ROLZ explanation?

[Trixter]

Is there a technical explanation of what ROLZ is? I've downloaded QUAD but the source doesn't make much sense to this beginner who only understands LZ77/LZ78/LZSS/LZW (Iila, you need to add some comments into your code once in a while

Does "reduced offset" mean the same thing as Bloom's LZP, where only the length is written as a code and the offset is maintained in a context table?

[encode]

Is there a technical explanation of what ROLZ is? I've downloaded QUAD but the source doesn't make much sense to this beginner who only understands LZ77/LZ78/LZSS/LZW (Iila, you need to add some comments into your code once in a while

Does "reduced offset" mean the same thing as Bloom's LZP, where only the length is written as a code and the offset is maintained in a context table?

Well, the idea is quite simple. Imagine LZP with multiple offsets stored - i.e. instead of just a recent one we keep hundreds or thousands of them. And instead of an actual offset as with LZ77, or instead of nothing as with LZP, we write to the output stream the match index. All in all ROLZ is an LZP with multiple offsets. Hope this helps...

[Bulat Ziganshin]

Is there a technical explanation of what ROLZ is? I've downloaded QUAD but the source doesn't make much sense to this beginner who only understands LZ77/LZ78/LZSS/LZW (Iila, you need to add some comments into your code once in a while

Does "reduced offset" mean the same thing as Bloom's LZP, where only the length is written as a code and the offset is maintained in a context table?

to the best of my knowledge, it's the same as lzrw5, so ?? was described by Ross

and idea is realy simple - for rolz-2, for example, you encode index of match among previous strings with the sme 2-byte prefix. for example:

ab123ab456ab789ab12

the last "12" encoded as -3 offset because you need to copy bytes after the third occurence of the "ab" context

[encode]

Even I can't understand anything from your explanation, Bulat...

[encode]

to the best of my knowledge, it's the same as lzrw5, so ?? was described by Ross

Not really, LZRW5 is about an LZW decoder improvement...
LZRW4 is the correct answer.

[Trixter]

Well, the idea is quite simple. Imagine LZP with multiple offsets stored - i.e. instead of just a recent one we keep hundreds or thousands of them. And instead of an actual offset as with LZ77, or instead of nothing as with LZP, we write to the output stream the match index. All in all ROLZ is an LZP with multiple offsets. Hope this helps...

It does, and it explains why the decompression time isn't as good as LZ77. Thanks for the help.