[CASSANDRA-13676] Some serializers depend on Stream-specific methods

• Type: Bug
• Status: Open
• Priority: Normal
• Resolution: Unresolved
• Component/s: Legacy/Core
• Severity: Normal
• Assignee: Unassigned
• Reporter: Hao Zhong
• Created: 06/Jul/17 06:16
• Updated: 16/Apr/19 09:30

Description

When fixing CASSANDRA-2382, Jonathan Ellis complained that some serializers did (do?) depend on Stream-specific methods. The buggy code is as follow:

public static class EstimatedHistogramSerializer implements ICompactSerializer<EstimatedHistogram>
    {
        public void serialize(EstimatedHistogram eh, DataOutputStream dos) throws IOException
        {
            long[] offsets = eh.getBucketOffsets();
            long[] buckets = eh.getBuckets(false);
            dos.writeInt(buckets.length);
            for (int i = 0; i < buckets.length; i++)
            {
                dos.writeLong(offsets[i == 0 ? 0 : i - 1]);
                dos.writeLong(buckets[i]);
            }
        }

        public EstimatedHistogram deserialize(DataInputStream dis) throws IOException
        {
            int size = dis.readInt();
            long[] offsets = new long[size - 1];
            long[] buckets = new long[size];

            for (int i = 0; i < size; i++) {
                offsets[i == 0 ? 0 : i - 1] = dis.readLong();
                buckets[i] = dis.readLong();
            }
            return new EstimatedHistogram(offsets, buckets);
        }
    }

The fixed code is:

public static class EstimatedHistogramSerializer implements ICompactSerializer2<EstimatedHistogram>
    {
        public void serialize(EstimatedHistogram eh, DataOutput dos) throws IOException
        {
            long[] offsets = eh.getBucketOffsets();
            long[] buckets = eh.getBuckets(false);
            dos.writeInt(buckets.length);
            for (int i = 0; i < buckets.length; i++)
            {
                dos.writeLong(offsets[i == 0 ? 0 : i - 1]);
                dos.writeLong(buckets[i]);
            }
        }

        public EstimatedHistogram deserialize(DataInput dis) throws IOException
        {
            int size = dis.readInt();
            long[] offsets = new long[size - 1];
            long[] buckets = new long[size];

            for (int i = 0; i < size; i++) {
                offsets[i == 0 ? 0 : i - 1] = dis.readLong();
                buckets[i] = dis.readLong();
            }
            return new EstimatedHistogram(offsets, buckets);
        }
    }

I notice that some serializers still depend on Stream-specific methods. For example, the IndexSummary_deserialize method has the following code:

 public IndexSummary deserialize(DataInputStream in, IPartitioner partitioner, int expectedMinIndexInterval, int maxIndexInterval) throws IOException
        {
            int minIndexInterval = in.readInt();
            if (minIndexInterval != expectedMinIndexInterval)
            {
                throw new IOException(String.format("Cannot read index summary because min_index_interval changed from %d to %d.",
                                                    minIndexInterval, expectedMinIndexInterval));
            }

            int offsetCount = in.readInt();
            long offheapSize = in.readLong();
            int samplingLevel = in.readInt();
            int fullSamplingSummarySize = in.readInt();

            int effectiveIndexInterval = (int) Math.ceil((BASE_SAMPLING_LEVEL / (double) samplingLevel) * minIndexInterval);
            if (effectiveIndexInterval > maxIndexInterval)
            {
                throw new IOException(String.format("Rebuilding index summary because the effective index interval (%d) is higher than" +
                                                    " the current max index interval (%d)", effectiveIndexInterval, maxIndexInterval));
            }

            Memory offsets = Memory.allocate(offsetCount * 4);
            Memory entries = Memory.allocate(offheapSize - offsets.size());
            try
            {
                FBUtilities.copy(in, new MemoryOutputStream(offsets), offsets.size());
                FBUtilities.copy(in, new MemoryOutputStream(entries), entries.size());
            }
            catch (IOException ioe)
            {
                offsets.free();
                entries.free();
                throw ioe;
            }
            // our on-disk representation treats the offsets and the summary data as one contiguous structure,
            // in which the offsets are based from the start of the structure. i.e., if the offsets occupy
            // X bytes, the value of the first offset will be X. In memory we split the two regions up, so that
            // the summary values are indexed from zero, so we apply a correction to the offsets when de/serializing.
            // In this case subtracting X from each of the offsets.
            for (int i = 0 ; i < offsets.size() ; i += 4)
                offsets.setInt(i, (int) (offsets.getInt(i) - offsets.size()));
            return new IndexSummary(partitioner, offsets, offsetCount, entries, entries.size(), fullSamplingSummarySize, minIndexInterval, samplingLevel);
        }

Is it worthy replacing the Stream-specific inputs as well?