@Override
public void setInput(Configuration conf, Map<String, Block> input, JsonNode json)
throws IOException, InterruptedException {
// #1. input block
inputBlock = (RubixMemoryBlock) input.get(JsonUtils.getText(json, "inputBlock"));
// #2. lookup column
String lookupColumn = json.get("lookupColumn").getTextValue();
BlockSchema inputSchema = inputBlock.getProperties().getSchema();
coord2offsets = BlockUtils.generateColumnIndex(inputBlock, lookupColumn);
// #3. meta data relation name
metaRelationName = new String(JsonUtils.getText(json, "metaRelationName"));
matchingMetaBlock = (Block) input.get(metaRelationName);
BlockSchema metaBlockSchema = matchingMetaBlock.getProperties().getSchema();
// #4. find indexes for coordinate column names in meta relation's schema
String[] coordinateColumns = JsonUtils.asArray(json.get("coordinateColumns"));
coordinateColumnIndexes = new int[coordinateColumns.length];
int idx = 0;
for (String s : JsonUtils.asArray(json.get("coordinateColumns")))
coordinateColumnIndexes[idx++] = metaBlockSchema.getIndex(s);
// #5. find index of identifier column in meta relation's schema
identifierColumnName = new String(JsonUtils.getText(json, "identifierColumn"));
identifierColumnIndex = metaBlockSchema.getIndex(identifierColumnName);
// #6. combine columns
ArrayNode combineColumns = (ArrayNode) json.get("combineColumns");
// setup info for sort operator
/*
* jsonForSort = JsonUtils.cloneNode(json); ((ObjectNode)
* jsonForSort).put("sortBy", combineColumns); sortedBlock = new
* TupleOperatorBlock(sortOp);
*/
// setup info for combiner operator
jsonForCombine = JsonUtils.createObjectNode();
((ObjectNode) jsonForCombine).put("pivotBy", combineColumns);
((ObjectNode) jsonForCombine).put("schema", inputSchema.toJson());
combinedBlock = new TupleOperatorBlock(combineOp, null);
// setup info for generate operator
jsonForGenerate = JsonUtils.createObjectNode();
}
@Override
public PostCondition getPostCondition(Map<String, PostCondition> preConditions, JsonNode json)
throws PreconditionException {
String inputBlockName = JsonUtils.getText(json, "input");
PostCondition inputCondition = preConditions.get(inputBlockName);
BlockSchema inputSchema = inputCondition.getSchema();
Map<String, CodeDictionary> dictionaryMap = new HashMap<String, CodeDictionary>();
if (json.has("columns")) {
String[] columns = JsonUtils.asArray(json, "columns");
for (String column : columns) dictionaryMap.put(column, new CodeDictionary());
} else {
JsonNode dictionary = json.get("dictionary");
// this is inline dictionary
Iterator<String> nameIterator = dictionary.getFieldNames();
while (nameIterator.hasNext()) {
String name = nameIterator.next();
ArrayNode values = (ArrayNode) dictionary.get(name);
CodeDictionary codeDictionary = new CodeDictionary();
for (JsonNode value : values) {
codeDictionary.addKey(value.getTextValue());
}
dictionaryMap.put(name, codeDictionary);
}
}
int numColumns = inputSchema.getNumColumns();
ColumnType[] columnTypes = new ColumnType[numColumns];
for (int i = 0; i < columnTypes.length; i++) {
ColumnType type;
final String name = inputSchema.getName(i);
if (dictionaryMap.containsKey(name)) {
// this column is decoded. Transform schema
type = new ColumnType(name, DataType.STRING);
} else {
// this column is not decoded. Reuse schema
type = inputSchema.getColumnType(i);
}
columnTypes[i] = type;
}
BlockSchema schema = new BlockSchema(columnTypes);
return new PostCondition(
schema, inputCondition.getPartitionKeys(), inputCondition.getSortKeys());
}
public String[] getSortKeys() throws IOException, ClassNotFoundException {
if (keyData == null) getKeyData();
return JsonUtils.asArray(metadataJson.get("sortKeys"));
}
