cajeta.io.net.TcpStream — a connected TCP stream: the blocking client/peer socket, obtained from connect (client side) or TcpListener.accept (server side). The async forms (connectAsync, readAsync, writeAsync, writeAllAsync, readWithin) are fiber-parking readiness loops — they never block the carrier thread. The static WOULD_BLOCK sentinel (-1) is returned by the non-blocking I/O forms when an op could not complete immediately. Socket options (TCP_NODELAY, SO_KEEPALIVE, buffer sizes, SO_LINGER, TTL) are exposed as set/get pairs, and the destructor closes the socket if close() wasn’t called, so a dropped TcpStream never leaks a descriptor.

SocketAddress addr #= SocketAddress.parse("127.0.0.1:7000");
TcpStream s #= TcpStream.connect(#addr);
int8[] buf = heap int8[4];
s.write(buf, (int64) 0, (int64) 4);
int64 got = s.read(buf, (int64) 0, (int64) 4);
s.close();

Methods

Signature
`static #TcpStream connect(#SocketAddress addr)` ⚑	Open a blocking TCP connection to `addr` and return the connected stream
`static #TcpStream connectAsync(#SocketAddress addr)` ⚑	Connect without blocking the carrier (the fiber parks until the connect completes)
`int64 read(int8[] dst, int64 offset, int64 length)`	Read up to `length` bytes into `dst[offset..offset+length)`; returns the count, `0` at EOF
`int64 write(int8[] data, int64 offset, int64 length)`	Write up to `length` bytes from `data[offset..)`; returns the count written
`void writeAll(int8[] data, int64 offset, int64 length)`	Write all of `data[offset..offset+length)`, looping past short writes
`int64 readAsync(int8[] buf, int64 offset, int64 length)`	Async read — parks the fiber on readability instead of blocking
`int64 writeAsync(int8[] data, int64 offset, int64 length)`	Async write — parks the fiber on writability
`void writeAllAsync(int8[] data, int64 offset, int64 length)`	Async write-all: every byte lands, fiber-parking between short writes
`int64 readWithin(int8[] buf, int64 offset, int64 length, int32 timeoutMs)`	Deadline-bounded async read: like `readAsync` but bounds the total wait by `timeoutMs`
`void setNoDelay(boolean on)` / `boolean getNoDelay()`	`TCP_NODELAY` — disable Nagle’s algorithm so small writes send immediately
`void setKeepAlive(boolean on)` / `boolean getKeepAlive()`	`SO_KEEPALIVE`
`void setRecvBufferSize(int32 bytes)` / `int32 getRecvBufferSize()`	Kernel receive buffer size (`SO_RCVBUF`)
`void setSendBufferSize(int32 bytes)` / `int32 getSendBufferSize()`	Kernel send buffer size (`SO_SNDBUF`)
`void setLinger(boolean on, int32 seconds)` / `boolean getLinger()`	`SO_LINGER`
`void setTtl(int32 ttl)` / `int32 getTtl()`	Outbound unicast TTL / hop limit (`0..255`)
`void shutdown(int32 how)`	Shut down part of the full-duplex connection without closing the descriptor
`void close()`	Close the underlying socket; idempotent
`~TcpStream()`	Destructor: closes the socket if `close` wasn’t called

⚑ = @EntryPoint

TcpStream

Methods

See also