use of org.apache.hadoop.hive.ql.exec.FooterBuffer in project hive by apache.
the class HiveContextAwareRecordReader method doNext.
public boolean doNext(K key, V value) throws IOException {
if (this.isSorted) {
if (this.getIOContext().shouldEndBinarySearch() || (!this.getIOContext().useSorted() && this.wasUsingSortedSearch)) {
beginLinearSearch();
this.wasUsingSortedSearch = false;
this.getIOContext().setEndBinarySearch(false);
}
if (this.getIOContext().useSorted()) {
if (this.genericUDFClassName == null && this.getIOContext().getGenericUDFClassName() != null) {
setGenericUDFClassName(this.getIOContext().getGenericUDFClassName());
}
if (this.getIOContext().isBinarySearching()) {
// Proceed with a binary search
if (this.getIOContext().getComparison() != null) {
switch(this.getIOContext().getComparison()) {
case GREATER:
case EQUAL:
// Indexes have only one entry per value, could go linear from here, if we want to
// use this for any sorted table, we'll need to continue the search
rangeEnd = previousPosition;
break;
case LESS:
rangeStart = previousPosition;
break;
default:
break;
}
}
long position = (rangeStart + rangeEnd) / 2;
sync(position);
long newPosition = getSyncedPosition();
// matching rows must be in the final block, so we can end the binary search.
if (newPosition == previousPosition || newPosition >= splitEnd) {
this.getIOContext().setBinarySearching(false);
sync(rangeStart);
}
previousPosition = newPosition;
} else if (foundAllTargets()) {
// Found all possible rows which will not be filtered
return false;
}
}
}
try {
/**
* When start reading new file, check header, footer rows.
* If file contains header, skip header lines before reading the records.
* If file contains footer, used a FooterBuffer to remove footer lines
* at the end of the table file.
*/
if (this.ioCxtRef.getCurrentBlockStart() == 0) {
// Check if the table file has header to skip.
footerBuffer = null;
Path filePath = this.ioCxtRef.getInputPath();
PartitionDesc part = null;
try {
if (pathToPartitionInfo == null) {
pathToPartitionInfo = Utilities.getMapWork(jobConf).getPathToPartitionInfo();
}
part = HiveFileFormatUtils.getFromPathRecursively(pathToPartitionInfo, filePath, IOPrepareCache.get().getPartitionDescMap());
} catch (AssertionError ae) {
LOG.info("Cannot get partition description from " + this.ioCxtRef.getInputPath() + "because " + ae.getMessage());
part = null;
} catch (Exception e) {
LOG.info("Cannot get partition description from " + this.ioCxtRef.getInputPath() + "because " + e.getMessage());
part = null;
}
TableDesc table = (part == null) ? null : part.getTableDesc();
if (table != null) {
headerCount = Utilities.getHeaderCount(table);
footerCount = Utilities.getFooterCount(table, jobConf);
}
// If input contains header, skip header.
if (!Utilities.skipHeader(recordReader, headerCount, (WritableComparable) key, (Writable) value)) {
return false;
}
if (footerCount > 0) {
footerBuffer = new FooterBuffer();
if (!footerBuffer.initializeBuffer(jobConf, recordReader, footerCount, (WritableComparable) key, (Writable) value)) {
return false;
}
}
}
if (footerBuffer == null) {
// Table files don't have footer rows.
return recordReader.next(key, value);
} else {
return footerBuffer.updateBuffer(jobConf, recordReader, (WritableComparable) key, (Writable) value);
}
} catch (Exception e) {
return HiveIOExceptionHandlerUtil.handleRecordReaderNextException(e, jobConf);
}
}
Aggregations