Example 1 with NextState
use of org.apache.hadoop.hbase.regionserver.ScannerContext.NextState in project hbase by apache.
the class StoreScanner method next.
/**
* Get the next row of values from this Store.
* @param outResult
* @param scannerContext
* @return true if there are more rows, false if scanner is done
*/
@Override
public boolean next(List<Cell> outResult, ScannerContext scannerContext) throws IOException {
if (scannerContext == null) {
throw new IllegalArgumentException("Scanner context cannot be null");
}
if (checkFlushed() && reopenAfterFlush()) {
return scannerContext.setScannerState(NextState.MORE_VALUES).hasMoreValues();
}
// return.
if (this.heap == null) {
// By this time partial close should happened because already heap is null
// Do all cleanup except heap.close()
close(false);
return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues();
}
Cell cell = this.heap.peek();
if (cell == null) {
// Do all cleanup except heap.close()
close(false);
return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues();
}
// comparison.
if (!scannerContext.hasAnyLimit(LimitScope.BETWEEN_CELLS) || matcher.currentRow() == null) {
this.countPerRow = 0;
matcher.setToNewRow(cell);
}
// Clear progress away unless invoker has indicated it should be kept.
if (!scannerContext.getKeepProgress()) {
scannerContext.clearProgress();
}
int count = 0;
long totalBytesRead = 0;
// track the cells for metrics only if it is a user read request.
boolean onlyFromMemstore = matcher.isUserScan();
try {
LOOP: do {
// the shipped method below.
if (kvsScanned % cellsPerHeartbeatCheck == 0 || (scanUsePread && readType == Scan.ReadType.DEFAULT && bytesRead > preadMaxBytes)) {
if (scannerContext.checkTimeLimit(LimitScope.BETWEEN_CELLS)) {
return scannerContext.setScannerState(NextState.TIME_LIMIT_REACHED).hasMoreValues();
}
}
// Do object compare - we set prevKV from the same heap.
if (prevCell != cell) {
++kvsScanned;
}
checkScanOrder(prevCell, cell, comparator);
int cellSize = PrivateCellUtil.estimatedSerializedSizeOf(cell);
bytesRead += cellSize;
if (scanUsePread && readType == Scan.ReadType.DEFAULT && bytesRead > preadMaxBytes) {
// return immediately if we want to switch from pread to stream. We need this because we
// can
// only switch in the shipped method, if user use a filter to filter out everything and
// rpc
// timeout is very large then the shipped method will never be called until the whole scan
// is finished, but at that time we have already scan all the data...
// See HBASE-20457 for more details.
// And there is still a scenario that can not be handled. If we have a very large row,
// which
// have millions of qualifiers, and filter.filterRow is used, then even if we set the flag
// here, we still need to scan all the qualifiers before returning...
scannerContext.returnImmediately();
}
prevCell = cell;
scannerContext.setLastPeekedCell(cell);
topChanged = false;
ScanQueryMatcher.MatchCode qcode = matcher.match(cell);
switch(qcode) {
case INCLUDE:
case INCLUDE_AND_SEEK_NEXT_ROW:
case INCLUDE_AND_SEEK_NEXT_COL:
Filter f = matcher.getFilter();
if (f != null) {
cell = f.transformCell(cell);
}
this.countPerRow++;
// also update metric accordingly
if (this.countPerRow > storeOffset) {
outResult.add(cell);
// Update local tracking information
count++;
totalBytesRead += cellSize;
/**
* Increment the metric if all the cells are from memstore.
* If not we will account it for mixed reads
*/
onlyFromMemstore = onlyFromMemstore && heap.isLatestCellFromMemstore();
// Update the progress of the scanner context
scannerContext.incrementSizeProgress(cellSize, cell.heapSize());
scannerContext.incrementBatchProgress(1);
if (matcher.isUserScan() && totalBytesRead > maxRowSize) {
String message = "Max row size allowed: " + maxRowSize + ", but the row is bigger than that, the row info: " + CellUtil.toString(cell, false) + ", already have process row cells = " + outResult.size() + ", it belong to region = " + store.getHRegion().getRegionInfo().getRegionNameAsString();
LOG.warn(message);
throw new RowTooBigException(message);
}
if (storeLimit > -1 && this.countPerRow >= (storeLimit + storeOffset)) {
// do what SEEK_NEXT_ROW does.
if (!matcher.moreRowsMayExistAfter(cell)) {
// Do all cleanup except heap.close()
close(false);
return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues();
}
matcher.clearCurrentRow();
seekToNextRow(cell);
break LOOP;
}
}
if (qcode == ScanQueryMatcher.MatchCode.INCLUDE_AND_SEEK_NEXT_ROW) {
if (!matcher.moreRowsMayExistAfter(cell)) {
// Do all cleanup except heap.close()
close(false);
return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues();
}
matcher.clearCurrentRow();
seekOrSkipToNextRow(cell);
} else if (qcode == ScanQueryMatcher.MatchCode.INCLUDE_AND_SEEK_NEXT_COL) {
seekOrSkipToNextColumn(cell);
} else {
this.heap.next();
}
if (scannerContext.checkBatchLimit(LimitScope.BETWEEN_CELLS)) {
break LOOP;
}
if (scannerContext.checkSizeLimit(LimitScope.BETWEEN_CELLS)) {
break LOOP;
}
continue;
case DONE:
// Optimization for Gets! If DONE, no more to get on this row, early exit!
if (get) {
// Then no more to this row... exit.
// Do all cleanup except heap.close()
close(false);
// update metric
return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues();
}
matcher.clearCurrentRow();
return scannerContext.setScannerState(NextState.MORE_VALUES).hasMoreValues();
case DONE_SCAN:
// Do all cleanup except heap.close()
close(false);
return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues();
case SEEK_NEXT_ROW:
// us if there is an endKey in the scan.
if (!matcher.moreRowsMayExistAfter(cell)) {
// Do all cleanup except heap.close()
close(false);
return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues();
}
matcher.clearCurrentRow();
seekOrSkipToNextRow(cell);
NextState stateAfterSeekNextRow = needToReturn(outResult);
if (stateAfterSeekNextRow != null) {
return scannerContext.setScannerState(stateAfterSeekNextRow).hasMoreValues();
}
break;
case SEEK_NEXT_COL:
seekOrSkipToNextColumn(cell);
NextState stateAfterSeekNextColumn = needToReturn(outResult);
if (stateAfterSeekNextColumn != null) {
return scannerContext.setScannerState(stateAfterSeekNextColumn).hasMoreValues();
}
break;
case SKIP:
this.heap.next();
break;
case SEEK_NEXT_USING_HINT:
Cell nextKV = matcher.getNextKeyHint(cell);
if (nextKV != null) {
int difference = comparator.compare(nextKV, cell);
if (((!scan.isReversed() && difference > 0) || (scan.isReversed() && difference < 0))) {
seekAsDirection(nextKV);
NextState stateAfterSeekByHint = needToReturn(outResult);
if (stateAfterSeekByHint != null) {
return scannerContext.setScannerState(stateAfterSeekByHint).hasMoreValues();
}
break;
}
}
heap.next();
break;
default:
throw new RuntimeException("UNEXPECTED");
}
} while ((cell = this.heap.peek()) != null);
if (count > 0) {
return scannerContext.setScannerState(NextState.MORE_VALUES).hasMoreValues();
}
// No more keys
// Do all cleanup except heap.close()
close(false);
return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues();
} finally {
// increment only if we have some result
if (count > 0 && matcher.isUserScan()) {
// if true increment memstore metrics, if not the mixed one
updateMetricsStore(onlyFromMemstore);
}
}
}