Github user joshelser commented on a diff in the pull request:
https://github.com/apache/accumulo/pull/42#discussion_r37206467
--- Diff:
core/src/main/java/org/apache/accumulo/core/iterators/user/SeekingFilter.java
---
@@ -0,0 +1,221 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.accumulo.core.iterators.user;
+
+import org.apache.accumulo.core.data.ByteSequence;
+import org.apache.accumulo.core.data.Key;
+import org.apache.accumulo.core.data.PartialKey;
+import org.apache.accumulo.core.data.Range;
+import org.apache.accumulo.core.data.Value;
+import org.apache.accumulo.core.iterators.IteratorEnvironment;
+import org.apache.accumulo.core.iterators.SortedKeyValueIterator;
+import org.apache.accumulo.core.iterators.WrappingIterator;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.io.IOException;
+import java.util.Collection;
+import java.util.EnumMap;
+import java.util.Map;
+
+/**
+ * Base class for filters that can skip over key-value pairs which do not
match their filter predicate. In addition to returning true/false to accept or
reject
+ * a kv pair, subclasses can return an extra field which indicates how far
the source iterator should be advanced.
+ *
+ * Note that the behaviour of the negate option is different from the
Filter class. If a KV pair fails the subclass' filter predicate and negate is
true, then
+ * the KV pair will pass the filter. However if the subclass advances the
source past a bunch of KV pairs, all those pairs will be implicitly rejected and
+ * negate will have no effect.
+ *
+ * @see org.apache.accumulo.core.iterators.Filter
+ */
+public abstract class SeekingFilter extends WrappingIterator {
+ private static final Logger log =
LoggerFactory.getLogger(SeekingFilter.class);
+
+ protected static final String NEGATE = "negate";
+
+ public enum AdvanceResult {
+ NEXT, NEXT_CQ, NEXT_CF, NEXT_ROW, USE_HINT
+ }
+
+ public static class FilterResult {
+ private static final EnumMap<AdvanceResult,FilterResult> PASSES = new
EnumMap<>(AdvanceResult.class);
+ private static final EnumMap<AdvanceResult,FilterResult> FAILS = new
EnumMap<>(AdvanceResult.class);
+ static {
+ for (AdvanceResult ar : AdvanceResult.values()) {
+ PASSES.put(ar, new FilterResult(true, ar));
+ FAILS.put(ar, new FilterResult(false, ar));
+ }
+ }
+
+ final boolean accept;
+ final AdvanceResult advance;
+
+ public FilterResult(boolean accept, AdvanceResult advance) {
+ this.accept = accept;
+ this.advance = advance;
+ }
+
+ public static FilterResult of(boolean accept, AdvanceResult advance) {
+ return accept ? PASSES.get(advance) : FAILS.get(advance);
+ }
+
+ public String toString() {
+ return "Acc: " + accept + " Adv: " + advance;
+ }
+ }
+
+ /**
+ * Subclasses must provide an implementation which examines the given
key and value and determines (1) whether to accept the KV pair and (2) how far
to
+ * advance the source iterator past the key.
+ *
+ * @param k
+ * a key
+ * @param v
+ * a value
+ * @return indicating whether to pass or block the key, and how far the
source iterator should be advanced.
+ */
+ public abstract FilterResult filter(Key k, Value v);
+
+ /**
+ * Whenever the subclass returns AdvanceResult.USE_HINT from its filter
predicate, this method will be called to see how far to advance the source
iterator.
+ * The return value must be a key which is greater than (sorts after)
the input key. If the subclass never returns USE_HINT, this method will never
be called
+ * and may safely return null.
