nbuff 0.1.14

Network buffer

To use this package, run the following command in your project's root directory:

Manual usage
Put the following dependency into your project's dependences section:


Nbuff is nogc and safe buffer. Goals:

  • give user immutable view on bytes
  • avoid data copy
  • avoid GC-allocations

Main consumer of Nbuff can be networking code. Usage scenario: user requested mutable ubyte[] from Nbuff, read data from network into this buffer, then append received data to Nbuff. Then you can find patterns in Nbuff content, apply range algorithms to it, split in on parts, as Nbuff behave as range of bytes - everything without data copy and memory allocations.

All Nbuff content is refcounted, so it will be authomatically freed if not referenced.

You can find detailed tutorial in the docs directory, and some examples in unittest section of buffer.d

Code sample:

    NbuffChunk receiveData(int socket, size_t n) @safe
        // Nbuff.get() - get unique reference for n mutable bytes
        // You can't have multiple references to this buffer (this is unique_ptr)
        auto buffer = Nbuff.get(n);

        // now read data from socket to buffer - unsafe as you pet pointer to buffer 
        auto rc = () @trusted {
            return .recv(socket, cast(void*)&buffer.data[ptr], n, 0);

        if ( rc > 0 )
            // if we received something - convert it to immutable buffer and return
            return NbuffChunk(buffer, rc);
            throw Exception("");
        // if you didn't attached "buffer" to anything - it will be freed on leaving scope
    void readReply(int socket, size_t n)
        Nbuff b;
            NbuffChunk r = receiveData(socket, n);
        catch (Exception e)
            // done

Usage scenario: You are reading bulk newline delimited lines (or any other way structured data) from TCP socket and process it as soon as possible. Every time you received something like:

from network to your socket buffer you can process 'line1' and 'line2' and then you have keep whole buffer (or copy 
and save it incomplete part 'li') just because you have some incomplete data.

This leads to unnecessary allocations and data movement (if you choose to free old buffer and save incomplete part)
or memory wasting (if you choose to preallocate very large buffer and keep processed and incomplete data).

Nbuff solve this problem by using memory pool and smart pointers - it take memory chunks from pool for reading
from network(file, etc...), and authomatically return buffer to pool as soon as you processed all data in it and moved
'processed' pointer forward.
So Nbuff looks like some "window" on the list of buffers, filled with network data, and as soon as buffer moves out of this
window and dereferenced it will be automatically returned to memory pool. Please note - there is no GC allocations, everything
is done using malloc.

  • Igor Khasilev
0.1.14 2021-Jun-17
0.1.13 2021-May-25
0.1.12 2020-Jul-16
0.1.11 2020-Jun-17
0.1.10 2020-Jun-16
Show all 22 versions
Download Stats:
  • 0 downloads today

  • 3 downloads this week

  • 9 downloads this month

  • 67 downloads total

Short URL: