C#•3d ago

Streaming zip extract

Is it possible to do a streaming zip extract to a folder (from a network stream)?
The ZipArchive constructor seems to copy the entire stream into memory if seeking isn't supported.
Perhaps there's a third-party library?

cap5lut•12/22/25, 12:29 PM

iirc the way the ZIP format is designed, it simply requires seeking, the only way around having the entire content in memory is to write it to a temporary file first.
but just like with the the memory stream, u will have to consume everything first, but at least its not all in memory at once

RuttieOP•12/22/25, 12:31 PM

I checked the source, and even with Read, if the stream is not seekable it simply copies to a memory stream

RuttieOP•12/22/25, 12:31 PM

see https://github.com/dotnet/runtime/blob/98237c4b423fad026f38c2530bac9954b9a9dcfc/src/libraries/System.IO.Compression/src/System/IO/Compression/ZipArchive.cs#L132

Ccap5lut iirc the way the ZIP format is designed, it simply requires seeking, the only wa...

RuttieOP•12/22/25, 12:34 PM

That's unfortunate
I guess I have no choice but to just read the entire thing to memory then

RuttieOP•12/22/25, 12:38 PM

I didn't look into it as much as ZipArchive, but from what I saw it just throws if the stream is not seekable

cap5lut•12/22/25, 12:39 PM

the main problem is that u basically have 3 "main sections" of data in a zip file:
entry headers -> compressed data -> "directory data table"
the headers might or might not contain the compressed data size, if they do u can probably stream + extract, but if not u have to get to the last section to get the actual offset and lengths

Original message was deleted

cap5lut•12/22/25, 12:48 PM

was digging through the code, SharpZipLib simply throws if the stream is not seekable

Original message was deleted

cap5lut•12/22/25, 2:59 PM

tbh, thats actually a problem ontop of the current problem, so just waiting for that wont fix it all

Original message was deleted

cap5lut•12/22/25, 3:33 PM

restarting a network stream is sorta impossible ;p
so yeah caching the stream somehow (either via a memory or file stream) would be required, but those will require ya to copy over the whole thing first

cap5lut•12/22/25, 3:36 PM

i guess u could write a wrapper around a network stream and some "buffer stream" if u know the total size length beforehand and block on

Seek

Seek

s to unavailable data,
but that will be in the end something extremely horrible as well

Ccap5lut restarting a network stream is sorta impossible ;p so yeah caching the stream so...

Cattywampus•12/22/25, 3:41 PM

yeah just load it to a memory stream, so no need for temp files

Cattywampus•12/22/25, 3:44 PM

its a bad idea if the zip size is massive kekw

Cattywampus•12/22/25, 3:51 PM

OOM easy peasy blobthumbsup

Cattywampus•12/22/25, 3:52 PM

thinking of chunking then again I dont know how nor do I've experience on how network packets work

CCattywampus thinking of chunking then again I dont know how nor do I've experience on how ne...

cap5lut•12/22/25, 3:54 PM

thats the main problem why non-seekable streams wont (always) work, u cant just forward read in all cases

Ccap5lut the main problem is that u basically have 3 "main sections" of data in a zip fil...

cap5lut•12/22/25, 3:55 PM

Cattywampus•12/22/25, 3:55 PM

cap5lut•12/22/25, 3:57 PM

and for first throwing it into a

FileStream

FileStream

MemoryStream

MemoryStream

CopyToAsync(targetStream)

CopyToAsync(targetStream)

is enough

cap5lut•12/22/25, 3:59 PM

tho if its a big file and keeping it in memory is okay, i probably wouldnt use a

MemoryStream

MemoryStream

. big continuous memory is bad, so writing something similar to

MemoryStream

MemoryStream

that uses multiple arrays will probably be better

cap5lut•12/22/25, 4:00 PM

(especially with an array pool if that happens a lot)

cap5lut•12/22/25, 4:04 PM

so @Ruttie if u want us to be of more help, i think u have to give us more details.
like if u know the average/max/min zip file sizes, how often this operation occurs, if u are in control of the server side and can change the archive type, etc.

RuttieOP•12/22/25, 4:06 PM

I am not in control of the server side
I'm currently planning to set an upper limit for in-memory reads, and am planning to use a temp file if the zip is above that limit

RuttieOP•12/22/25, 4:07 PM

the operation doesn't happen often luckily, so that should be fine

cap5lut•12/22/25, 4:10 PM

do u know the size of the zip beforehand?

RuttieOP•12/22/25, 11:22 PM

I mean, from the http request yeah, but the actual size can change wildly

RuttieOP•12/22/25, 11:22 PM

though it'll probably stick between 15kb and ~500mb (though larger is still possible)

Streaming zip extract

Similar Threads

Similar Threads

Similar Threads