Compare commits

..

No commits in common. "594ee8ac0c5620be279c7520e9f4478ecf4eb582" and "93e6e2eed3f855786d1686dcfce1b2f4f21e4fb7" have entirely different histories.

3 changed files with 548 additions and 894 deletions

View file

@ -2,25 +2,14 @@ use std::collections::HashMap;
use crate::parser::{parse_from_str, Event, Parser, ParserError};
#[derive(PartialEq, Eq, Hash, Copy, Clone, PartialOrd, Ord, Debug)]
#[derive(PartialEq, Eq, Hash, Copy, Clone, PartialOrd, Ord)]
struct SectionId(usize);
#[derive(Debug, PartialEq, Eq)]
enum LookupTreeNode<'a> {
Terminal(Vec<SectionId>),
NonTerminal(HashMap<&'a str, Vec<SectionId>>),
}
#[derive(Debug, PartialEq, Eq)]
pub enum GitConfigError<'a> {
/// The requested section does not exist.
SectionDoesNotExist(&'a str),
/// The requested subsection does not exist.
SubSectionDoesNotExist(Option<&'a str>),
/// The key does not exist in the requested section.
KeyDoesNotExist(&'a str),
}
/// High level `git-config` reader and writer.
pub struct GitConfig<'a> {
front_matter_events: Vec<Event<'a>>,
@ -38,13 +27,14 @@ impl<'a> GitConfig<'a> {
}
pub fn from_parser(parser: Parser<'a>) -> Self {
let mut new_self = Self {
front_matter_events: vec![],
sections: HashMap::new(),
section_lookup_tree: HashMap::new(),
section_header_separators: HashMap::new(),
section_id_counter: 0,
};
// Monotonically increasing
let mut section_id_counter: usize = 0;
// Fields for the struct
let mut front_matter_events: Vec<Event<'a>> = vec![];
let mut sections: HashMap<SectionId, Vec<Event<'a>>> = HashMap::new();
let mut section_lookup_tree: HashMap<&str, Vec<LookupTreeNode>> = HashMap::new();
let mut section_header_separators = HashMap::new();
// Current section that we're building
let mut current_section_name: Option<&str> = None;
@ -53,63 +43,16 @@ impl<'a> GitConfig<'a> {
for event in parser.into_iter() {
match event {
Event::SectionHeader(header) => {
new_self.push_section(
&mut current_section_name,
&mut current_subsection_name,
&mut maybe_section,
);
// Initialize new section
let (name, subname) = (header.name, header.subsection_name);
maybe_section = Some(vec![]);
current_section_name = Some(name);
current_subsection_name = subname;
// We need to store the new, current id counter, so don't
// use new_section_id here and use the already incremented
// section id value.
new_self
.section_header_separators
.insert(SectionId(new_self.section_id_counter), header.separator);
}
e @ Event::Key(_)
| e @ Event::Value(_)
| e @ Event::ValueNotDone(_)
| e @ Event::ValueDone(_) => maybe_section
.as_mut()
.expect("Got a section-only event before a section")
.push(e),
e @ Event::Comment(_) | e @ Event::Newline(_) | e @ Event::Whitespace(_) => {
match maybe_section {
e @ Event::Comment(_) => match maybe_section {
Some(ref mut section) => section.push(e),
None => new_self.front_matter_events.push(e),
}
}
}
}
// The last section doesn't get pushed since we only push if there's a
// new section header, so we need to call push one more time.
new_self.push_section(
&mut current_section_name,
&mut current_subsection_name,
&mut maybe_section,
);
new_self
}
fn push_section(
&mut self,
current_section_name: &mut Option<&'a str>,
current_subsection_name: &mut Option<&'a str>,
maybe_section: &mut Option<Vec<Event<'a>>>,
) {
let new_section_id = SectionId(self.section_id_counter);
None => front_matter_events.push(e),
},
Event::SectionHeader(header) => {
// Push current section to struct
let new_section_id = SectionId(section_id_counter);
if let Some(section) = maybe_section.take() {
self.sections.insert(new_section_id, section);
let lookup = self
.section_lookup_tree
sections.insert(new_section_id, section);
let lookup = section_lookup_tree
.entry(current_section_name.unwrap())
.or_default();
@ -127,7 +70,7 @@ impl<'a> GitConfig<'a> {
}
if !found_node {
let mut map = HashMap::new();
map.insert(*subsection_name, vec![new_section_id]);
map.insert(subsection_name, vec![new_section_id]);
lookup.push(LookupTreeNode::NonTerminal(map));
}
} else {
@ -142,40 +85,58 @@ impl<'a> GitConfig<'a> {
lookup.push(LookupTreeNode::Terminal(vec![new_section_id]))
}
}
self.section_id_counter += 1;
section_id_counter += 1;
}
// Initialize new section
let (name, subname) = (header.name, header.subsection_name);
maybe_section = Some(vec![]);
current_section_name = Some(name);
current_subsection_name = subname;
// We need to store the new, current id counter, so don't
// use new_section_id here and use the already incremented
// section id value.
section_header_separators
.insert(SectionId(section_id_counter), header.separator);
}
e @ Event::Key(_) => maybe_section
.as_mut()
.expect("Got a section-only event before a section")
.push(e),
e @ Event::Value(_) => maybe_section
.as_mut()
.expect("Got a section-only event before a section")
.push(e),
e @ Event::Newline(_) => match maybe_section {
Some(ref mut section) => section.push(e),
None => front_matter_events.push(e),
},
e @ Event::ValueNotDone(_) => maybe_section
.as_mut()
.expect("Got a section-only event before a section")
.push(e),
e @ Event::ValueDone(_) => maybe_section
.as_mut()
.expect("Got a section-only event before a section")
.push(e),
e @ Event::Whitespace(_) => match maybe_section {
Some(ref mut section) => section.push(e),
None => front_matter_events.push(e),
},
}
}
/// Returns an uninterpreted value given a section and optional subsection
/// and key.
///
/// Note that `git-config` follows a "last-one-wins" rule for single values.
/// If multiple sections contain the same key, then the last section's last
/// key's value will be returned.
///
/// Concretely, if you have the following config:
///
/// ```text
/// [core]
/// a = b
/// [core]
/// a = c
/// a = d
/// ```
///
/// Then this function will return `d`:
///
/// ```
/// # use serde_git_config::config::GitConfig;
/// # let git_config = GitConfig::from_str("[core]a=b\n[core]\na=c\na=d").unwrap();
/// assert_eq!(git_config.get_raw_value("core", None, "a"), Ok("d"));
/// ```
///
/// # Errors
///
/// This function will return an error if the key is not in the requested
/// section and subsection.
pub fn get_raw_value<'b>(
Self {
front_matter_events,
section_lookup_tree,
sections,
section_header_separators,
section_id_counter,
}
}
pub fn get_raw_single_value<'b>(
&self,
section_name: &'b str,
subsection_name: Option<&'b str>,
@ -184,38 +145,33 @@ impl<'a> GitConfig<'a> {
// Note: cannot wrap around the raw_multi_value method because we need
// to guarantee that the highest section id is used (so that we follow
// the "last one wins" resolution strategy by `git-config`).
let section_id = self.get_section_id_by_name_and_subname(section_name, subsection_name)?;
let section_id = self
.get_section_id_by_name_and_subname(section_name, subsection_name)
.ok_or(GitConfigError::SubSectionDoesNotExist(subsection_name))?;
// section_id is guaranteed to exist in self.sections, else we have a
// violated invariant.
let events = self.sections.get(&section_id).unwrap();
let mut found_key = false;
let mut latest_value = None;
for event in events {
match event {
Event::Key(event_key) if *event_key == key => found_key = true,
Event::Value(v) if found_key => {
found_key = false;
latest_value = Some(*v);
}
Event::Value(v) if found_key => return Ok(v),
_ => (),
}
}
latest_value.ok_or(GitConfigError::KeyDoesNotExist(key))
Err(GitConfigError::KeyDoesNotExist(key))
}
fn get_section_id_by_name_and_subname<'b>(
&'a self,
section_name: &'b str,
subsection_name: Option<&'b str>,
) -> Result<SectionId, GitConfigError<'b>> {
) -> Option<SectionId> {
self.get_section_ids_by_name_and_subname(section_name, subsection_name)
.map(|vec| {
// get_section_ids_by_name_and_subname is guaranteed to return
// a non-empty vec, so max can never return empty.
*vec.into_iter().max().unwrap()
})
.map(|vec| vec.into_iter().max())
.flatten()
}
pub fn get_raw_multi_value<'b>(
@ -225,33 +181,26 @@ impl<'a> GitConfig<'a> {
key: &'b str,
) -> Result<Vec<&'a str>, GitConfigError<'b>> {
let values = self
.get_section_ids_by_name_and_subname(section_name, subsection_name)?
.get_section_ids_by_name_and_subname(section_name, subsection_name)
.ok_or(GitConfigError::SubSectionDoesNotExist(subsection_name))?
.iter()
.map(|section_id| {
let mut found_key = false;
let mut events = vec![];
// section_id is guaranteed to exist in self.sections, else we have a
// violated invariant.
for event in self.sections.get(section_id).unwrap() {
match event {
Event::Key(event_key) if *event_key == key => found_key = true,
Event::Value(v) if found_key => {
events.push(*v);
found_key = false;
}
Event::Value(v) if found_key => return Ok(*v),
_ => (),
}
}
if events.is_empty() {
Err(GitConfigError::KeyDoesNotExist(key))
} else {
Ok(events)
}
})
.filter_map(Result::ok)
.flatten()
.collect::<Vec<_>>();
if values.is_empty() {
Err(GitConfigError::KeyDoesNotExist(key))
} else {
@ -263,324 +212,32 @@ impl<'a> GitConfig<'a> {
&'a self,
section_name: &'b str,
subsection_name: Option<&'b str>,
) -> Result<&[SectionId], GitConfigError<'b>> {
let section_ids = self
.section_lookup_tree
.get(section_name)
.ok_or(GitConfigError::SectionDoesNotExist(section_name))?;
let mut maybe_ids = None;
// Don't simplify if and matches here -- the for loop currently needs
// `n + 1` checks, while the if and matches will result in the for loop
// needing `2n` checks.
) -> Option<Vec<SectionId>> {
let section_ids = self.section_lookup_tree.get(section_name)?;
if let Some(subsect_name) = subsection_name {
let mut maybe_ids = None;
for node in section_ids {
if let LookupTreeNode::NonTerminal(subsection_lookup) = node {
maybe_ids = subsection_lookup.get(subsect_name);
break;
}
}
maybe_ids.map(|vec| vec.clone())
} else {
let mut maybe_ids = None;
for node in section_ids {
if let LookupTreeNode::Terminal(subsection_lookup) = node {
maybe_ids = Some(subsection_lookup);
maybe_ids = subsection_lookup.iter().max();
break;
}
}
maybe_ids.map(|v| vec![*v])
}
maybe_ids
.map(Vec::as_slice)
.ok_or(GitConfigError::SubSectionDoesNotExist(subsection_name))
}
}
#[cfg(test)]
mod from_parser {
use super::*;
#[test]
fn parse_empty() {
let config = GitConfig::from_str("").unwrap();
assert!(config.section_header_separators.is_empty());
assert_eq!(config.section_id_counter, 0);
assert!(config.section_lookup_tree.is_empty());
assert!(config.sections.is_empty());
}
#[test]
fn parse_single_section() {
let config = GitConfig::from_str("[core]\na=b\nc=d").unwrap();
let expected_separators = {
let mut map = HashMap::new();
map.insert(SectionId(0), None);
map
};
assert_eq!(config.section_header_separators, expected_separators);
assert_eq!(config.section_id_counter, 1);
let expected_lookup_tree = {
let mut tree = HashMap::new();
tree.insert("core", vec![LookupTreeNode::Terminal(vec![SectionId(0)])]);
tree
};
assert_eq!(config.section_lookup_tree, expected_lookup_tree);
let expected_sections = {
let mut sections = HashMap::new();
sections.insert(
SectionId(0),
vec![
Event::Newline("\n"),
Event::Key("a"),
Event::Value("b"),
Event::Newline("\n"),
Event::Key("c"),
Event::Value("d"),
],
);
sections
};
assert_eq!(config.sections, expected_sections);
}
#[test]
fn parse_single_subsection() {
let config = GitConfig::from_str("[core.subsec]\na=b\nc=d").unwrap();
let expected_separators = {
let mut map = HashMap::new();
map.insert(SectionId(0), Some("."));
map
};
assert_eq!(config.section_header_separators, expected_separators);
assert_eq!(config.section_id_counter, 1);
let expected_lookup_tree = {
let mut tree = HashMap::new();
let mut inner_tree = HashMap::new();
inner_tree.insert("subsec", vec![SectionId(0)]);
tree.insert("core", vec![LookupTreeNode::NonTerminal(inner_tree)]);
tree
};
assert_eq!(config.section_lookup_tree, expected_lookup_tree);
let expected_sections = {
let mut sections = HashMap::new();
sections.insert(
SectionId(0),
vec![
Event::Newline("\n"),
Event::Key("a"),
Event::Value("b"),
Event::Newline("\n"),
Event::Key("c"),
Event::Value("d"),
],
);
sections
};
assert_eq!(config.sections, expected_sections);
}
#[test]
fn parse_multiple_sections() {
let config = GitConfig::from_str("[core]\na=b\nc=d\n[other]e=f").unwrap();
let expected_separators = {
let mut map = HashMap::new();
map.insert(SectionId(0), None);
map.insert(SectionId(1), None);
map
};
assert_eq!(config.section_header_separators, expected_separators);
assert_eq!(config.section_id_counter, 2);
let expected_lookup_tree = {
let mut tree = HashMap::new();
tree.insert("core", vec![LookupTreeNode::Terminal(vec![SectionId(0)])]);
tree.insert("other", vec![LookupTreeNode::Terminal(vec![SectionId(1)])]);
tree
};
assert_eq!(config.section_lookup_tree, expected_lookup_tree);
let expected_sections = {
let mut sections = HashMap::new();
sections.insert(
SectionId(0),
vec![
Event::Newline("\n"),
Event::Key("a"),
Event::Value("b"),
Event::Newline("\n"),
Event::Key("c"),
Event::Value("d"),
Event::Newline("\n"),
],
);
sections.insert(SectionId(1), vec![Event::Key("e"), Event::Value("f")]);
sections
};
assert_eq!(config.sections, expected_sections);
}
#[test]
fn parse_multiple_duplicate_sections() {
let config = GitConfig::from_str("[core]\na=b\nc=d\n[core]e=f").unwrap();
let expected_separators = {
let mut map = HashMap::new();
map.insert(SectionId(0), None);
map.insert(SectionId(1), None);
map
};
assert_eq!(config.section_header_separators, expected_separators);
assert_eq!(config.section_id_counter, 2);
let expected_lookup_tree = {
let mut tree = HashMap::new();
tree.insert(
"core",
vec![LookupTreeNode::Terminal(vec![SectionId(0), SectionId(1)])],
);
tree
};
assert_eq!(config.section_lookup_tree, expected_lookup_tree);
let expected_sections = {
let mut sections = HashMap::new();
sections.insert(
SectionId(0),
vec![
Event::Newline("\n"),
Event::Key("a"),
Event::Value("b"),
Event::Newline("\n"),
Event::Key("c"),
Event::Value("d"),
Event::Newline("\n"),
],
);
sections.insert(SectionId(1), vec![Event::Key("e"), Event::Value("f")]);
sections
};
assert_eq!(config.sections, expected_sections);
}
}
#[cfg(test)]
mod get_raw_value {
use super::*;
#[test]
fn single_section() {
let config = GitConfig::from_str("[core]\na=b\nc=d").unwrap();
assert_eq!(config.get_raw_value("core", None, "a"), Ok("b"));
assert_eq!(config.get_raw_value("core", None, "c"), Ok("d"));
}
#[test]
fn last_one_wins_respected_in_section() {
let config = GitConfig::from_str("[core]\na=b\na=d").unwrap();
assert_eq!(config.get_raw_value("core", None, "a"), Ok("d"));
}
#[test]
fn last_one_wins_respected_across_section() {
let config = GitConfig::from_str("[core]\na=b\n[core]\na=d").unwrap();
assert_eq!(config.get_raw_value("core", None, "a"), Ok("d"));
}
#[test]
fn section_not_found() {
let config = GitConfig::from_str("[core]\na=b\nc=d").unwrap();
assert_eq!(
config.get_raw_value("foo", None, "a"),
Err(GitConfigError::SectionDoesNotExist("foo"))
);
}
#[test]
fn subsection_not_found() {
let config = GitConfig::from_str("[core]\na=b\nc=d").unwrap();
assert_eq!(
config.get_raw_value("core", Some("a"), "a"),
Err(GitConfigError::SubSectionDoesNotExist(Some("a")))
);
}
#[test]
fn key_not_found() {
let config = GitConfig::from_str("[core]\na=b\nc=d").unwrap();
assert_eq!(
config.get_raw_value("core", None, "aaaaaa"),
Err(GitConfigError::KeyDoesNotExist("aaaaaa"))
);
}
#[test]
fn subsection_must_be_respected() {
let config = GitConfig::from_str("[core]a=b\n[core.a]a=c").unwrap();
assert_eq!(config.get_raw_value("core", None, "a"), Ok("b"));
assert_eq!(config.get_raw_value("core", Some("a"), "a"), Ok("c"));
}
}
#[cfg(test)]
mod get_raw_multi_value {
use super::*;
#[test]
fn single_value_is_identical_to_single_value_query() {
let config = GitConfig::from_str("[core]\na=b\nc=d").unwrap();
assert_eq!(
vec![config.get_raw_value("core", None, "a").unwrap()],
config.get_raw_multi_value("core", None, "a").unwrap()
);
}
#[test]
fn multi_value_in_section() {
let config = GitConfig::from_str("[core]\na=b\na=c").unwrap();
assert_eq!(
config.get_raw_multi_value("core", None, "a").unwrap(),
vec!["b", "c"]
);
}
#[test]
fn multi_value_across_sections() {
let config = GitConfig::from_str("[core]\na=b\na=c\n[core]a=d").unwrap();
assert_eq!(
config.get_raw_multi_value("core", None, "a").unwrap(),
vec!["b", "c", "d"]
);
}
#[test]
fn section_not_found() {
let config = GitConfig::from_str("[core]\na=b\nc=d").unwrap();
assert_eq!(
config.get_raw_multi_value("foo", None, "a"),
Err(GitConfigError::SectionDoesNotExist("foo"))
);
}
#[test]
fn subsection_not_found() {
let config = GitConfig::from_str("[core]\na=b\nc=d").unwrap();
assert_eq!(
config.get_raw_multi_value("core", Some("a"), "a"),
Err(GitConfigError::SubSectionDoesNotExist(Some("a")))
);
}
#[test]
fn key_not_found() {
let config = GitConfig::from_str("[core]\na=b\nc=d").unwrap();
assert_eq!(
config.get_raw_multi_value("core", None, "aaaaaa"),
Err(GitConfigError::KeyDoesNotExist("aaaaaa"))
);
}
#[test]
fn subsection_must_be_respected() {
let config = GitConfig::from_str("[core]a=b\n[core.a]a=c").unwrap();
assert_eq!(
config.get_raw_multi_value("core", None, "a").unwrap(),
vec!["b"]
);
assert_eq!(
config.get_raw_multi_value("core", Some("a"), "a").unwrap(),
vec!["c"]
);
}
pub enum GitConfigError<'a> {
SectionDoesNotExist(&'a str),
SubSectionDoesNotExist(Option<&'a str>),
KeyDoesNotExist(&'a str),
}

View file

@ -1,19 +1,18 @@
//! This module handles parsing a `git-config`. Generally speaking, you want to
//! use a higher abstraction such as [`GitConfig`] unless you have some explicit
//! reason to work with events instead.
