implement comment foldable whitespace

This commit is contained in:
Quentin 2023-06-12 16:05:06 +02:00
parent 7d3b3ff053
commit 6e76fed684
Signed by: quentin
GPG key ID: E9602264D639FF68
6 changed files with 213 additions and 62 deletions

View file

@ -1,3 +1,7 @@
# imf-codec
**Work in progress, do not use in production**
**Focus: correctness over performance**
**This is currently only a decoder (parser), encoding is not supported.**

View file

@ -1,50 +0,0 @@
use nom::{
IResult,
branch::alt,
bytes::complete::{tag, take_while1},
character::complete::{crlf, space0, space1},
combinator::opt,
sequence::terminated,
};
/// ABNF rfc5234
/// Permissive CRLF
///
/// Theoretically, all lines must end with \r\n
/// but mail servers support malformated emails,
/// for example with only \n eol. It works because
/// \r\n is allowed nowhere else, so we also add this support.
pub fn perm_crlf(input: &str) -> IResult<&str, &str> {
alt((crlf, tag("\r"), tag("\n")))(input)
}
// Note: WSP = SP / HTAB = %x20 / %x09
// nom::*::space0 = *WSP
// nom::*::space1 = 1*WSP
/// Parse a folding white space
///
/// Folding white space are used for long headers splitted on multiple lines
///
/// ```abnf
/// FWS = ([*WSP CRLF] 1*WSP) / obs-FWS
/// obs-FWS = 1*WSP *(CRLF 1*WSP)
/// ```
pub fn fws(input: &str) -> IResult<&str, &str> {
let (input, _) = opt(terminated(space0, perm_crlf))(input)?;
// @FIXME: not implemented obs-FWS
space1(input)
}
/// Sequence of visible chars with the UTF-8 extension
///
/// ```abnf
/// VCHAR = %x21-7E
/// ; visible (printing) characters
/// VCHAR =/ UTF8-non-ascii
/// SEQ = 1*VCHAR
///```
pub fn vchar_seq(input: &str) -> IResult<&str, &str> {
take_while1(|c: char| (c >= '\x21' && c <= '\x7E') || !c.is_ascii())(input)
}

View file

@ -11,18 +11,18 @@ use nom::{
sequence::tuple,
};
use crate::abnf::{fws, vchar_seq, perm_crlf};
use crate::model::{HeaderSection, HeaderDate};
use crate::tokens::{perm_fws, vchar_seq, perm_crlf};
use crate::model::{PermissiveHeaderSection, HeaderDate, MailboxRef};
/// HEADERS
/// Header section
///
/// See: https://www.rfc-editor.org/rfc/rfc5322.html#section-2.2
pub fn header_section(input: &str) -> IResult<&str, HeaderSection> {
pub fn header_section(input: &str) -> IResult<&str, PermissiveHeaderSection> {
let (input, headers) = fold_many0(
header_field,
HeaderSection::default,
PermissiveHeaderSection::default,
|mut section, head| {
match head {
HeaderField::Date(d) => {
@ -114,7 +114,11 @@ fn header_field(input: &str) -> IResult<&str, HeaderField> {
};
(input, HeaderField::Date(date))
},
//"From" => unimplemented!(),
"From" => {
let (input, mbx) = mailbox(input)?;
//many0(
unimplemented!()
},
"Sender" => unimplemented!(),
"Subject" => {
let (input, body) = unstructured(input)?;
@ -136,17 +140,17 @@ fn header_field(input: &str) -> IResult<&str, HeaderField> {
/// unstructured = (*([FWS] VCHAR_SEQ) *WSP) / obs-unstruct
/// ```
fn unstructured(input: &str) -> IResult<&str, String> {
let (input, r) = many0(tuple((opt(fws), vchar_seq)))(input)?;
let (input, r) = many0(tuple((opt(perm_fws), vchar_seq)))(input)?;
let (input, _) = space0(input)?;
// Try to optimize for the most common cases
let body = match r.as_slice() {
[(None, content)] => content.to_string(),
[(Some(ws), content)] => ws.to_string() + content,
[(Some(_), content)] => " ".to_string() + content,
lines => lines.iter().fold(String::with_capacity(255), |acc, item| {
let (may_ws, content) = item;
match may_ws {
Some(ws) => acc + ws + content,
Some(ws) => acc + " " + content,
None => acc + content,
}
}),
@ -155,3 +159,6 @@ fn unstructured(input: &str) -> IResult<&str, String> {
Ok((input, body))
}
fn mailbox(input: &str) -> IResult<&str, MailboxRef> {
unimplemented!();
}

