Deflate stream question

[Razor12911]

Hi all,


I'm making a deflate or rather a zlib precompressor written in pascal, already managed to make something that can look for all streams in a file, am now dealing with the precompression part, now there is something I want to know before precompressing a certain stream which is the sliding window / window size that was used either from deflate headers or just by decompressing that stream, is there a way to know that? I want to sort out the streams with 16K and 32K window size from the other ones with different sliding window then fed them to different functions that can handle each.

Thanks in advance.

[Shelwien]

For zlib there is a way - http://stackoverflow.com/a/9050274/395609
Otherwise you just have to presume that the window is always 32k.

[schnaader]

As Shelwien said, the zlib header is the only way to detect it for sure. Anyway, one further comment to this: In other cases where you don't know window size, presume 32K like Shelwien said, but have a look at the biggest distance value used - this determines which window size settings can be used. E.g. if the biggest distance value is 5K, recompression will give the same results for 8K, 16K and 32K window size (except zlib header), but not for values <= 4K.

[Razor12911]

For zlib there is a way - http://stackoverflow.com/a/9050274/395609
Otherwise you just have to presume that the window is always 32k.

Thank you for response however, I already know this much, the program I'm trying to create detects deflate streams instead of zlib streams which have the headers from the link you provided.

As Shelwien said, the zlib header is the only way to detect it for sure. Anyway, one further comment to this: In other cases where you don't know window size, presume 32K like Shelwien said, but have a look at the biggest distance value used - this determines which window size settings can be used. E.g. if the biggest distance value is 5K, recompression will give the same results for 8K, 16K and 32K window size (except zlib header), but not for values <= 4K.

Anyways I don't know if you guys understand pascal but I found this piece of code so what you're telling me is that even if I pursue the idea that was used here to calculate window size or rather the distance, which is this part in the attachment I uploaded:
Distance := dfc_DistanceBase[Symbol] + ((BitBuffer shr SymbolCodeLen) and AbExtractMask[ExtraBitCount]);


where dfc_DistanceBase is


dfc_DistanceBase : array [0..31] of word =
(1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, 257,
385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145, 8193, 12289,
16385, 24577, 32769, 49153);

[Shelwien]

As schnaader said, you can actually decode first block of deflate stream (and you kinda have to do that for detection)
and use maximum distance from matches there to determine the window size for the block... and the min window size for the stream.
But overall there's no way to detect the window size in advance. In fact, it may be not a good idea to do that at all, as there're streams
made by concatenation of separately compressed blocks (eg. .docx files are like that), in which case the window size may be different
for parts of deflate stream.

[Razor12911]

Ok thanks, good to know.