View Javadoc

1   /**
2    * Licensed to the Apache Software Foundation (ASF) under one
3    * or more contributor license agreements.  See the NOTICE file
4    * distributed with this work for additional information
5    * regarding copyright ownership.  The ASF licenses this file
6    * to you under the Apache License, Version 2.0 (the
7    * "License"); you may not use this file except in compliance
8    * with the License.  You may obtain a copy of the License at
9    *
10   *     http://www.apache.org/licenses/LICENSE-2.0
11   *
12   * Unless required by applicable law or agreed to in writing, software
13   * distributed under the License is distributed on an "AS IS" BASIS,
14   * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
15   * See the License for the specific language governing permissions and
16   * limitations under the License.
17   */
18  package org.apache.hadoop.hbase.regionserver;
19  
20  import java.util.List;
21  
22  import org.apache.commons.logging.Log;
23  import org.apache.commons.logging.LogFactory;
24  import org.apache.hadoop.hbase.HBaseInterfaceAudience;
25  import org.apache.hadoop.hbase.classification.InterfaceAudience;
26  import org.apache.hadoop.hbase.classification.InterfaceStability;
27  import org.apache.hadoop.hbase.client.metrics.ServerSideScanMetrics;
28  
29  /**
30   * ScannerContext instances encapsulate limit tracking AND progress towards those limits during
31   * invocations of {@link InternalScanner#next(java.util.List)} and
32   * {@link RegionScanner#next(java.util.List)}.
33   * <p>
34   * A ScannerContext instance should be updated periodically throughout execution whenever progress
35   * towards a limit has been made. Each limit can be checked via the appropriate checkLimit method.
36   * <p>
37   * Once a limit has been reached, the scan will stop. The invoker of
38   * {@link InternalScanner#next(java.util.List)} or {@link RegionScanner#next(java.util.List)} can
39   * use the appropriate check*Limit methods to see exactly which limits have been reached.
40   * Alternatively, {@link #checkAnyLimitReached(LimitScope)} is provided to see if ANY limit was
41   * reached
42   * <p>
43   * {@link NoLimitScannerContext#NO_LIMIT} is an immutable static definition that can be used
44   * whenever a {@link ScannerContext} is needed but limits do not need to be enforced.
45   * <p>
46   * NOTE: It is important that this class only ever expose setter methods that can be safely skipped
47   * when limits should be NOT enforced. This is because of the necessary immutability of the class
48   * {@link NoLimitScannerContext}. If a setter cannot be safely skipped, the immutable nature of
49   * {@link NoLimitScannerContext} will lead to incorrect behavior.
50   */
51  @InterfaceAudience.LimitedPrivate(HBaseInterfaceAudience.COPROC)
52  @InterfaceStability.Evolving
53  public class ScannerContext {
54    private static final Log LOG = LogFactory.getLog(ScannerContext.class);
55  
56    /**
57     * Two sets of the same fields. One for the limits, another for the progress towards those limits
58     */
59    LimitFields limits;
60    LimitFields progress;
61  
62    /**
63     * The state of the scanner after the invocation of {@link InternalScanner#next(java.util.List)}
64     * or {@link RegionScanner#next(java.util.List)}.
65     */
66    NextState scannerState;
67    private static final NextState DEFAULT_STATE = NextState.MORE_VALUES;
68  
69    /**
70     * Used as an indication to invocations of {@link InternalScanner#next(java.util.List)} and
71     * {@link RegionScanner#next(java.util.List)} that, if true, the progress tracked within this
72     * {@link ScannerContext} instance should be considered while evaluating the limits. Useful for
73     * enforcing a set of limits across multiple calls (i.e. the limit may not be reached in a single
74     * invocation, but any progress made should be considered in future invocations)
75     * <p>
76     * Defaulting this value to false means that, by default, any tracked progress will be wiped clean
77     * on invocations to {@link InternalScanner#next(java.util.List)} and
78     * {@link RegionScanner#next(java.util.List)} and the call will be treated as though no progress
79     * has been made towards the limits so far.
80     * <p>
81     * This is an important mechanism. Users of Internal/Region scanners expect that they can define
82     * some limits and then repeatedly invoke {@link InternalScanner#next(List)} or
83     * {@link RegionScanner#next(List)} where each invocation respects these limits separately.
84     * <p>
85     * For example: <code><pre>
86     * ScannerContext context = new ScannerContext.newBuilder().setBatchLimit(5).build();
87     * RegionScanner scanner = ...
