bat/build/syntax_mapping.rs

294 lines
9 KiB
Rust

use std::{
convert::Infallible,
env, fs,
path::{Path, PathBuf},
str::FromStr,
};
use anyhow::{anyhow, bail};
use indexmap::IndexMap;
use itertools::Itertools;
use once_cell::sync::Lazy;
use regex::Regex;
use serde_derive::Deserialize;
use serde_with::DeserializeFromStr;
use walkdir::WalkDir;
/// Known mapping targets.
///
/// Corresponds to `syntax_mapping::MappingTarget`.
#[allow(clippy::enum_variant_names)]
#[derive(Clone, Debug, Eq, PartialEq, Hash, DeserializeFromStr)]
pub enum MappingTarget {
MapTo(String),
MapToUnknown,
MapExtensionToUnknown,
}
impl FromStr for MappingTarget {
type Err = Infallible;
fn from_str(s: &str) -> Result<Self, Self::Err> {
match s {
"MappingTarget::MapToUnknown" => Ok(Self::MapToUnknown),
"MappingTarget::MapExtensionToUnknown" => Ok(Self::MapExtensionToUnknown),
syntax => Ok(Self::MapTo(syntax.into())),
}
}
}
impl MappingTarget {
fn codegen(&self) -> String {
match self {
Self::MapTo(syntax) => format!(r###"MappingTarget::MapTo(r#"{syntax}"#)"###),
Self::MapToUnknown => "MappingTarget::MapToUnknown".into(),
Self::MapExtensionToUnknown => "MappingTarget::MapExtensionToUnknown".into(),
}
}
}
#[derive(Clone, Debug, PartialEq, Eq, Hash, DeserializeFromStr)]
/// A single matcher.
///
/// Codegen converts this into a `Lazy<Option<GlobMatcher>>`.
struct Matcher(Vec<MatcherSegment>);
/// Parse a matcher.
///
/// Note that this implementation is rather strict: it will greedily interpret
/// every valid environment variable replacement as such, then immediately
/// hard-error if it finds a '$' anywhere in the remaining text segments.
///
/// The reason for this strictness is I currently cannot think of a valid reason
/// why you would ever need '$' as plaintext in a glob pattern. Therefore any
/// such occurrences are likely human errors.
///
/// If we later discover some edge cases, it's okay to make it more permissive.
///
/// Revision history:
/// - 2024-02-20: allow `{` and `}` (glob brace expansion)
impl FromStr for Matcher {
type Err = anyhow::Error;
fn from_str(s: &str) -> Result<Self, Self::Err> {
use MatcherSegment as Seg;
static VAR_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(r"\$\{([\w\d_]+)\}").unwrap());
let mut segments = vec![];
let mut text_start = 0;
for capture in VAR_REGEX.captures_iter(s) {
let match_0 = capture.get(0).unwrap();
// text before this var
let text_end = match_0.start();
segments.push(Seg::Text(s[text_start..text_end].into()));
text_start = match_0.end();
// this var
segments.push(Seg::Env(capture.get(1).unwrap().as_str().into()));
}
// possible trailing text
segments.push(Seg::Text(s[text_start..].into()));
// cleanup empty text segments
let non_empty_segments = segments
.into_iter()
.filter(|seg| seg.text().map(|t| !t.is_empty()).unwrap_or(true))
.collect_vec();
// sanity check
if non_empty_segments
.windows(2)
.any(|segs| segs[0].is_text() && segs[1].is_text())
{
unreachable!("Parsed into consecutive text segments: {non_empty_segments:?}");
}
// guard empty case
if non_empty_segments.is_empty() {
bail!(r#"Parsed an empty matcher: "{s}""#);
}
// guard variable syntax leftover fragments
if non_empty_segments
.iter()
.filter_map(Seg::text)
.any(|t| t.contains('$'))
{
bail!(r#"Invalid matcher: "{s}""#);
}
Ok(Self(non_empty_segments))
}
}
impl Matcher {
fn codegen(&self) -> String {
match self.0.len() {
0 => unreachable!("0-length matcher should never be created"),
// if-let guard would be ideal here
// see: https://github.com/rust-lang/rust/issues/51114
1 if self.0[0].is_text() => {
let s = self.0[0].text().unwrap();
format!(r###"Lazy::new(|| Some(build_matcher_fixed(r#"{s}"#)))"###)
}
// parser logic ensures that this case can only happen when there are dynamic segments
_ => {
let segs = self.0.iter().map(MatcherSegment::codegen).join(", ");
format!(r###"Lazy::new(|| build_matcher_dynamic(&[{segs}]))"###)
}
}
}
}
/// A segment in a matcher.
