The new-tokio-smtp crate provides an extendible SMTP (Simple Mail Transfer Protocol) implementation using tokio.
This crate provides only SMTP functionality, this means it does neither provides functionality for creating mails, nor for e.g. retrying sending a mail if the receiver was temporary not available.
While it only provides SMTP functionality it is written in a way to make it easy to integrate with higher level libraries. The interoperability is provided through two mechanisms:
-
SMTP commands are defined in a way which allow library user to define there own commands, all commands provided by this library could theoretically have been implemented in an external library, this includes some of the more special commands like
STARTTLS
,EHLO
andDATA
. Moreover aConnection
can be converted into aIo
instance which provides a number of useful functionalities for easily implementing new commands, e.g.Io.parse_response
. -
syntactic construct's like e.g.
Domain
orClientId
can be parsed but also have "unchecked" constructors, this allows libraries which have there own validation to skip redundant validations, e.g. if a mail library might provide aMailbox
type of mail addresses and names, which is guaranteed to be syntactically correct if can implement a simpleFrom
/Into
impl to cheaply convert it to anForward-Path
. (Alternative they also could implement their ownMail
cmd if this has any benefit for them) -
provided commands (and syntax constructs) are written in a robust way, allowing for example extensions like
SMTPUTF8
to be implemented on it. The only drawback of this is that it trusts that parts created by more higher level libraries are valid, e.g. it won't validate that the mail given to it is actually 7bit ascii or that it does not contain "orphan"'\n'
(or'\r'
) chars. But this is fine as this library is for using smtp to send mails, but not for creating such mails. (Note that while it is trusting it does validate if a command can be used through checking the result from the lastEHLO
command, i.e. it wont allow you to send aSTARTTLS
command on a mail server not supporting it) -
handling logic errors (i.e. server responded with code 550) separately from more fatal errors like e.g. a broken pipe
extern crate futures;
extern crate tokio;
extern crate new_tokio_smtp;
#[macro_use]
extern crate vec1;
extern crate rpassword;
use std::io::{stdin, stdout, Write};
use futures::stream::Stream;
use futures::future::lazy;
use new_tokio_smtp::error::GeneralError;
use new_tokio_smtp::{command, Connection, ConnectionConfig, Domain};
use new_tokio_smtp::send_mail::{
Mail, EncodingRequirement,
MailAddress, MailEnvelop,
};
struct Request {
config: ConnectionConfig<command::auth::Plain>,
mails: Vec<MailEnvelop>
}
fn main() {
let Request { config, mails } = read_request();
// We only have iter map overhead because we
// don't have a failable mail encoding step, which normally is required.
let mails = mails.into_iter().map(|m| -> Result<_, GeneralError> { Ok(m) });
println!("[now starting tokio]");
tokio::run(lazy(move || {
println!("[start connect_send_quit]");
Connection::connect_send_quit(config, mails)
//Stream::for_each is design wise broken in futures v0.1
.then(|result| Ok(result))
.for_each(|result| {
if let Err(err) = result {
println!("[sending mail failed]: {}", err);
} else {
println!("[successfully send mail]")
}
Ok(())
})
}))
}
fn read_request() -> Request {
println!("preparing to send mail with ethereal.email");
let sender = read_email();
let passwd = read_password();
// The `from_unchecked` will turn into a `.parse()` in the future.
let config = ConnectionConfig
::builder(Domain::from_unchecked("smtp.ethereal.email"))
.expect("resolving domain failed")
.auth(command::auth::Plain::from_username(sender.clone(), passwd)
.expect("username/password can not contain \\0 bytes"))
.build();
// the from_unchecked normally can be used if we know the address is valid
// a mail address parser will be added at some point in the future
let send_to = MailAddress::from_unchecked("[email protected]");
// using string fmt to crate mails IS A
// REALLY BAD IDEA there are a ton of ways
// this can go wrong, so don't do this in
// practice, use some library to crate mails
let raw_mail = format!(concat!(
"Date: Thu, 14 Jun 2018 11:22:18 +0000\r\n",
"From: You <{}>\r\n",
//ethereal doesn't delivers any mail so it's fine
"To: Invalid <{}>\r\n",
"Subject: I am spam?\r\n",
"\r\n",
"...\r\n"
), sender.as_str(), send_to.as_str());
// this normally adapts to a higher level abstraction
// of mail then this crate provides
let mail_data = Mail::new(EncodingRequirement::None, raw_mail.to_owned());
let mail = MailEnvelop::new(sender, vec1![ send_to ], mail_data);
Request {
config,
mails: vec![ mail ]
}
}
fn read_email() -> MailAddress {
let stdout = stdout();
let mut handle = stdout.lock();
write!(handle, "enter ethereal.email mail address\n[Note mail is not validated in this example]: ")
.unwrap();
handle.flush().unwrap();
let mut line = String::new();
stdin().read_line(&mut line).unwrap();
MailAddress::from_unchecked(line.trim())
}
fn read_password() -> String {
rpassword::prompt_password_stdout("password: ").unwrap()
}
The concept of behind the library is explained in the notes/concept.md file.
