Deflate (String function)
Compress a longstring with deflate (String class)
[Introduced in Sirius Mods 7.4]
This function compresses a longstring using the "deflate" algorithm. The deflate algorithm is described completely in RFC 1951. It is very effective with HTML and XML data.
Syntax
%outString = string:Deflate[( [FixedCode= boolean], [LazyMatch= boolean], - [MaxChain= number])]
Syntax terms
| %outString | The resulting compressed string. |
|---|---|
| string | The string to be compressed. |
| FixedCode | FixedCode is an optional, name required, parameter that is a Boolean value that specifies whether the compression uses fixed codes or is dynamic, based on the contents of the input string.
The default value for this argument is False (use dynamic compression). |
| LazyMatch | LazyMatch is an optional, name required, parameter that is a Boolean value that specifies whether to use "lazy match" compression, as specified in RFC 1951.
The default value for this argument is False (do not use "lazy match" compression). Note: This parameter has no effect when hardware compression is used (see Hardware vs. software compression). |
| MaxChain | MaxChain is an optional, name required, parameter that is a numeric value that specifies the maximum hash chain length, as explained in RFC 1951.
The default value for this argument is 0. If specified, it must be between 0 and 99, inclusive. Note: This parameter has no effect when hardware compression is used (see Hardware vs. software compression). |
Usage notes
- The NCMPBUF parameter must be set to a non-0 value during Model 204 initialization to allow use of the Deflate function; otherwise, invoking Deflate causes request cancellation.
- As with any compression scheme, it is possible that a particular string will become longer after compression. This would happen, for example, if a deflated string were passed to Deflate.
- Short strings (less than 128 bytes) typically compress better with
fixedCode=true.
Hardware vs. software compression
[Introduced in Model 204 version 8.0]
On IBM z15 and above processors, Deflate automatically uses the DFLTCC (Deflate Conversion Call) hardware instruction to perform compression. On older processors without DFLTCC support, the existing software implementation is used. The choice is made automatically and is transparent to the application.
There are some behavioral differences between hardware and software compression that should be noted:
- The LazyMatch and MaxChain parameters are ignored when hardware compression is used. The DFLTCC hardware uses its own internal compression strategy for match searching, and these parameters have no effect. The FixedCode parameter is honored: when
fixedCode=true, the hardware uses fixed Huffman codes; whenfixedCode=false(the default), the hardware generates a Dynamic Huffman Table (DHT) from the input data for each deflate block. - Hardware compression is non-deterministic: compressing the same input data multiple times may produce different compressed output, even with identical parameters. This is because the DFLTCC instruction uses an implementation-dependent hash function for duplicate string matching, as documented in the z/Architecture Principles of Operation. The compressed output is always valid RFC 1951 data and can be decompressed by Inflate regardless, but applications should not depend on a specific compressed length or byte-for-byte reproducibility of compressed data.
- Hardware compression generally provides significantly better throughput than software compression, particularly for larger input data.
These differences also apply to the Gzip and Zip methods, which use Deflate compression internally.
Examples
In the following example, %out is set to the compressed version of the given string:
%out = 'How much wood could a woodchuck chuck':deflate(fixedCode=true)
See also
- Inflate is used to decompress the string to its original value.
- Other related methods: