documented parsererror

src/parser.rs

@@ -10,17 +10,18 @@
 //! [`GitConfig`]: crate::config::GitConfig
 
 use bstr::{BStr, ByteSlice};
+use nom::branch::alt;
 use nom::bytes::complete::{escaped, tag, take_till, take_while};
 use nom::character::complete::{char, none_of, one_of};
 use nom::character::{is_newline, is_space};
 use nom::combinator::{map, opt};
 use nom::error::{Error as NomError, ErrorKind};
+use nom::multi::{many0, many1};
 use nom::sequence::delimited;
 use nom::IResult;
-use nom::{branch::alt, multi::many0};
-use std::borrow::Cow;
 use std::fmt::Display;
 use std::iter::FusedIterator;
+use std::{borrow::Cow, error::Error};
 
 /// Syntactic events that occurs in the config. Despite all these variants
 /// holding a [`Cow`] instead over a simple reference, the parser will only emit
@@ -158,20 +159,77 @@ impl Display for ParsedComment<'_> {
     }
 }
 
-/// The various parsing failure reasons.
+/// A parser error reports the one-indexed line number where the parsing error
+/// occurred, as well as the last parser node and the remaining data to be
+/// parsed.
 #[derive(PartialEq, Debug)]
-pub enum ParserError<'a> {
-    /// A parsing error occurred.
-    InvalidInput(nom::Err<NomError<&'a [u8]>>),
-    /// The config was successfully parsed, but we had extraneous data after the
-    /// config file.
-    ConfigHasExtraData(&'a BStr),
+pub struct ParserError<'a> {
+    line_number: usize,
+    last_attempted_parser: ParserNode,
+    parsed_until: &'a [u8],
 }
 
-#[doc(hidden)]
-impl<'a> From<nom::Err<NomError<&'a [u8]>>> for ParserError<'a> {
-    fn from(e: nom::Err<NomError<&'a [u8]>>) -> Self {
-        Self::InvalidInput(e)
+impl ParserError<'_> {
+    /// The one-indexed line number where the error occurred. This is determined
+    /// by the number of newlines that were successfully parsed.
+    pub fn line_number(&self) -> usize {
+        self.line_number + 1
+    }
+
+    /// The remaining data that was left unparsed.
+    pub fn remaining_data(&self) -> &[u8] {
+        self.parsed_until
+    }
+}
+
+impl Display for ParserError<'_> {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        let data_size = self.parsed_until.len();
+        let data = std::str::from_utf8(self.parsed_until);
+        write!(
+            f,
+            "Got an unexpected token on line {} while trying to parse a {}: ",
+            self.line_number + 1,
+            self.last_attempted_parser,
+        )?;
+
+        match (data, data_size) {
+            (Ok(data), _) if data_size > 10 => write!(
+                f,
+                "'{}' ... ({} characters omitted)",
+                &data[..10],
+                data_size - 10
+            ),
+            (Ok(data), _) => write!(f, "'{}'", data),
+            (Err(_), _) if data_size > 10 => write!(
+                f,
+                "'{:02x?}' ... ({} characters omitted)",
+                &self.parsed_until[..10],
+                data_size - 10
+            ),
+            (Err(_), _) => write!(f, "'{:02x?}'", self.parsed_until),
+        }
+    }
+}
+
+impl Error for ParserError<'_> {}
+
+#[derive(PartialEq, Debug, Clone, Copy)]
+enum ParserNode {
+    SectionHeader,
+    ConfigName,
+    ConfigValue,
+    Comment,
+}
+
+impl Display for ParserNode {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::SectionHeader => write!(f, "section header"),
+            Self::ConfigName => write!(f, "config name"),
+            Self::ConfigValue => write!(f, "config value"),
+            Self::Comment => write!(f, "comment"),
+        }
     }
 }
 
@@ -499,19 +557,56 @@ pub fn parse_from_str(input: &str) -> Result<Parser<'_>, ParserError> {
 /// This generally is due to either invalid names or if there's extraneous
 /// data succeeding valid `git-config` data.
 pub fn parse_from_bytes(input: &[u8]) -> Result<Parser<'_>, ParserError> {
+    let mut newlines = 0;
     let (i, frontmatter) = many0(alt((
         map(comment, Event::Comment),
         map(take_spaces, |whitespace| {
             Event::Whitespace(Cow::Borrowed(whitespace.into()))
         }),
-        map(take_newline, |newline| {
+        map(take_newline, |(newline, counter)| {
+            newlines += counter;
             Event::Newline(Cow::Borrowed(newline.into()))
         }),
-    )))(input.as_bytes())?;
-    let (i, sections) = many0(section)(i)?;
+    )))(input.as_bytes())
+    // I don't think this can panic. many0 errors if the child parser returns
+    // a success where the input was not consumed, but alt will only return Ok
+    // if one of its children succeed. However, all of it's children are
+    // guaranteed to consume something if they succeed, so the Ok(i) == i case
+    // can never occur.
+    .expect("many0(alt(...)) panicked. Likely a bug in one of the children parser.");
 