+ *
+ * @param k
+ * a key
+ * @param v
+ * a value
+ * @return as above
+ */
+ public abstract Key getNextKeyHint(Key k, Value v);
+
+ private Collection<ByteSequence> columnFamilies;
+ private boolean inclusive;
+ private Range seekRange;
+ private boolean negate;
+
+ private AdvanceResult advance;
+
+ private boolean advancedPastSeek = false;
+
+ @Override
+ public void next() throws IOException {
+ findTop();
+ }
+
+ @Override
+ public boolean hasTop() {
+ return !advancedPastSeek && super.hasTop();
+ }
+
+ @Override
+ public void seek(Range range, Collection<ByteSequence> columnFamilies,
boolean inclusive) throws IOException {
+ super.seek(range, columnFamilies, inclusive);
+ advance = null;
+ this.columnFamilies = columnFamilies;
+ this.inclusive = inclusive;
+ seekRange = range;
+ advancedPastSeek = false;
+ findTop();
+ }
+
+ @Override
+ public void init(SortedKeyValueIterator<Key,Value> source,
Map<String,String> options, IteratorEnvironment env) throws IOException {
+ super.init(source, options, env);
+ negate = Boolean.parseBoolean(options.get(NEGATE));
+ }
+
+ @Override
+ public SortedKeyValueIterator<Key,Value> deepCopy(IteratorEnvironment
env) {
+ SeekingFilter newInstance;
+ try {
+ newInstance = this.getClass().newInstance();
+ } catch (Exception e) {
+ throw new RuntimeException(e);
+ }
+ newInstance.setSource(getSource().deepCopy(env));
+ newInstance.negate = negate;
+ return newInstance;
+ }
+
+ protected void findTop() throws IOException {
+ SortedKeyValueIterator<Key,Value> src = getSource();
+ // advance could be null if we've just been seeked
+ if (src.hasTop() && advance != null) {
+ advanceSource(src, advance);
+ }
+ advance = null;
+ while (src.hasTop() && !advancedPastSeek) {
+ if (src.getTopKey().isDeleted()) {
+ // as per. o.a.a.core.iterators.Filter, deleted keys always pass
through the filter.
+ advance = AdvanceResult.NEXT;
+ return;
+ }
+ FilterResult f = filter(src.getTopKey(), src.getTopValue());
+ if (log.isTraceEnabled()) {
+ log.trace("Filtered: " + src.getTopKey() + " result == " + f + "
hint == " + getNextKeyHint(src.getTopKey(), src.getTopValue()));
+ }
+ if (f.accept != negate) {
+ // advance will be processed next time findTop is called
+ advance = f.advance;
+ break;
+ } else {
+ advanceSource(src, f.advance);
+ }
+ }
+ }
+
+ private void advanceSource(SortedKeyValueIterator<Key,Value> src,
AdvanceResult adv) throws IOException {
+ Key topKey = src.getTopKey();
+ Range advRange = null;
+ switch (adv) {
+ case NEXT:
+ src.next();
+ return;
+ case NEXT_CQ:
+ advRange = new
Range(topKey.followingKey(PartialKey.ROW_COLFAM_COLQUAL), null);
+ break;
+ case NEXT_CF:
+ advRange = new Range(topKey.followingKey(PartialKey.ROW_COLFAM),
null);
+ break;
+ case NEXT_ROW:
+ advRange = new Range(topKey.followingKey(PartialKey.ROW), null);
+ break;
+ case USE_HINT:
+ Value topVal = src.getTopValue();
+ Key hintKey = getNextKeyHint(topKey, topVal);
+ if (hintKey != null && hintKey.compareTo(topKey) > 0) {
+ advRange = new Range(hintKey, null);
+ } else {
+ String msg = "Filter returned USE_HINT for " + topKey + " but
invalid hint: " + hintKey;
+ throw new IOException(msg);
+ }
+ break;
+ }
+ if (advRange == null) {
+ // Should never get here. Just a safeguard in case somebody adds a
new type of AdvanceRange and forgets to handle it here.
+ throw new IOException("Unable to determine range to advance to for
AdvanceResult " + adv);
+ }
+ advRange = advRange.clip(seekRange, true);
+ if (advRange == null) {
+ // the advanced range is outside the seek range. the source is
exhausted.
+ advancedPastSeek = true;
+ } else {
+ src.seek(advRange, columnFamilies, inclusive);
--- End diff --
As an optimization later on (mentioning it now before I forget), it _may_
be more efficient to try to encode a lookahead of 10s of key-values instead of
always seeking. This is intrinsically done for you in the stack already, but,
if memory serves, you still benefit from encoding that logic here (as there are
some costs getting to the optimization lower in the stack). Not something that
needs to be changed here, just might be a fun experiment.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