The library provides a number of usability helpers:
-
chain::chain
provides a easy way to chain a number of SMTP commands, sending each command when the previous command in the chain did not fail in any way. -
mock_support
feature: Extends the Socket abstraction to not only abstract over the socket being either aTcpStream
or aTlsStream
but also adds another variant which is a boxedMockStream
, making the smtp libraries, but also libraries build on top of it more testable. -
mock::MockStream
(use the featuresmock-impl
) A simple implementation for aMockStream
which allows you to test which data was send to it and mock responses for it. (Through it's currently limited to a fixed predefined conversation, if more is needed a customMockStream
impl. has to be used) -
future_ext::ResultWithContextExt
: Provides actx_and_then
andctx_or_else
methods making it easier to handle results resolving as Item to an tuple of a context (here the connection) and aResult
belonging to an different abstraction level than the futuresError
(here a possibleCommandError
while the futureError
is an connection error like e.g. a broken pipe)
Like mentioned before this library has some limitations as it's meant to only do SMTP and nothing more. Through there are some other limitations, which will be likely to be fixed in future versions:
-
no mail address parser for
send_mail::MailAddress
and neither a parser forForwardPath
/ReversePath
(they can be constructed usingfrom_unchecked
). This will be fixed when I find a library "just" doing mail addresses and doing it right. -
no "build-in" support for extended status codes, this is mainly the way because I hadn't had time for this, changing this in a nice build-in way might require some API changes wrt. to the
Response
type and it should be done beforev1.0
-
The number of provided commands is currently limited to a small but useful subset, commands which would be nice to provide include
BDAT
and more variations ofAUTH
(currently provided arePLAIN
and simpleLOGIN
which is enough for most cases but supporting e.g.OAuth2
would be good) -
no support for
PIPELINING
, while most extensions can be implemented using custom commands, this is not true for pipelining. While there exists a concept how pipelining can be implemented without to much API brakeage this is for now not planed due to time limitations. -
no stable version (
v1.0
) for now, astokio
is not stable yet. When tokio becomes stable a stable version should be released, through another one might have to be released at some point ifPIPELINING
is implemented later one (through in the current concept for implementing it there are little braking changes, except for implementors of custom commands)
Documentation can be viewed on docs.rs.
Licensed under either of
- Apache License, Version 2.0, (LICENSE-APACHE or https://www.apache.org/licenses/LICENSE-2.0)
- MIT license (LICENSE-MIT or https://opensource.org/licenses/MIT)
at your option.
Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.
-
v0.4
:- renamed
from_str_unchecked
tofrom_unchecked
Cmd.exec
accepts nowIo
instead ofConnection
- replace
CmdFuture
withExecFuture
Connection.send_simple_cmd
is nowIo.exec_simple_cmd
- replace
- renamed
-
v0.5
:- improved
ClientId
- renamed
ClientIdentity
toClientId
- added
hostname()
constructor
- renamed
- added builder for
ConnectionConfig
- removed old
with_
constructors
- removed old
- placed all
Auth*
commands into aauth
module (e.g.AuthPlain
=>auth::Plain
) - changed feature naming schema
- improved
-
v0.6
- added connection builder for local non secure connections
- addec constructors for builders on types they build
- renamed
auth::plain::NullCodePoint
toauth::plain::NullCodePointError
-
v0.7
send_all_mails
andconnect_send_quit
now accept aIntoIterable
instead of stream- you need to have all values already ready when sending so
Stream
didn't fit well - it also means you can now pass in a
Vec
or astd::iter:once
- you need to have all values already ready when sending so
GeneralError
as no longer thePreviousRequestKilledConnection
error variant instead astd::io::Error::new(std::io::ErrorKind::NoConnection, "...")
is returned which makes it easier to adapt to by libraries using it and fit the semantics as good as previous solutionsend_all_mails
andconnect_send_quit
now return a stream instead of a future resolving to a stream.
-
v0.7.1
- updated dependencies, makes sure tokio doesn't produces a deprecation warning as a import moved to a different place in tokio
-
v0.8.0
- update
tokio-tls
/native-tls
to v0.2.x - renamed method creating a builder from
build*
tobuilder*
- update