View file

@ -1,3 +1,3 @@
pub mod headers;
pub mod model;
mod abnf;
mod tokens;

View file

@ -9,10 +9,37 @@ pub enum HeaderDate {
None,
}
#[derive(Debug)]
pub struct MailboxRef<'a> {
// The actual "email address" like hello@example.com
pub addrspec: &'a str,
pub name: Option<&'a str>,
}
#[derive(Debug)]
pub struct GroupRef<'a> {
pub name: &'a str,
pub mbx: Vec<MailboxRef<'a>>,
}
#[derive(Debug)]
pub enum AddressRef<'a> {
Single(MailboxRef<'a>),
Many(GroupRef<'a>),
}
/// Permissive Header Section
///
/// This is a structure intended for parsing/decoding,
/// hence it's support cases where the email is considered
/// as invalid according to RFC5322 but for which we can
/// still extract some data.
#[derive(Debug, Default)]
pub struct HeaderSection<'a> {
pub struct PermissiveHeaderSection<'a> {
pub subject: Option<String>,
pub from: Vec<String>,
pub from: Vec<MailboxRef<'a>>,
pub sender: Option<MailboxRef<'a>>,
pub reply_to: Vec<AddressRef<'a>>,
pub date: HeaderDate,
pub optional: HashMap<&'a str, String>,
}
@ -21,7 +48,15 @@ enum InvalidEmailErr {
NoUsableDate,
}
impl<'a> HeaderSection<'a> {
impl<'a> PermissiveHeaderSection<'a> {
/// Check validity of the email
///
/// Especially check that there is no missing fields,
/// or no unique fields declared multiple times.
///
/// See: https://www.rfc-editor.org/rfc/rfc5322#section-3.6
//@FIXME could be changed to a to_StrictHeaderSection call. All fixed errors would be returned in
// a vec of errors.
fn is_valid(&self) -> Result<(), InvalidEmailErr> {
match self.date {
HeaderDate::Parsed(_) => (),