88     * List<Cell> results = new ArrayList<Cell>();
89     * while(scanner.next(results, context)) {
90     *   // Do something with a batch of 5 cells
91     * }
92     * </pre></code> However, in the case of RPCs, the server wants to be able to define a set of
93     * limits for a particular RPC request and have those limits respected across multiple
94     * invocations. This means that the progress made towards the limits in earlier calls will be
95     * saved and considered in future invocations
96     */
97    boolean keepProgress;
98    private static boolean DEFAULT_KEEP_PROGRESS = false;
99  
100   /**
101    * Tracks the relevant server side metrics during scans. null when metrics should not be tracked
102    */
103   final ServerSideScanMetrics metrics;
104 
105   ScannerContext(boolean keepProgress, LimitFields limitsToCopy, boolean trackMetrics) {
106     this.limits = new LimitFields();
107     if (limitsToCopy != null) this.limits.copy(limitsToCopy);
108 
109     // Progress fields are initialized to 0
110     progress = new LimitFields(0, LimitFields.DEFAULT_SCOPE, 0, LimitFields.DEFAULT_SCOPE, 0);
111 
112     this.keepProgress = keepProgress;
113     this.scannerState = DEFAULT_STATE;
114     this.metrics = trackMetrics ? new ServerSideScanMetrics() : null;
115   }
116 
117   boolean isTrackingMetrics() {
118     return this.metrics != null;
119   }
120 
121   /**
122    * Get the metrics instance. Should only be called after a call to {@link #isTrackingMetrics()}
123    * has been made to confirm that metrics are indeed being tracked.
124    * @return {@link ServerSideScanMetrics} instance that is tracking metrics for this scan
125    */
126   ServerSideScanMetrics getMetrics() {
127     assert isTrackingMetrics();
128     return this.metrics;
129   }
130 
131   /**
132    * @return true if the progress tracked so far in this instance will be considered during an
133    *         invocation of {@link InternalScanner#next(java.util.List)} or
134    *         {@link RegionScanner#next(java.util.List)}. false when the progress tracked so far
135    *         should not be considered and should instead be wiped away via {@link #clearProgress()}
136    */
137   boolean getKeepProgress() {
138     return keepProgress;
139   }
140 
141   void setKeepProgress(boolean keepProgress) {
142     this.keepProgress = keepProgress;
143   }
144 
145   /**
146    * Progress towards the batch limit has been made. Increment internal tracking of batch progress
147    */
148   void incrementBatchProgress(int batch) {
149     int currentBatch = progress.getBatch();
150     progress.setBatch(currentBatch + batch);
151   }
152 
153   /**
154    * Progress towards the size limit has been made. Increment internal tracking of size progress
155    */
156   void incrementSizeProgress(long size) {
157     long currentSize = progress.getSize();
158     progress.setSize(currentSize + size);
159   }
160 
161   /**
162    * Update the time progress with {@link System#currentTimeMillis()}
163    */
164   void updateTimeProgress() {
165     progress.setTime(System.currentTimeMillis());
166   }
167 
168   int getBatchProgress() {
169     return progress.getBatch();
170   }
171 
172   long getSizeProgress() {
173     return progress.getSize();
174   }
175 
176   long getTimeProgress() {
177     return progress.getTime();
178   }
179 
180   void setProgress(int batchProgress, long sizeProgress, long timeProgress) {
181     setBatchProgress(batchProgress);
182     setSizeProgress(sizeProgress);
183     setTimeProgress(timeProgress);
184   }
185 
186   void setSizeProgress(long sizeProgress) {
187     progress.setSize(sizeProgress);
188   }
189 
190   void setBatchProgress(int batchProgress) {
191     progress.setBatch(batchProgress);
192   }
193 
194   void setTimeProgress(long timeProgress) {
195     progress.setTime(timeProgress);
196   }
197 
198   /**
199    * Clear away any progress that has been made so far. All progress fields are reset to initial
200    * values
201    */
202   void clearProgress() {
203     progress.setFields(0, LimitFields.DEFAULT_SCOPE, 0, LimitFields.DEFAULT_SCOPE, 0);
204   }
205 
206   /**
207    * Note that this is not a typical setter. This setter returns the {@link NextState} that was
208    * passed in so that methods can be invoked against the new state. Furthermore, this pattern
209    * allows the {@link NoLimitScannerContext} to cleanly override this setter and simply return the
210    * new state, thus preserving the immutability of {@link NoLimitScannerContext}
211    * @param state
212    * @return The state that was passed in.
