[IOTDB-5443] Implement Chimp encoding in IoTDB

[panagiotisl]

Description

This PR adds the Chimp compression algorithm for double and single precision floating point data.

Algorithm.

Chimp was recently presented in VLDB 2022 (https://www.vldb.org/pvldb/vol15/p3058-liakos.pdf):
Panagiotis Liakos, Katia Papakonstantinopoulou, Yannis Kotidis:
Chimp: Efficient Lossless Floating Point Compression for Time Series Databases. Proc. VLDB Endow. 15(11): 3058-3070 (2022)
The algorithm focuses exclusively on floating point data and takes advantage of more than one earlier values encountered to significantly outperform the state-of-the-art Gorilla algorithm in terms of compression ratio, while preserving its speed.
The implementations provided here focus on the Chimp128 variation for double precision, that uses 128 previous earlier values, and Chimp64 for single precision, that uses 64 earlier values.

Indicative results for many different time series datasets, that highlight the significant benefits expected with the adoption of Chimp in terms of space savings:

Adoption

Chimp is already used in the latest releases of DuckDB (duckdb/duckdb#4878)

Implementation

The algorithm is implemented to reuse code from GorillaEncoderV2.java and GorillaDecoderV2.java, and Long, DoublePrecision, Int and Float versions have been built on top of it. Method organization, design, and naming has been based on the respective Gorilla classes.

Testing

A new class named ChimpDecoderTest executes all tests implemented in the GorillaDecoderV2Test class, ensuring the threshold for code coverage

---

This PR has:

• [X ] been self-reviewed.
  • concurrent read
  • concurrent write
  • concurrent read and write
• added documentation for new or modified features or behaviors.
• added Javadocs for most classes and all non-trivial methods.
• added or updated version, license, or notice information
• added comments explaining the "why" and the intent of the code wherever would not be obvious
  for an unfamiliar reader.
• [X ] added unit tests or modified existing tests to cover new code paths, ensuring the threshold
  for code coverage.
• added integration tests.
• been tested in a test IoTDB cluster.

---

Key changed/added classes (or packages if there are too many classes) in this PR

tsfile/src/main/java/org/apache/iotdb/tsfile/encoding/encoder/IntChimpEncoder.java
tsfile/src/main/java/org/apache/iotdb/tsfile/encoding/encoder/LongChimpEncoder.java
tsfile/src/main/java/org/apache/iotdb/tsfile/encoding/encoder/SinglePrecisionChimpEncoder.java
tsfile/src/main/java/org/apache/iotdb/tsfile/encoding/encoder/DoublePrecisionChimpEncoder.java

tsfile/src/main/java/org/apache/iotdb/tsfile/encoding/encoder/IntChimpDecoder.java
tsfile/src/main/java/org/apache/iotdb/tsfile/encoding/encoder/LongChimpDecoder.java
tsfile/src/main/java/org/apache/iotdb/tsfile/encoding/encoder/SinglePrecisionChimpDecoder.java
tsfile/src/main/java/org/apache/iotdb/tsfile/encoding/encoder/DoublePrecisionChimpDecoder.java

[github-actions]

Hi, this is your first pull request in IoTDB project. Thanks for your contribution! IoTDB will be better because of you.

[panagiotisl]

Some checks are failing but I don't think my PR has anything to do with the failures.

[qiaojialin]

Hi Panagiotis, thanks for contribution! Please update the UserGuide, then users could see this encoding:

docs/UserGuide/Data-Concept/Encoding.md
docs/zh/UserGuide/Data-Concept/Encoding.md

You could just copy the English version into docs/zh/UserGuide/Data-Concept/Encoding.md, then we could translate in the review.

[HTHou]

Hi, there are more code to be update, in order to make Chimp can be used in IoTDB.

server/src/main/java/org/apache/iotdb/db/utils/SchemaUtils.java
tsfile/src/main/java/org/apache/iotdb/tsfile/encoding/encoder/TSEncodingBuilder.java
tsfile/src/main/java/org/apache/iotdb/tsfile/encoding/decoder/Decoder.java
tsfile/src/main/java/org/apache/iotdb/tsfile/file/metadata/enums/TSEncoding.java getTsEncoding method
client-cpp/src/main/Session.h
client-py/iotdb/utils/IoTDBConstants.py

[HTHou]

I did a simple test. After executed the following sql.

The tsfile size with Chimp is 243 bytes, the tsfile size with Gorilla is 248 bytes.

[panagiotisl]

I have used the above example to make tests with 4 different configurations using the first 1,000 values of the basel-wind dataset (see Figure in the PR text). The results are:

Chimp (with snappy): 4695 bytes
Gorilla (with snappy): 8527 bytes
Chimp (uncompressed): 4812 bytes
Gorilla (uncompressed): 8653 bytes

[panagiotisl]

Also some timing comparisons/experiments in milliseconds with three different datasets:

5,000,000 values (Stocks-Germany)

GORILLA Encoding time: 1163
GORILLA Decoding time: 158
CHIMP Encoding time: 553
CHIMP Decoding time: 114

2,905,887 values (city-temperature)

GORILLA Encoding time: 855
GORILLA Decoding time: 97
CHIMP Encoding time: 412
CHIMP Decoding time: 99

8,927 values (SSD-benchmark)

GORILLA Encoding time: 28
GORILLA Decoding time: 7
CHIMP Encoding time: 7
CHIMP Decoding time: 6

These timings refer to encoding the values to a byte array and decoding them using the Java code only. If writing the data to disk is involved the speed up that Chimp can offer will be more evident, as it usually needs to write and read much less bytes to and from the disk.