Skip to content

GH-14909: [Java] Prevent potential memory leak of ListSubfieldEncoder and StructSubfieldEncoder #14910

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
lidavidm merged 2 commits into from
Dec 12, 2022
Merged

GH-14909: [Java] Prevent potential memory leak of ListSubfieldEncoder and StructSubfieldEncoder #14910

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
315b831
prevent potential memory leak
gf2121 Dec 11, 2022
e1f9f16
review iter
gf2121 Dec 12, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
58 changes: 34 additions & 24 deletions java/vector/src/main/java/org/apache/arrow/vector/dictionary/ListSubfieldEncoder.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -85,20 +85,25 @@ public BaseListVector encodeListSubField(BaseListVector vector) {

// clone list vector and initialize data vector
BaseListVector encoded = cloneVector(vector, allocator);
encoded.initializeChildrenFromFields(Collections.singletonList(valueField));
BaseIntVector indices = (BaseIntVector) getDataVector(encoded);

ValueVector dataVector = getDataVector(vector);
for (int i = 0; i < valueCount; i++) {
if (!vector.isNull(i)) {
int start = vector.getElementStartIndex(i);
int end = vector.getElementEndIndex(i);

DictionaryEncoder.buildIndexVector(dataVector, indices, hashTable, start, end);
try {
encoded.initializeChildrenFromFields(Collections.singletonList(valueField));
BaseIntVector indices = (BaseIntVector) getDataVector(encoded);

ValueVector dataVector = getDataVector(vector);
for (int i = 0; i < valueCount; i++) {
if (!vector.isNull(i)) {
int start = vector.getElementStartIndex(i);
int end = vector.getElementEndIndex(i);

DictionaryEncoder.buildIndexVector(dataVector, indices, hashTable, start, end);
}
}
}

return encoded;
return encoded;
} catch (Exception e) {
encoded.close();
throw e;
}
}

/**
Expand Down Expand Up @@ -132,24 +137,29 @@ public static BaseListVector decodeListSubField(BaseListVector vector,

// clone list vector and initialize data vector
BaseListVector decoded = cloneVector(vector, allocator);
Field dataVectorField = getDataVector(dictionaryVector).getField();
decoded.initializeChildrenFromFields(Collections.singletonList(dataVectorField));
try {
Field dataVectorField = getDataVector(dictionaryVector).getField();
decoded.initializeChildrenFromFields(Collections.singletonList(dataVectorField));

// get data vector
ValueVector dataVector = getDataVector(decoded);
// get data vector
ValueVector dataVector = getDataVector(decoded);

TransferPair transfer = getDataVector(dictionaryVector).makeTransferPair(dataVector);
BaseIntVector indices = (BaseIntVector) getDataVector(vector);
TransferPair transfer = getDataVector(dictionaryVector).makeTransferPair(dataVector);
BaseIntVector indices = (BaseIntVector) getDataVector(vector);

for (int i = 0; i < valueCount; i++) {
for (int i = 0; i < valueCount; i++) {

if (!vector.isNull(i)) {
int start = vector.getElementStartIndex(i);
int end = vector.getElementEndIndex(i);
if (!vector.isNull(i)) {
int start = vector.getElementStartIndex(i);
int end = vector.getElementEndIndex(i);

DictionaryEncoder.retrieveIndexVector(indices, transfer, dictionaryValueCount, start, end);
DictionaryEncoder.retrieveIndexVector(indices, transfer, dictionaryValueCount, start, end);
}
}
return decoded;
} catch (Exception e) {
decoded.close();
throw e;

@lidavidm lidavidm Dec 12, 2022

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Perhaps use AutoCloseables#close(Throwable, AutoCloseable...) for this kind of thing?

}
return decoded;
}
}
94 changes: 52 additions & 42 deletions java/vector/src/main/java/org/apache/arrow/vector/dictionary/StructSubfieldEncoder.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -118,23 +118,28 @@ public StructVector encode(StructVector vector, Map<Integer, Long> columnToDicti

// clone list vector and initialize data vector
StructVector encoded = cloneVector(vector, allocator);
encoded.initializeChildrenFromFields(childrenFields);
encoded.setValueCount(valueCount);

for (int index = 0; index < childCount; index++) {
FieldVector childVector = getChildVector(vector, index);
FieldVector encodedChildVector = getChildVector(encoded, index);
Long dictionaryId = columnToDictionaryId.get(index);
if (dictionaryId != null) {
BaseIntVector indices = (BaseIntVector) encodedChildVector;
DictionaryEncoder.buildIndexVector(childVector, indices, dictionaryIdToHashTable.get(dictionaryId),
0, valueCount);
} else {
childVector.makeTransferPair(encodedChildVector).splitAndTransfer(0, valueCount);
try {
encoded.initializeChildrenFromFields(childrenFields);
encoded.setValueCount(valueCount);

for (int index = 0; index < childCount; index++) {
FieldVector childVector = getChildVector(vector, index);
FieldVector encodedChildVector = getChildVector(encoded, index);
Long dictionaryId = columnToDictionaryId.get(index);
if (dictionaryId != null) {
BaseIntVector indices = (BaseIntVector) encodedChildVector;
DictionaryEncoder.buildIndexVector(childVector, indices, dictionaryIdToHashTable.get(dictionaryId),
0, valueCount);
} else {
childVector.makeTransferPair(encodedChildVector).splitAndTransfer(0, valueCount);
}
}
}

return encoded;
return encoded;
} catch (Exception e) {
encoded.close();
throw e;
}
}

/**
Expand Down Expand Up @@ -167,36 +172,41 @@ public static StructVector decode(StructVector vector,

// clone list vector and initialize child vectors
StructVector decoded = cloneVector(vector, allocator);
List<Field> childFields = new ArrayList<>();
for (int i = 0; i < childCount; i++) {
FieldVector childVector = getChildVector(vector, i);
Dictionary dictionary = getChildVectorDictionary(childVector, provider);
// childVector is not encoded.
if (dictionary == null) {
childFields.add(childVector.getField());
} else {
childFields.add(dictionary.getVector().getField());
try {
List<Field> childFields = new ArrayList<>();
for (int i = 0; i < childCount; i++) {
FieldVector childVector = getChildVector(vector, i);
Dictionary dictionary = getChildVectorDictionary(childVector, provider);
// childVector is not encoded.
if (dictionary == null) {
childFields.add(childVector.getField());
} else {
childFields.add(dictionary.getVector().getField());
}
}
}
decoded.initializeChildrenFromFields(childFields);
decoded.setValueCount(valueCount);

for (int index = 0; index < childCount; index++) {
// get child vector
FieldVector childVector = getChildVector(vector, index);
FieldVector decodedChildVector = getChildVector(decoded, index);
Dictionary dictionary = getChildVectorDictionary(childVector, provider);
if (dictionary == null) {
childVector.makeTransferPair(decodedChildVector).splitAndTransfer(0, valueCount);
} else {
TransferPair transfer = dictionary.getVector().makeTransferPair(decodedChildVector);
BaseIntVector indices = (BaseIntVector) childVector;

DictionaryEncoder.retrieveIndexVector(indices, transfer, valueCount, 0, valueCount);
decoded.initializeChildrenFromFields(childFields);
decoded.setValueCount(valueCount);

for (int index = 0; index < childCount; index++) {
// get child vector
FieldVector childVector = getChildVector(vector, index);
FieldVector decodedChildVector = getChildVector(decoded, index);
Dictionary dictionary = getChildVectorDictionary(childVector, provider);
if (dictionary == null) {
childVector.makeTransferPair(decodedChildVector).splitAndTransfer(0, valueCount);
} else {
TransferPair transfer = dictionary.getVector().makeTransferPair(decodedChildVector);
BaseIntVector indices = (BaseIntVector) childVector;

DictionaryEncoder.retrieveIndexVector(indices, transfer, valueCount, 0, valueCount);
}
}
}

return decoded;
return decoded;
} catch (Exception e) {
decoded.close();
throw e;
}
}

/**
Expand Down
126 changes: 126 additions & 0 deletions java/vector/src/test/java/org/apache/arrow/vector/TestDictionaryVector.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -916,6 +916,132 @@ public void testNoMemoryLeak() {
assertEquals("decode memory leak", 0, allocator.getAllocatedMemory());
}

@Test
public void testListNoMemoryLeak() {
// Create a new value vector
try (final ListVector vector = ListVector.empty("vector", allocator);
final ListVector dictionaryVector = ListVector.empty("dict", allocator)) {

UnionListWriter writer = vector.getWriter();
writer.allocate();
writeListVector(writer, new int[]{10, 20});
writer.setValueCount(1);

UnionListWriter dictWriter = dictionaryVector.getWriter();
dictWriter.allocate();
writeListVector(dictWriter, new int[]{10});
dictionaryVector.setValueCount(1);

Dictionary dictionary = new Dictionary(dictionaryVector, new DictionaryEncoding(1L, false, null));
ListSubfieldEncoder encoder = new ListSubfieldEncoder(dictionary, allocator);

try (final ListVector encoded = (ListVector) encoder.encodeListSubField(vector)) {
fail("There should be an exception when encoding");
} catch (Exception e) {
assertEquals("Dictionary encoding not defined for value:" + 20, e.getMessage());

@lidavidm lidavidm Dec 12, 2022

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: why concat here, when the string is fully hardcoded below?