+    if i.is_empty() {
+        return Ok(Parser {
+            frontmatter,
+            sections: vec![],
+        });
+    }
+
+    let mut node = ParserNode::SectionHeader;
+
+    let maybe_sections = many1(|i| section(i, &mut node))(i);
+    let (i, sections) = maybe_sections.map_err(|_| ParserError {
+        line_number: newlines,
+        last_attempted_parser: node,
+        parsed_until: i,
+    })?;
+
+    let sections = sections
+        .into_iter()
+        .map(|(section, additional_newlines)| {
+            newlines += additional_newlines;
+            section
+        })
+        .collect();
+
+    // This needs to happen after we collect sections, otherwise the line number
+    // will be off.
     if !i.is_empty() {
-        return Err(ParserError::ConfigHasExtraData(i.into()));
+        return Err(ParserError {
+            line_number: newlines,
+            last_attempted_parser: node,
+            parsed_until: i,
+        });
     }
 
     Ok(Parser {
@@ -532,28 +627,43 @@ fn comment(i: &[u8]) -> IResult<&[u8], ParsedComment> {
     ))
 }
 
-fn section(i: &[u8]) -> IResult<&[u8], ParsedSection> {
+fn section<'a, 'b>(
+    i: &'a [u8],
+    node: &'b mut ParserNode,
+) -> IResult<&'a [u8], (ParsedSection<'a>, usize)> {
     let (i, section_header) = section_header(i)?;
+
+    let mut newlines = 0;
+    // todo: unhack this (manually implement many0 and alt to avoid closure moves)
+    let node = std::sync::Mutex::new(node);
     let (i, items) = many0(alt((
         map(take_spaces, |space| {
             vec![Event::Whitespace(Cow::Borrowed(space.into()))]
         }),
-        map(take_newline, |newline| {
+        map(take_newline, |(newline, counter)| {
+            newlines += counter;
             vec![Event::Newline(Cow::Borrowed(newline.into()))]
         }),
-        map(section_body, |(key, values)| {
+        map(
+            |i| section_body(i, *node.lock().unwrap()),
+            |(key, values)| {
                 let mut vec = vec![Event::Key(Cow::Borrowed(key.into()))];
                 vec.extend(values);
                 vec
-        }),
+            },
+        ),
         map(comment, |comment| vec![Event::Comment(comment)]),
     )))(i)?;
+
     Ok((
         i,
+        (
             ParsedSection {
                 section_header,
                 events: items.into_iter().flatten().collect(),
             },
+            newlines,
+        ),
     ))
 }
 
@@ -588,7 +698,6 @@ fn section_header(i: &[u8]) -> IResult<&[u8], ParsedSectionHeader> {
     // Section header must be using modern subsection syntax at this point.
 
     let (i, whitespace) = take_spaces(i)?;
-
     let (i, subsection_name) = delimited(
         char('"'),
         opt(escaped(none_of("\"\\\n\0"), '\\', one_of(r#""\"#))),
@@ -609,11 +718,19 @@ fn section_header(i: &[u8]) -> IResult<&[u8], ParsedSectionHeader> {
     ))
 }
 
-fn section_body(i: &[u8]) -> IResult<&[u8], (&[u8], Vec<Event>)> {
+fn section_body<'a, 'b>(
+    i: &'a [u8],
+    node: &'b mut ParserNode,
+) -> IResult<&'a [u8], (&'a [u8], Vec<Event<'a>>)> {
     // maybe need to check for [ here
+    *node = ParserNode::ConfigName;
     let (i, name) = config_name(i)?;
+
     let (i, whitespace) = opt(take_spaces)(i)?;
+
+    *node = ParserNode::ConfigValue;
     let (i, value) = config_value(i)?;
+
     if let Some(whitespace) = whitespace {
         let mut events = vec![Event::Whitespace(Cow::Borrowed(whitespace.into()))];
         events.extend(value);
@@ -639,7 +756,6 @@ fn config_name(i: &[u8]) -> IResult<&[u8], &[u8]> {
             code: ErrorKind::Alpha,
         }));
     }
-
     take_while(|c: u8| (c as char).is_alphanumeric() || c == b'-')(i)
 }
 