155
src/tokens.rs Normal file
View file

@ -0,0 +1,155 @@
use nom::{
IResult,
branch::alt,
bytes::complete::{tag, take_while1},
character::complete::{crlf, satisfy, space0, space1},
combinator::{recognize, opt},
multi::{many0, many1},
sequence::{preceded, terminated, tuple},
};
/// Lexical tokens
///
/// Approx. maps to section 3.2 of the RFC
/// https://www.rfc-editor.org/rfc/rfc5322#section-3.2
/// Also https://datatracker.ietf.org/doc/html/rfc6532
/// Permissive CRLF
///
/// Theoretically, all lines must end with \r\n
/// but some mail servers like Dovecot support malformated emails,
/// for example with only \n eol. It works because
/// \r or \n is allowed nowhere else, so we also add this support.
pub fn perm_crlf(input: &str) -> IResult<&str, &str> {
alt((crlf, tag("\r"), tag("\n")))(input)
}
// Note: WSP = SP / HTAB = %x20 / %x09
// nom::*::space0 = *WSP
// nom::*::space1 = 1*WSP
/// Quoted pair
///
/// ```abnf
/// quoted-pair = ("\" (VCHAR / WSP)) / obs-qp
/// ```
pub fn quoted_pair(input: &str) -> IResult<&str, char> {
preceded(tag("\\"), satisfy(|c| is_vchar(c) || c == '\t' || c == ' '))(input)
}
/// Permissive foldable white space
///
/// Folding white space are used for long headers splitted on multiple lines.
/// The obsolete syntax allowes multiple lines without content; implemented for compatibility
/// reasons
pub fn perm_fws(input: &str) -> IResult<&str, &str> {
alt((recognize(many1(fold_marker)), space1))(input)
}
fn fold_marker(input: &str) -> IResult<&str, &str> {
let (input, _) = space0(input)?;
let (input, _) = perm_crlf(input)?;
space1(input)
}
/// Folding White Space with Comment
///
/// Note: we drop the comments for now...
///
/// ctext = %d33-39 / ; Printable US-ASCII
/// %d42-91 / ; characters not including
/// %d93-126 / ; "(", ")", or "\"
/// obs-ctext
///
/// ccontent = ctext / quoted-pair / comment
///
/// comment = "(" *([FWS] ccontent) [FWS] ")"
///
/// CFWS = (1*([FWS] comment) [FWS]) / FWS
/// ```
pub fn cfws(input: &str) -> IResult<&str, &str> {
alt((perm_fws, recognize(comments)))(input)
}
pub fn comments(input: &str) -> IResult<&str, ()> {
let (input, _) = many1(tuple((opt(perm_fws), comment)))(input)?;
let (input, _) = opt(perm_fws)(input)?;
Ok((input, ()))
}
pub fn comment(input: &str) -> IResult<&str, ()> {
let (input, _) = tag("(")(input)?;
let (input, _) = many0(tuple((opt(perm_fws), ccontent)))(input)?;
let (input, _) = opt(perm_fws)(input)?;
let (input, _) = tag(")")(input)?;
Ok((input, ()))
}
pub fn ccontent(input: &str) -> IResult<&str, &str> {
alt((recognize(ctext), recognize(quoted_pair), recognize(comment)))(input)
}
pub fn ctext(input: &str) -> IResult<&str, char> {
satisfy(is_ctext)(input)
}
/// Check if it's a comment text character
///
/// ```abnf
/// ctext = %d33-39 / ; Printable US-ASCII
/// %d42-91 / ; characters not including
/// %d93-126 / ; "(", ")", or "\"
/// obs-ctext
///```
pub fn is_ctext(c: char) -> bool {
(c >= '\x21' && c <= '\x27') || (c >= '\x2A' && c <= '\x5B') || (c >= '\x5D' && c <= '\x7E') || !c.is_ascii()
}
/// VCHAR definition
pub fn is_vchar(c: char) -> bool {
(c >= '\x21' && c <= '\x7E') || !c.is_ascii()
}
/// Sequence of visible chars with the UTF-8 extension
///
/// ```abnf
/// VCHAR = %x21-7E
/// ; visible (printing) characters
/// VCHAR =/ UTF8-non-ascii
/// SEQ = 1*VCHAR
///```
pub fn vchar_seq(input: &str) -> IResult<&str, &str> {
take_while1(is_vchar)(input)
}
#[cfg(test)]
mod tests {
use super::*;
use nom;
#[test]
fn test_vchar_seq() {
assert_eq!(vchar_seq("hello world"), Ok((" world", "hello")));
assert_eq!(vchar_seq("hello👋 world"), Ok((" world", "hello👋")));
}
#[test]
fn test_perm_crlf() {
assert_eq!(perm_crlf("\rworld"), Ok(("world", "\r")));
assert_eq!(perm_crlf("\r\nworld"), Ok(("world", "\r\n")));
assert_eq!(perm_crlf("\nworld"), Ok(("world", "\n")));
}
#[test]
fn test_perm_fws() {
assert_eq!(perm_fws("\r\n world"), Ok(("world", "\r\n ")));
assert_eq!(perm_fws(" \r\n \r\n world"), Ok(("world", " \r\n \r\n ")));
assert_eq!(perm_fws(" world"), Ok(("world", " ")));
assert!(perm_fws("\r\nFrom: test").is_err());
}
#[test]
fn test_cfws() {
assert_eq!(cfws("(A nice \\) chap) <pete(his account)@silly.test(his host)>"), Ok(("<pete(his account)@silly.test(his host)>", "(A nice \\) chap) ")));
}
}