Download InputStream

[llyppi]

project.downloadObject
InputStream close after 64mg writed
Works if use ByteStreams.copy(download, out)
But that away its needed write byte[] in server side,after create InputStream to download client side
It would be better to send Inputstream directly to the client side response to download object

[mniewrzal]

I'm not sure I understand main problem. Could you give me a code sample to visualize it?

[llyppi]

code.zip

[Erikvv]

Thanks for your report!

InputStream close after 64mg writed

At 64 MiB int ObjectInputStream.read(byte[]) will return 0 because there are no more bytes in the segment. You need to call read(byte[]) again to start reading the next segment.

This looks like a bug, it doesn't satisfy the specification:

https://docs.oracle.com/en/java/javase/17/docs/api/java.base/java/io/InputStream.html#read(byte%5B%5D,int,int):
(...)
at least one byte is read and stored into b

For sure we also have a bug in int ObjectInputStream.read() which throws IOException("invalid state") at 64 MiB.

[Erikvv]

Works if use ByteStreams.copy(download, out)
But that away its needed write byte[] in server side,after create InputStream to download client side
It would be better to send Inputstream directly to the client side response to download object

It sounds like you're asking for an API change because you think it would be faster. Is ObjectInputStream.transferTo(OutputStream) maybe suitable? I don't think you'll be able to measure a difference though.

[llyppi]

FileInputStream works good but its need exist file in server side to download
Better case is on the moment open
ObjectInputStream download = project.downloadObject(bucket, new File(file).getName());
use that InputStream download and redirect to response client side
No needed write file in server and after create FileInputStream

[Erikvv]

I'm sorry. I don't understand your use case and proposal.

[llyppi]

I have a panel web to explorer my buckets
When client side request download, first the object will be write o server in file,after FileInputstream return response to client request
But that away all object will stored in server, not good
Using ByteOutputstream the memory server its out of memory

[Erikvv]

I have a panel web to explorer my buckets
When client side request download, first the object will be write o server in file,after FileInputstream return response to client request
But that away all object will stored in server, not good
Using ByteOutputstream the memory server its out of memory

"Doctor it hurts when I move like this"
-> "Then don't move like that"

It looks you can use jakarta.servlet.http.HttpServletResponse.getOutputStream() to immediately write from the HTTP server to the client, without storing anything on the server.

Your HTTP server framework may even accept an InputStream directly like io.javalin.http.Context.result(InputStream resultStream)

(FWIW, I'm not a Java guru)

[Erikvv]

With regards to the 0 bytes at 64 MiB, created storj/uplink#99

[llyppi]

i try owerite close method