213    */
214   NextState setScannerState(NextState state) {
215     if (!NextState.isValidState(state)) {
216       throw new IllegalArgumentException("Cannot set to invalid state: " + state);
217     }
218 
219     this.scannerState = state;
220     return state;
221   }
222 
223   /**
224    * @return true when a partial result is formed. A partial result is formed when a limit is
225    *         reached in the middle of a row.
226    */
227   boolean partialResultFormed() {
228     return scannerState == NextState.SIZE_LIMIT_REACHED_MID_ROW
229         || scannerState == NextState.TIME_LIMIT_REACHED_MID_ROW;
230   }
231 
232   /**
233    * @param checkerScope
234    * @return true if the batch limit can be enforced in the checker's scope
235    */
236   boolean hasBatchLimit(LimitScope checkerScope) {
237     return limits.canEnforceBatchLimitFromScope(checkerScope) && limits.getBatch() > 0;
238   }
239 
240   /**
241    * @param checkerScope
242    * @return true if the size limit can be enforced in the checker's scope
243    */
244   boolean hasSizeLimit(LimitScope checkerScope) {
245     return limits.canEnforceSizeLimitFromScope(checkerScope) && limits.getSize() > 0;
246   }
247 
248   /**
249    * @param checkerScope
250    * @return true if the time limit can be enforced in the checker's scope
251    */
252   boolean hasTimeLimit(LimitScope checkerScope) {
253     return limits.canEnforceTimeLimitFromScope(checkerScope) && limits.getTime() > 0;
254   }
255 
256   /**
257    * @param checkerScope
258    * @return true if any limit can be enforced within the checker's scope
259    */
260   boolean hasAnyLimit(LimitScope checkerScope) {
261     return hasBatchLimit(checkerScope) || hasSizeLimit(checkerScope) || hasTimeLimit(checkerScope);
262   }
263 
264   /**
265    * @param scope The scope in which the size limit will be enforced
266    */
267   void setSizeLimitScope(LimitScope scope) {
268     limits.setSizeScope(scope);
269   }
270 
271   /**
272    * @param scope The scope in which the time limit will be enforced
273    */
274   void setTimeLimitScope(LimitScope scope) {
275     limits.setTimeScope(scope);
276   }
277 
278   int getBatchLimit() {
279     return limits.getBatch();
280   }
281 
282   long getSizeLimit() {
283     return limits.getSize();
284   }
285 
286   long getTimeLimit() {
287     return limits.getTime();
288   }
289 
290   /**
291    * @param checkerScope The scope that the limit is being checked from
292    * @return true when the limit is enforceable from the checker's scope and it has been reached
293    */
294   boolean checkBatchLimit(LimitScope checkerScope) {
295     return hasBatchLimit(checkerScope) && progress.getBatch() >= limits.getBatch();
296   }
297 
298   /**
299    * @param checkerScope The scope that the limit is being checked from
300    * @return true when the limit is enforceable from the checker's scope and it has been reached
301    */
302   boolean checkSizeLimit(LimitScope checkerScope) {
303     return hasSizeLimit(checkerScope) && progress.getSize() >= limits.getSize();
304   }
305 
306   /**
307    * @param checkerScope The scope that the limit is being checked from. The time limit is always
308    *          checked against {@link System#currentTimeMillis()}
309    * @return true when the limit is enforceable from the checker's scope and it has been reached
310    */
311   boolean checkTimeLimit(LimitScope checkerScope) {
312     return hasTimeLimit(checkerScope) && progress.getTime() >= limits.getTime();
313   }
314 
315   /**
316    * @param checkerScope The scope that the limits are being checked from
317    * @return true when some limit is enforceable from the checker's scope and it has been reached
318    */
319   boolean checkAnyLimitReached(LimitScope checkerScope) {
320     return checkSizeLimit(checkerScope) || checkBatchLimit(checkerScope)
321         || checkTimeLimit(checkerScope);
322   }
323 
324   @Override
325   public String toString() {
326     StringBuilder sb = new StringBuilder();
327     sb.append("{");
328 
329     sb.append("limits:");
330     sb.append(limits);
331 
332     sb.append(", progress:");
333     sb.append(progress);
334 
335     sb.append(", keepProgress:");
336     sb.append(keepProgress);
337 
338     sb.append(", state:");
339     sb.append(scannerState);
340 
341     sb.append("}");
342     return sb.toString();
343   }
344 
345   public static Builder newBuilder() {
346     return new Builder();
347   }
348 
349   public static Builder newBuilder(boolean keepProgress) {
350     return new Builder(keepProgress);
351   }
352 
353   public static final class Builder {
354     boolean keepProgress = DEFAULT_KEEP_PROGRESS;
355     boolean trackMetrics = false;
356     LimitFields limits = new LimitFields();
357 
358     private Builder() {
359     }
360 
361     private Builder(boolean keepProgress) {
362       this.keepProgress = keepProgress;
363     }
364 
365     public Builder setKeepProgress(boolean keepProgress) {