///
/// Corresponds to `syntax_mapping::MatcherSegment`.
#[derive(Debug, Clone, PartialEq, Eq, Hash)]
enum MatcherSegment {
Text(String),
Env(String),
}
#[allow(dead_code)]
impl MatcherSegment {
fn is_text(&self) -> bool {
matches!(self, Self::Text(_))
}
fn is_env(&self) -> bool {
matches!(self, Self::Env(_))
}
fn text(&self) -> Option<&str> {
match self {
Self::Text(t) => Some(t),
Self::Env(_) => None,
}
}
fn env(&self) -> Option<&str> {
match self {
Self::Text(_) => None,
Self::Env(t) => Some(t),
}
}
fn codegen(&self) -> String {
match self {
Self::Text(s) => format!(r###"MatcherSegment::Text(r#"{s}"#)"###),
Self::Env(s) => format!(r###"MatcherSegment::Env(r#"{s}"#)"###),
}
}
}
/// A struct that models a single .toml file in /src/syntax_mapping/builtins/.
#[derive(Clone, Debug, Deserialize)]
struct MappingDefModel {
mappings: IndexMap<MappingTarget, Vec<Matcher>>,
}
impl MappingDefModel {
fn into_mapping_list(self) -> MappingList {
let list = self
.mappings
.into_iter()
.flat_map(|(target, matchers)| {
matchers
.into_iter()
.map(|matcher| (matcher, target.clone()))
.collect::<Vec<_>>()
})
.collect();
MappingList(list)
}
}
#[derive(Clone, Debug)]
struct MappingList(Vec<(Matcher, MappingTarget)>);
impl MappingList {
fn codegen(&self) -> String {
let array_items: Vec<_> = self
.0
.iter()
.map(|(matcher, target)| {
format!("({m}, {t})", m = matcher.codegen(), t = target.codegen())
})
.collect();
let len = array_items.len();
format!(
"/// Generated by build script from /src/syntax_mapping/builtins/.\n\
pub(crate) static BUILTIN_MAPPINGS: [(Lazy<Option<GlobMatcher>>, MappingTarget); {len}] = [\n{items}\n];",
items = array_items.join(",\n")
)
}
}
/// Get the list of paths to all mapping definition files that should be
/// included for the current target platform.
fn get_def_paths() -> anyhow::Result<Vec<PathBuf>> {
let source_subdirs = [
"common",
#[cfg(target_family = "unix")]
"unix-family",
#[cfg(any(
target_os = "freebsd",
target_os = "netbsd",
target_os = "openbsd",
target_os = "macos"
))]
"bsd-family",
#[cfg(target_os = "linux")]
"linux",
#[cfg(target_os = "macos")]
"macos",
#[cfg(target_os = "windows")]
"windows",
];
let mut toml_paths = vec![];
for subdir in source_subdirs {
let wd = WalkDir::new(Path::new("src/syntax_mapping/builtins").join(subdir));
let paths = wd
.into_iter()
.filter_map_ok(|entry| {
let path = entry.path();
(path.is_file() && path.extension().map(|ext| ext == "toml").unwrap_or(false))
.then(|| path.to_owned())
})
.collect::<Result<Vec<_>, _>>()?;
toml_paths.extend(paths);
}
toml_paths.sort_by_key(|path| {
path.file_name()
.expect("file name should not terminate in ..")
.to_owned()
});
Ok(toml_paths)
}
fn read_all_mappings() -> anyhow::Result<MappingList> {
let mut all_mappings = vec![];
for path in get_def_paths()? {
let toml_string = fs::read_to_string(path)?;
let mappings = toml::from_str::<MappingDefModel>(&toml_string)?.into_mapping_list();
all_mappings.extend(mappings.0);
}
let duplicates = all_mappings
.iter()
.duplicates_by(|(matcher, _)| matcher)
.collect_vec();
if !duplicates.is_empty() {
bail!("Rules with duplicate matchers found: {duplicates:?}");
}
Ok(MappingList(all_mappings))
}
/// Build the static syntax mappings defined in /src/syntax_mapping/builtins/
/// into a .rs source file, which is to be inserted with `include!`.
pub fn build_static_mappings() -> anyhow::Result<()> {
println!("cargo:rerun-if-changed=src/syntax_mapping/builtins/");
let mappings = read_all_mappings()?;
let codegen_path = Path::new(&env::var_os("OUT_DIR").ok_or(anyhow!("OUT_DIR is unset"))?)
.join("codegen_static_syntax_mappings.rs");
fs::write(codegen_path, mappings.codegen())?;
Ok(())
}