@@ -751,19 +867,7 @@ fn value_impl(i: &[u8]) -> IResult<&[u8], Vec<Event>> {
 }
 
 fn take_spaces(i: &[u8]) -> IResult<&[u8], &[u8]> {
-    take_common(i, |c| (c as char).is_ascii() && is_space(c))
-}
-
-fn take_newline(i: &[u8]) -> IResult<&[u8], &[u8]> {
-    take_common(i, is_char_newline)
-}
-
-fn is_char_newline(c: u8) -> bool {
-    (c as char).is_ascii() && is_newline(c)
-}
-
-fn take_common<F: Fn(u8) -> bool>(i: &[u8], f: F) -> IResult<&[u8], &[u8]> {
-    let (i, v) = take_while(f)(i)?;
+    let (i, v) = take_while(|c| (c as char).is_ascii() && is_space(c))(i)?;
     if v.is_empty() {
         Err(nom::Err::Error(NomError {
             input: i,
@@ -774,15 +878,24 @@ fn take_common<F: Fn(u8) -> bool>(i: &[u8], f: F) -> IResult<&[u8], &[u8]> {
     }
 }
 
-#[cfg(test)]
-fn fully_consumed<T>(t: T) -> (&'static [u8], T) {
-    (&[], t)
+fn take_newline(i: &[u8]) -> IResult<&[u8], (&[u8], usize)> {
+    let mut counter = 0;
+    let (i, v) = take_while(|c| (c as char).is_ascii() && is_newline(c))(i)?;
+    counter += v.len();
+    if v.is_empty() {
+        Err(nom::Err::Error(NomError {
+            input: i,
+            code: ErrorKind::Eof,
+        }))
+    } else {
+        Ok((i, (v, counter)))
+    }
 }
 
 #[cfg(test)]
 mod comments {
     use super::*;
-    use crate::test_util::comment as parsed_comment;
+    use crate::test_util::{comment as parsed_comment, fully_consumed};
 
     #[test]
     fn semicolon() {
@@ -812,7 +925,7 @@ mod comments {
 #[cfg(test)]
 mod section_headers {
     use super::*;
-    use crate::test_util::section_header as parsed_section_header;
+    use crate::test_util::{fully_consumed, section_header as parsed_section_header};
 
     #[test]
     fn no_subsection() {
@@ -884,6 +997,7 @@ mod section_headers {
 #[cfg(test)]
 mod config_name {
     use super::*;
+    use crate::test_util::fully_consumed;
 
     #[test]
     fn just_name() {
@@ -912,8 +1026,9 @@ mod section_body {
 
     #[test]
     fn whitespace_is_not_ambigious() {
+        let mut node = ParserNode::SectionHeader;
         assert_eq!(
-            section_body(b"a =b").unwrap().1,
+            section_body(b"a =b", &mut node).unwrap().1,
             (
                 "a".as_bytes(),
                 vec![
@@ -924,7 +1039,7 @@ mod section_body {
             )
         );
         assert_eq!(
-            section_body(b"a= b").unwrap().1,
+            section_body(b"a= b", &mut node).unwrap().1,
             (
                 "a".as_bytes(),
                 vec![
@@ -940,7 +1055,7 @@ mod section_body {
 #[cfg(test)]
 mod value_no_continuation {
     use super::*;
-    use crate::test_util::value_event;
+    use crate::test_util::{fully_consumed, value_event};
 
     #[test]
     fn no_comment() {
@@ -1019,7 +1134,7 @@ mod value_no_continuation {
 #[cfg(test)]
 mod value_continuation {
     use super::*;
-    use crate::test_util::{newline_event, value_done_event, value_not_done_event};
+    use crate::test_util::{fully_consumed, newline_event, value_done_event, value_not_done_event};
 
     #[test]
     fn simple_continuation() {
@@ -1082,30 +1197,37 @@ mod value_continuation {
 mod section {
     use super::*;
     use crate::test_util::{
-        comment_event, name_event, newline_event, section_header as parsed_section_header,
-        value_done_event, value_event, value_not_done_event, whitespace_event,
+        comment_event, fully_consumed, name_event, newline_event,
+        section_header as parsed_section_header, value_done_event, value_event,
+        value_not_done_event, whitespace_event,
     };
 
     #[test]
     fn empty_section() {
+        let mut node = ParserNode::SectionHeader;
         assert_eq!(
-            section(b"[test]").unwrap(),
-            fully_consumed(ParsedSection {
+            section(b"[test]", &mut node).unwrap(),
+            fully_consumed((
+                ParsedSection {
                     section_header: parsed_section_header("test", None),
                     events: vec![]
-            })
+                },
+                0
+            )),
         );
     }
 
     #[test]
     fn simple_section() {
+        let mut node = ParserNode::SectionHeader;
         let section_data = br#"[hello]
             a = b
             c
             d = "lol""#;
         assert_eq!(
-            section(section_data).unwrap(),
-            fully_consumed(ParsedSection {
+            section(section_data, &mut node).unwrap(),
+            fully_consumed((
+                ParsedSection {
                     section_header: parsed_section_header("hello", None),
                     events: vec![
                         newline_event(),
@@ -1127,31 +1249,39 @@ mod section {
                         whitespace_event(" "),
                         value_event("\"lol\"")
                     ]
-            })
+                },
+                3
+            ))
         )
     }
 
     #[test]
     fn section_single_line() {
+        let mut node = ParserNode::SectionHeader;
         assert_eq!(
-            section(b"[hello] c").unwrap(),
-            fully_consumed(ParsedSection {
+            section(b"[hello] c", &mut node).unwrap(),
+            fully_consumed((
+                ParsedSection {
                     section_header: parsed_section_header("hello", None),
                     events: vec![whitespace_event(" "), name_event("c"), value_event("")]
-            })
+                },
+                0
+            ))
         );
     }
 
     #[test]
     fn section_very_commented() {
+        let mut node = ParserNode::SectionHeader;
         let section_data = br#"[hello] ; commentA
             a = b # commentB
             ; commentC
             ; commentD
             c = d"#;
         assert_eq!(
-            section(section_data).unwrap(),
-            fully_consumed(ParsedSection {
+            section(section_data, &mut node).unwrap(),
+            fully_consumed((
+                ParsedSection {
                     section_header: parsed_section_header("hello", None),
                     events: vec![
                         whitespace_event(" "),
@@ -1179,16 +1309,24 @@ mod section {
                         whitespace_event(" "),
                         value_event("d"),
                     ]
-            })
+                },
+                4
+            ))
         );
     }
 
     #[test]
     fn complex_continuation() {
+        let mut node = ParserNode::SectionHeader;
         // This test is absolute hell. Good luck if this fails.
         assert_eq!(
-            section(b"[section] a = 1    \"\\\"\\\na ; e \"\\\"\\\nd # \"b\t ; c").unwrap(),
-            fully_consumed(ParsedSection {
+            section(
+                b"[section] a = 1    \"\\\"\\\na ; e \"\\\"\\\nd # \"b\t ; c",
+                &mut node
+            )
+            .unwrap(),
+            fully_consumed((
+                ParsedSection {
                     section_header: parsed_section_header("section", None),
                     events: vec![
                         whitespace_event(" "),
@@ -1204,15 +1342,19 @@ mod section {
                         whitespace_event(" "),
                         comment_event('#', " \"b\t ; c"),
                     ]
-            })
+                },
+                0
+            ))
         );
     }
 
     #[test]
     fn quote_split_over_two_lines() {
+        let mut node = ParserNode::SectionHeader;
         assert_eq!(
-            section(b"[section \"a\"] b =\"\\\n;\";a").unwrap(),
-            fully_consumed(ParsedSection {
+            section(b"[section \"a\"] b =\"\\\n;\";a", &mut node).unwrap(),
+            fully_consumed((
+                ParsedSection {
                     section_header: parsed_section_header("section", (" ", "a")),
                     events: vec![
                         whitespace_event(" "),
@@ -1224,15 +1366,19 @@ mod section {
                         value_done_event(";\""),
                         comment_event(';', "a"),
                     ]
-            })
+                },
+                0
+            ))
         )
     }
 
     #[test]
     fn section_handles_extranous_whitespace_before_comment() {
+        let mut node = ParserNode::SectionHeader;
         assert_eq!(
-            section(b"[s]hello             #world").unwrap(),
-            fully_consumed(ParsedSection {
+            section(b"[s]hello             #world", &mut node).unwrap(),
+            fully_consumed((
+                ParsedSection {
                     section_header: parsed_section_header("s", None),
                     events: vec![
                         name_event("hello"),
@@ -1240,7 +1386,9 @@ mod section {
                         value_event(""),
                         comment_event('#', "world"),
                     ]
-            })
+                },
+                0
+            ))
         );
     }
 }

src/test_util.rs

@@ -67,3 +67,7 @@ pub(crate) fn comment(comment_tag: char, comment: &'static str) -> ParsedComment
         comment: Cow::Borrowed(comment.into()),
     }
 }
+
+pub(crate) fn fully_consumed<T>(t: T) -> (&'static [u8], T) {
+    (&[], t)
+}

tests/parser_integration_tests.rs

@@ -216,3 +216,13 @@ fn newline_events_are_merged() {
         vec![newline_custom("\n\n\n\n\n")]
     );
 }
+
+#[test]
+fn error() {
+    let input = "[core] a=b\n 4a=3";
+    println!("{}", parse_from_str(input).unwrap_err());
+    let input = "[core] a=b\n =3";
+    println!("{}", parse_from_str(input).unwrap_err());
+    let input = "[core";
+    println!("{}", parse_from_str(input).unwrap_err());
+}