}
}
assertEquals("list encode memory leak", 0, allocator.getAllocatedMemory());

try (final ListVector indices = ListVector.empty("indices", allocator);
final ListVector dictionaryVector = ListVector.empty("dict", allocator)) {

UnionListWriter writer = indices.getWriter();
writer.allocate();
writeListVector(writer, new int[]{3});
writer.setValueCount(1);

UnionListWriter dictWriter = dictionaryVector.getWriter();
dictWriter.allocate();
writeListVector(dictWriter, new int[]{10, 20});
dictionaryVector.setValueCount(1);

Dictionary dictionary =
new Dictionary(dictionaryVector, new DictionaryEncoding(1L, false, null));

try (final ValueVector decoded = ListSubfieldEncoder.decodeListSubField(indices, dictionary, allocator)) {
fail("There should be an exception when decoding");
} catch (Exception e) {
assertEquals("Provided dictionary does not contain value for index 3", e.getMessage());
}
}
assertEquals("list decode memory leak", 0, allocator.getAllocatedMemory());
}

@Test
public void testStructNoMemoryLeak() {
try (final StructVector vector = StructVector.empty("vector", allocator);
final VarCharVector dictVector1 = new VarCharVector("f0", allocator);
final VarCharVector dictVector2 = new VarCharVector("f1", allocator)) {

vector.addOrGet("f0", FieldType.nullable(ArrowType.Utf8.INSTANCE), VarCharVector.class);
vector.addOrGet("f1", FieldType.nullable(ArrowType.Utf8.INSTANCE), VarCharVector.class);

NullableStructWriter writer = vector.getWriter();
writer.allocate();
writeStructVector(writer, "aa", "baz");
writer.setValueCount(1);

DictionaryProvider.MapDictionaryProvider provider = new DictionaryProvider.MapDictionaryProvider();
setVector(dictVector1,
"aa".getBytes(StandardCharsets.UTF_8));
setVector(dictVector2,
"foo".getBytes(StandardCharsets.UTF_8));

provider.put(new Dictionary(dictVector1, new DictionaryEncoding(1L, false, null)));
provider.put(new Dictionary(dictVector2, new DictionaryEncoding(2L, false, null)));

StructSubfieldEncoder encoder = new StructSubfieldEncoder(allocator, provider);
Map<Integer, Long> columnToDictionaryId = new HashMap<>();
columnToDictionaryId.put(0, 1L);
columnToDictionaryId.put(1, 2L);

try (final StructVector encoded = (StructVector) encoder.encode(vector, columnToDictionaryId)) {
fail("There should be an exception when encoding");
} catch (Exception e) {
assertEquals("Dictionary encoding not defined for value:" + "baz", e.getMessage());
}
}
assertEquals("struct encode memory leak", 0, allocator.getAllocatedMemory());

try (final StructVector indices = StructVector.empty("indices", allocator);
final VarCharVector dictVector1 = new VarCharVector("f0", allocator);
final VarCharVector dictVector2 = new VarCharVector("f1", allocator)) {

DictionaryProvider.MapDictionaryProvider provider = new DictionaryProvider.MapDictionaryProvider();
setVector(dictVector1,
"aa".getBytes(StandardCharsets.UTF_8));
setVector(dictVector2,
"foo".getBytes(StandardCharsets.UTF_8));

provider.put(new Dictionary(dictVector1, new DictionaryEncoding(1L, false, null)));
provider.put(new Dictionary(dictVector2, new DictionaryEncoding(2L, false, null)));

ArrowType int32 = new ArrowType.Int(32, true);
indices.addOrGet("f0",
new FieldType(true, int32, provider.lookup(1L).getEncoding()),
IntVector.class);
indices.addOrGet("f1",
new FieldType(true, int32, provider.lookup(2L).getEncoding()),
IntVector.class);

NullableStructWriter writer = indices.getWriter();
writer.allocate();
writer.start();
writer.integer("f0").writeInt(1);
writer.integer("f1").writeInt(3);
writer.end();
writer.setValueCount(1);

try (final StructVector decode = StructSubfieldEncoder.decode(indices, provider, allocator)) {
fail("There should be an exception when decoding");
} catch (Exception e) {
assertEquals("Provided dictionary does not contain value for index 3", e.getMessage());
}
}
assertEquals("struct decode memory leak", 0, allocator.getAllocatedMemory());
}

private void testDictionary(Dictionary dictionary, ToIntBiFunction<ValueVector, Integer> valGetter) {
try (VarCharVector vector = new VarCharVector("vector", allocator)) {
setVector(vector, "1", "3", "5", "7", "9");
Expand Down