366       this.keepProgress = keepProgress;
367       return this;
368     }
369 
370     public Builder setTrackMetrics(boolean trackMetrics) {
371       this.trackMetrics = trackMetrics;
372       return this;
373     }
374 
375     public Builder setSizeLimit(LimitScope sizeScope, long sizeLimit) {
376       limits.setSize(sizeLimit);
377       limits.setSizeScope(sizeScope);
378       return this;
379     }
380 
381     public Builder setTimeLimit(LimitScope timeScope, long timeLimit) {
382       limits.setTime(timeLimit);
383       limits.setTimeScope(timeScope);
384       return this;
385     }
386 
387     public Builder setBatchLimit(int batchLimit) {
388       limits.setBatch(batchLimit);
389       return this;
390     }
391 
392     public ScannerContext build() {
393       return new ScannerContext(keepProgress, limits, trackMetrics);
394     }
395   }
396 
397   /**
398    * The possible states a scanner may be in following a call to {@link InternalScanner#next(List)}
399    */
400   public enum NextState {
401     MORE_VALUES(true, false),
402     NO_MORE_VALUES(false, false),
403     SIZE_LIMIT_REACHED(true, true),
404 
405     /**
406      * Special case of size limit reached to indicate that the size limit was reached in the middle
407      * of a row and thus a partial results was formed
408      */
409     SIZE_LIMIT_REACHED_MID_ROW(true, true),
410     TIME_LIMIT_REACHED(true, true),
411 
412     /**
413      * Special case of time limit reached to indicate that the time limit was reached in the middle
414      * of a row and thus a partial results was formed
415      */
416     TIME_LIMIT_REACHED_MID_ROW(true, true),
417     BATCH_LIMIT_REACHED(true, true);
418 
419     private boolean moreValues;
420     private boolean limitReached;
421 
422     private NextState(boolean moreValues, boolean limitReached) {
423       this.moreValues = moreValues;
424       this.limitReached = limitReached;
425     }
426 
427     /**
428      * @return true when the state indicates that more values may follow those that have been
429      *         returned
430      */
431     public boolean hasMoreValues() {
432       return this.moreValues;
433     }
434 
435     /**
436      * @return true when the state indicates that a limit has been reached and scan should stop
437      */
438     public boolean limitReached() {
439       return this.limitReached;
440     }
441 
442     public static boolean isValidState(NextState state) {
443       return state != null;
444     }
445 
446     public static boolean hasMoreValues(NextState state) {
447       return isValidState(state) && state.hasMoreValues();
448     }
449   }
450 
451   /**
452    * The various scopes where a limit can be enforced. Used to differentiate when a limit should be
453    * enforced or not.
454    */
455   public enum LimitScope {
456     /**
457      * Enforcing a limit between rows means that the limit will not be considered until all the
458      * cells for a particular row have been retrieved
459      */
460     BETWEEN_ROWS(0),
461 
462     /**
463      * Enforcing a limit between cells means that the limit will be considered after each full cell
464      * has been retrieved
465      */
466     BETWEEN_CELLS(1);
467 
468     /**
469      * When enforcing a limit, we must check that the scope is appropriate for enforcement.
470      * <p>
471      * To communicate this concept, each scope has a depth. A limit will be enforced if the depth of
472      * the checker's scope is less than or equal to the limit's scope. This means that when checking
473      * limits, the checker must know their own scope (i.e. are they checking the limits between
474      * rows, between cells, etc...)
475      */
476     int depth;
477 
478     LimitScope(int depth) {
479       this.depth = depth;
480     }
481 
482     int depth() {
483       return depth;
484     }
485 
486     /**
487      * @param checkerScope The scope in which the limit is being checked
488      * @return true when the checker is in a scope that indicates the limit can be enforced. Limits
489      *         can be enforced from "higher or equal" scopes (i.e. the checker's scope is at a
490      *         lesser depth than the limit)
491      */
492     boolean canEnforceLimitFromScope(LimitScope checkerScope) {
493       return checkerScope != null && checkerScope.depth() <= depth;
494     }
495   }
496 
497   /**
498    * The different fields that can be used as limits in calls to
499    * {@link InternalScanner#next(java.util.List)} and {@link RegionScanner#next(java.util.List)}
500    */
501   private static class LimitFields {
502     /**
503      * Default values of the limit fields. Defined such that if a field does NOT change from its
504      * default, it will not be enforced
505      */
506     private static int DEFAULT_BATCH = -1;
507     private static long DEFAULT_SIZE = -1L;
508     private static long DEFAULT_TIME = -1L;
509 
510     /**
511      * Default scope that is assigned to a limit if a scope is not specified.
512      */
513     private static final LimitScope DEFAULT_SCOPE = LimitScope.BETWEEN_ROWS;
514 
515     // The batch limit will always be enforced between cells, thus, there isn't a field to hold the
516     // batch scope
517     int batch = DEFAULT_BATCH;
518 
519     LimitScope sizeScope = DEFAULT_SCOPE;
520     long size = DEFAULT_SIZE;
521 
522     LimitScope timeScope = DEFAULT_SCOPE;
523     long time = DEFAULT_TIME;
524 
525     /**
526      * Fields keep their default values.
527      */
528     LimitFields() {
529     }
530 
531     LimitFields(int batch, LimitScope sizeScope, long size, LimitScope timeScope, long time) {
532       setFields(batch, sizeScope, size, timeScope, time);
533     }
534 
535     void copy(LimitFields limitsToCopy) {
536       if (limitsToCopy != null) {
537         setFields(limitsToCopy.getBatch(), limitsToCopy.getSizeScope(), limitsToCopy.getSize(),
538           limitsToCopy.getTimeScope(), limitsToCopy.getTime());
539       }
540     }
541 
542     /**
543      * Set all fields together.
544      * @param batch
545      * @param sizeScope
546      * @param size
547      */
548     void setFields(int batch, LimitScope sizeScope, long size, LimitScope timeScope, long time) {
549       setBatch(batch);
550       setSizeScope(sizeScope);
551       setSize(size);
552       setTimeScope(timeScope);
553       setTime(time);
554     }
555 
556     int getBatch() {
557       return this.batch;
558     }
559 
560     void setBatch(int batch) {
561       this.batch = batch;
562     }
563 
564     /**
565      * @param checkerScope
566      * @return true when the limit can be enforced from the scope of the checker
567      */
568     boolean canEnforceBatchLimitFromScope(LimitScope checkerScope) {
569       return LimitScope.BETWEEN_CELLS.canEnforceLimitFromScope(checkerScope);
570     }
571 
572     long getSize() {
573       return this.size;
574     }
575 
576     void setSize(long size) {
577       this.size = size;
578     }
579 
580     /**
581      * @return {@link LimitScope} indicating scope in which the size limit is enforced
582      */
583     LimitScope getSizeScope() {
584       return this.sizeScope;
585     }
586 
587     /**
588      * Change the scope in which the size limit is enforced
589      */
590     void setSizeScope(LimitScope scope) {
591       this.sizeScope = scope;
592     }
593 
594     /**
595      * @param checkerScope
596      * @return true when the limit can be enforced from the scope of the checker
597      */
598     boolean canEnforceSizeLimitFromScope(LimitScope checkerScope) {
599       return this.sizeScope.canEnforceLimitFromScope(checkerScope);
600     }
601 
602     long getTime() {
603       return this.time;
604     }
605 
606     void setTime(long time) {
607       this.time = time;
608     }
609 
610     /**
611      * @return {@link LimitScope} indicating scope in which the time limit is enforced
612      */
613     LimitScope getTimeScope() {
614       return this.timeScope;
615     }
616 
617     /**
618      * Change the scope in which the time limit is enforced
619      */
620     void setTimeScope(LimitScope scope) {
621       this.timeScope = scope;
622     }
623 
624     /**
625      * @param checkerScope
626      * @return true when the limit can be enforced from the scope of the checker
627      */
628     boolean canEnforceTimeLimitFromScope(LimitScope checkerScope) {
629       return this.sizeScope.canEnforceLimitFromScope(checkerScope);
630     }
631 
632     @Override
633     public String toString() {
634       StringBuilder sb = new StringBuilder();
635       sb.append("{");
636 
637       sb.append("batch:");
638       sb.append(batch);
639 
640       sb.append(", size:");
641       sb.append(size);
642 
643       sb.append(", sizeScope:");
644       sb.append(sizeScope);
645 
646       sb.append(", time:");
647       sb.append(time);
648 
649       sb.append(", timeScope:");
650       sb.append(timeScope);
651 
652       sb.append("}");
653       return sb.toString();
654     }
655   }
656 }