * Get string value.
    * Pass string value to Base64 decoder, get bytes
    * Pass bytes to vector, using vector's code to understand byte meaning
    
    On encoding:
    
    * Get the vector value as a byte array
    * Base64 encode the array
    * Write it to JSON as a string
    
    All that said, Strings need not be Base64 encoded if we have a JSON encoder that will
escape JSON special characters. This works as long as the JSON file itself is UTF-8 encoded
because then each Unicode string has a UTF-8 equivalent.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

Mime
View raw message