FileSystem fs;
try {
db = Hive.get(conf);
fs = FileSystem.get(conf);
createTableDesc crtTbl = work.getCreateTblDesc();
if (crtTbl != null) {
// create the table
Table tbl = new Table(crtTbl.getTableName());
tbl.setFields(crtTbl.getCols());
StorageDescriptor tblStorDesc = tbl.getTTable().getSd();
if (crtTbl.getBucketCols() != null)
tblStorDesc.setBucketCols(crtTbl.getBucketCols());
if (crtTbl.getSortCols() != null)
tbl.setSortCols(crtTbl.getSortCols());
if (crtTbl.getPartCols() != null)
tbl.setPartCols(crtTbl.getPartCols());
if (crtTbl.getNumBuckets() != -1)
tblStorDesc.setNumBuckets(crtTbl.getNumBuckets());
if (crtTbl.getSerName() != null) {
tbl.setSerializationLib(crtTbl.getSerName());
if (crtTbl.getMapProp() != null) {
Iterator<Map.Entry<String, String>> iter = crtTbl.getMapProp().entrySet().iterator();
while (iter.hasNext()) {
Map.Entry<String, String> m = (Map.Entry)iter.next();
tbl.setSerdeParam(m.getKey(), m.getValue());
}
}
}
else
{
if (crtTbl.getFieldDelim() != null)
{
tbl.setSerdeParam(Constants.FIELD_DELIM, crtTbl.getFieldDelim());
tbl.setSerdeParam(Constants.SERIALIZATION_FORMAT, crtTbl.getFieldDelim());
}
if (crtTbl.getCollItemDelim() != null)
tbl.setSerdeParam(Constants.COLLECTION_DELIM, crtTbl.getCollItemDelim());
if (crtTbl.getMapKeyDelim() != null)
tbl.setSerdeParam(Constants.MAPKEY_DELIM, crtTbl.getMapKeyDelim());
if (crtTbl.getLineDelim() != null)
tbl.setSerdeParam(Constants.LINE_DELIM, crtTbl.getLineDelim());
}
/**
* For now, if the user specifies either the map or the collections delimiter, we infer the
* table to DynamicSerDe/TCTLSeparatedProtocol.
* In the future, we should infer this for any delimiters specified, but this will break older
* hive tables, so not for now.
*/
if (crtTbl.getCollItemDelim() != null || crtTbl.getMapKeyDelim() != null) {
tbl.setSerializationLib(org.apache.hadoop.hive.serde2.dynamic_type.DynamicSerDe.class.getName());
tbl.setSerdeParam(org.apache.hadoop.hive.serde.Constants.SERIALIZATION_FORMAT, org.apache.hadoop.hive.serde2.thrift.TCTLSeparatedProtocol.class.getName());
}
if (crtTbl.getComment() != null)
tbl.setProperty("comment", crtTbl.getComment());
if (crtTbl.getLocation() != null)
tblStorDesc.setLocation(crtTbl.getLocation());
if (crtTbl.isSequenceFile()) {
tbl.setInputFormatClass(SequenceFileInputFormat.class);
tbl.setOutputFormatClass(SequenceFileOutputFormat.class);
}
else {
tbl.setOutputFormatClass(IgnoreKeyTextOutputFormat.class);
tbl.setInputFormatClass(TextInputFormat.class);
}
if (crtTbl.isExternal())
tbl.setProperty("EXTERNAL", "TRUE");
// If the sorted columns is a superset of bucketed columns, store this fact. It can be later used to
// optimize some group-by queries. Note that, the order does not matter as long as it in the first
// 'n' columns where 'n' is the length of the bucketed columns.