Snappy is a fast data compression and decompression library written in C++ by Google based on ideas from LZ77 and open-sourced in 2011. It does not aim for maximum compression, or compatibility with any other compression library; instead, it aims for very high speeds and reasonable compression. Compression speed is 250 MB/s and decompression speed is 500 MB/s using a single core of a circa 2011 "Westmere" 2.26 GHz Core i7 processor running in 64-bit mode. The compression ratio is 20–100% lower than gzip. Snappy is widely used in Google projects like Bigtable, MapReduce and in compressing data for Google's internal RPC systems. It can be used in open-source projects like MariaDB ColumnStore, Cassandra, Couchbase, Hadoop, LevelDB, MongoDB, RocksDB, Lucene, Spark, and InfluxDB. Decompression is tested to detect any errors in the compressed stream. Snappy does not use inline assembler and is portable.
00 – Literal – uncompressed data; upper 6 bits are used to store length of data. Lengths larger than 60 are stored in a 1-4 byte integer indicated by a 6 bit length of 60 to 63.
01 – Copy with length stored as 3 bits and offset stored as 11 bits; one byte after tag byte is used for part of offset;
10 – Copy with length stored as 6 bits of tag byte and offset stored as two-byte integer after the tag byte;
11 – Copy with length stored as 6 bits of tag byte and offset stored as four-byte little-endian integer after the tag byte;
The copy refers to the dictionary. The offset is the shift from the current position back to the already decompressed stream. The length is the number of bytes to copy from the dictionary. The size of the dictionary was limited by the 1.0 Snappy compressor to 32,768 bytes, and updated to 65,536 in version 1.1.
Example of a compressed stream
The text may be compressed to this, shown as hex data with explanations: 0000000: ca02f042 5769 6b69 7065 6469 6120 6973 ...BWikipedia is The first 2 bytes, ca02 are the length, as a little-endian varint. Thus the most-significant byte is '02'. 0x02ca = 0x014a = 330 bytes. The next two bytes, 0xf042, indicate that a literal of 66+1 bytes follows 0000010: 2061 2066 7265 652c 2077 6562 2d62 6173 a free, web-bas 0000020: 6564 2c20 636f 6c6c 6162 6f72 6174 6976 ed, collaborativ 0000030: 652c 206d 756c 7469 6c69 6e67 7561 6c20 e, multilingual 0000040: 656e 6379 636c 6f09 3ff0 8170 726f 6a65 encyclo.?..proje 0x09 is tag-byte of type 01 with length - 4 = 0102 = 210 and offset = 0x03f = 63 or "pedia "; 0xf081 is a literal with length of 129+1 bytes 0000050: 6374 2e00 0000 0000 0000 0000 0000 0000 ct. In this example, all common substrings with four or more characters were eliminated by the compression process. More common compressors can compress this better. Unlike compression methods such as gzip and bzip2, there is no entropy encoding used to pack alphabet into the bit stream.
