andi/perfdb
1
0
Fork 0
Code Issues 1 Pull Requests Packages Projects Releases Activity

performance improvement

Browse Source 
Queries like "firstname=John and lastname=???" were slightly
inefficient.
They fetched all firstnames, filtered to those that matched the prefix
(e.g. John or Jonathan is this example) and then iterated over all those
values and return the lastnames.
Fixed by having two implementations. One for the case that only a few
of the values in fieldA match and one for the case that many match.
This commit is contained in:
Andreas Huber
2019-08-31 19:30:54 +02:00
parent d8a114dbaf
commit 8579974051
2 changed files with 25 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
data-store/src/main/java/org/lucares/pdb/datastore/internal/QueryCompletionIndex.java
View File

@@ -342,18 +342,10 @@ public class QueryCompletionIndex implements AutoCloseable {
*
* @param dateRange the date range
* @param fieldA the other field of the and expression
* @param valueA the value of the other field
* @param valueA {@link GlobMatcher} for the value of the other field
* @param fieldB the field we are searching values for
* @return values of fieldB
*/
public SortedSet<String> find(final DateTimeRange dateRange, final String fieldA, final String valueA,
final String fieldB) {
final Tag tag = new Tag(fieldA, valueA);
Preconditions.checkGreaterOrEqual(tag.getKey(), 0, "The field ''{0}'' is unkown", fieldA);
Preconditions.checkGreaterOrEqual(tag.getValue(), 0, "The value ''{0}'' is unkown", valueA);
return find(dateRange, tag, fieldB);
}
public SortedSet<String> find(final DateTimeRange dateRange, final String fieldA, final GlobMatcher valueA,
final String fieldB) {

31
data-store/src/main/java/org/lucares/pdb/datastore/lang/FindValuesForQueryCompletion.java
View File

@@ -43,23 +43,40 @@ public class FindValuesForQueryCompletion extends ExpressionVisitor<SortedSet<St
@Override
public SortedSet<String> visit(final Property property) {
final long start = System.nanoTime();
final TreeSet<String> result = new TreeSet<>();
final SortedSet<String> result;
final String fieldA = property.getField();
final String valueA = property.getValue().getValue();
final boolean hasField = index.hasField(dateTimeRange, fieldA);
if (hasField) {
final SortedSet<String> allValuesForField = index.findAllValuesForField(dateTimeRange, fieldA);
final SortedSet<String> valuesA = GloblikePattern.filterValues(allValuesForField, valueA, TreeSet::new);
for (final String v : valuesA) {
final SortedSet<String> tmp = index.find(dateTimeRange, fieldA, v, field);
result.addAll(tmp);
}
}
final double valueInFieldAMatchPercentage = valuesA.size() / (double) allValuesForField.size();
final boolean useMultiFetch = valuesA.size() <= 1 || valueInFieldAMatchPercentage < 0.5; // 50% was
// chosen
// arbitrarily
if (useMultiFetch) {
result = new TreeSet<>();
METRIC_AND_CARET_LOGGER.debug("{}: {}ms", property, (System.nanoTime() - start) / 1_000_000.0);
for (final String v : valuesA) {
final Tag tagA = new Tag(fieldA, v);
final SortedSet<String> tmp = index.find(dateTimeRange, tagA, field);
result.addAll(tmp);
}
} else {
result = index.find(dateTimeRange, fieldA, new GlobMatcher(valueA), field);
}
METRIC_AND_CARET_LOGGER.debug("{}: {} and {}=???: {}ms matches in fieldA {} ({}%)",
useMultiFetch ? "multi-fetch" : "single-fetch", property, field,
(System.nanoTime() - start) / 1_000_000.0, valuesA.size(), valueInFieldAMatchPercentage * 100);
} else {
result = new TreeSet<>();
}
return result;
}