Data Compression Book with LZMA description [!]

[encode]

Check out David Solomon's Data Compression The Complete Reference - Fourth Edition book which contains the LZMA description!

datacompression.rar (~7 MB)

Note: For evaluation purposes only. If you like this book - purchase it.

[Bulat Ziganshin]

Ilya, can you please stop advertizing your program in such way?

[encode]

Why not?

EDIT:
It reminds me that Malcolm Taylor sends all files in RK format. OK, actually I'll repack the book! Anyway, I'm using PIM archiver myself!

[Bulat Ziganshin]

big thanks for book!

[LovePimple]

Thanks Ilia!

[IsName]

Ilya, can you please stop advertizing your program in such way?

Bulat, dont you use such method?

[Bulat Ziganshin]

afair, the only case when i've published file compressed by my programs is *update* to freearc which was pretty useless without freearc already installed

Ilya, the book says almost nothing about rar&lzma. it only contains desciption of binary tree - hope that this will help us to understand it

[toffer]

I've bought the 3rd edition, which is a very easy to understand reference. I already knew about the 7zip describtion in the 4th edition. But the information is poor, these string search mechanisms aren't anything new. It doesn't contain information about lzma's "offset reduction by markov chains", or am i wrong here? Only the "last few references as special indices"-trick. I've only quickly overflown the chapter about lzma.

[encode]

It doesnt contain information about lzmas "offset reduction by markov chains", or am i wrong here?

Yep, youre wrong. Nothing special with LZMAs offset encoding. LZMA can be regarded as a extended LZX - LZARI with Optimal Parsing and some tricks. For example LZMA can encode literals with matched byte - we use the byte at the rep0 (recent offset) as a context for literal coding, if we got wrong bit (unmatched) we stop coding bits with such context, and decode/encode literal as usual.

Ilya, the book says almost nothing about rar&lzma.

Better than nothing!

[Bulat Ziganshin]

author is pretty ignorant in lz77 area (compared to me ). he just included markeing info from programs. there is no any info on rar/lzma except that lzma uses rep codes (invented by Roshal) and half-described bin. trees (rebalancing algorithm isnt described). lzma!=lzari, otherwise tornado/lzpm will have the same compression ratio as lzma on binary data. he erroneously wrote that lzma uses lzrw4-like hash while actually its exactly the same as in deflate/rar

Better than nothing!

nothing new for me. i know much more and you too

[encode]

lzma!=lzari, otherwise tornado/lzpm will have the same compression ratio as lzma on binary data.

Test the LZMA at Fast modes.

The LZMAs parsing makes the difference. LZPM is just an experimental ROLZ implementation. I position BALZ as the extended Deflate (at least with Lazy Matching it is, I hope soon youll see BALZ with two modes).


Dont forget that LZPM/Tornado/BALZ are too young compared to the LZMA!

[Bulat Ziganshin]

Test the LZMA at Fast modes.

ive made a lot of tests. unfortunately for me, lzma still outperform tornado at binary files. and you know a lot of ideas which helps lzma to do it