//! use a higher abstraction unless you have some explicit reason to work with
//! events instead.
//!
//! The general workflow for interacting with this is to use one of the
//! `parse_from_*` function variants. These will return a [`Parser`] on success,
//! which can be converted into an [`Event`] iterator. The [`Parser`] also has
//! additional methods for accessing leading comments or events by section.
//!
//! [`GitConfig`]: crate::config::GitConfig
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::many1;
use nom::sequence::delimited;
use nom::IResult;
use nom::{branch::alt, multi::many0};
@ -361,7 +360,7 @@ impl<'a> Parser<'a> {
/// data succeeding valid `git-config` data.
pub fn parse_from_str(input: &str) -> Result<Parser<'_>, ParserError> {
let (i, comments) = many0(comment)(input)?;
let (i, sections) = many0(section)(i)?;
let (i, sections) = many1(section)(i)?;
if !i.is_empty() {
return Err(ParserError::ConfigHasExtraData(i));
@ -618,17 +617,18 @@ fn take_common<'a, F: Fn(char) -> bool>(i: &'a str, f: F) -> IResult<&'a str, &'
Ok((i, v))
}
}
#[cfg(test)]
fn fully_consumed<T>(t: T) -> (&'static str, T) {
mod parse {
use super::*;
fn fully_consumed<T>(t: T) -> (&'static str, T) {
("", t)
}
}
#[cfg(test)]
fn gen_section_header(
fn gen_section_header(
name: &str,
subsection: impl Into<Option<(&'static str, &'static str)>>,
) -> ParsedSectionHeader<'_> {
) -> ParsedSectionHeader<'_> {
if let Some((separator, subsection_name)) = subsection.into() {
ParsedSectionHeader {
name,
@ -642,10 +642,10 @@ fn gen_section_header(
subsection_name: None,
}
}
}
}
#[cfg(test)]
mod comments {
mod comments {
use super::super::*;
use super::*;
#[test]
@ -680,10 +680,10 @@ mod comments {
})
);
}
}
}
#[cfg(test)]
mod section_headers {
mod section_headers {
use super::super::*;
use super::*;
#[test]
@ -751,10 +751,10 @@ mod section_headers {
fully_consumed(gen_section_header("hello", (" ", "]")))
);
}
}
}
#[cfg(test)]
mod config_name {
mod config_name {
use super::super::*;
use super::*;
#[test]
@ -772,10 +772,10 @@ mod config_name {
fn cannot_be_empty() {
assert!(config_name("").is_err())
}
}
}
#[cfg(test)]
mod value_no_continuation {
mod value_no_continuation {
use super::super::*;
use super::*;
#[test]
@ -850,10 +850,10 @@ mod value_no_continuation {
fn garbage_after_continution_is_err() {
assert!(value_impl("hello \\afwjdls").is_err());
}
}
}
#[cfg(test)]
mod value_continuation {
mod value_continuation {
use super::super::*;
use super::*;
#[test]
@ -911,10 +911,10 @@ mod value_continuation {
)
)
}
}
}
#[cfg(test)]
mod section {
mod section {
use super::super::*;
use super::*;
#[test]
@ -1089,4 +1089,5 @@ mod section {
})
);
}
}
}

View file

@ -62,7 +62,8 @@ fn personal_config() {
assert_eq!(
parse_from_str(config)
.unwrap()
.into_vec(),
.into_iter()
.collect::<Vec<_>>(),
vec![
gen_section_header("user", None),
newline(),
@ -162,8 +163,3 @@ fn personal_config() {
]
);
}
#[test]
fn parse_empty() {
assert_eq!(parse_from_str("").unwrap().into_vec(), vec![]);
}