use std::ffi::{OsStr, OsString};
use std::io::{self, BufWriter, IsTerminal, Read, Write};
use std::path::Path;
use std::process::ExitCode;
use std::sync::atomic::{AtomicUsize, Ordering};
use std::sync::{Arc, Mutex};
use jsonvet::{Dialect, Report, Severity};
mod args;
mod glob;
mod json;
mod output;
mod rules;
mod sarif;
mod schema;
mod settings;
use args::{Command, STDIN};
use output::Printer;
use schema::Schema;
use settings::{Resolver, Settings};
const SCHEMA_MAX_DEPTH: usize = 512;
const STACK: usize = 64 << 20;
const CHUNK_BYTES: usize = 64 << 20;
const SCHEMA_BYTES: usize = 20;
const SPLIT_BYTES: usize = 1 << 20;
pub fn run() -> ExitCode {
on_big_stack(run_here)
}
fn on_big_stack<T: Send + 'static>(f: fn() -> T) -> T {
let thread = std::thread::Builder::new().name("jsonvet".into());
match thread.stack_size(STACK).spawn(f) {
Ok(thread) => thread
.join()
.unwrap_or_else(|panic| std::panic::resume_unwind(panic)),
Err(_) => f(),
}
}
fn run_here() -> ExitCode {
let annotate = std::env::var_os("GITHUB_ACTIONS").is_some_and(|v| v == "true");
let stdout = io::stdout().lock();
let summary = stdout.is_terminal();
match run_with(
std::env::args_os().skip(1),
annotate,
summary,
BufWriter::new(stdout),
) {
Ok(code) => ExitCode::from(code),
Err(message) => {
eprintln!("jsonvet: {message}");
ExitCode::from(2)
}
}
}
fn run_with(
args: impl IntoIterator<Item = OsString>,
annotate: bool,
summary: bool,
out: impl Write,
) -> Result<u8, String> {
let args = match args::parse(args).map_err(|e| {
format!(
"{}\n\nRun `jsonvet --help` for usage.",
output::printable(&e)
)
})? {
Command::Help => return print(out, args::HELP),
Command::Version => return print(out, concat!("jsonvet ", env!("CARGO_PKG_VERSION"))),
Command::ListRules => return print(out, &rules::list()),
Command::Explain(text) => return print(out, &text),
Command::Check(args) => args,
};
let resolver = Arc::new(Mutex::new(Resolver::new(&args.config, args.settings)?));
let settings_for = |path: &Path| lock(&resolver).for_file(path);
let threads = std::thread::available_parallelism().map_or(1, std::num::NonZero::get);
let files = expand(args.files, &resolver, threads)?;
let schema = args
.schema
.as_deref()
.map(|path| Schema::load(path, &*settings_for(path)?))
.transpose()
.map_err(|e| output::printable(&e))?;
let stdin_name = args.stdin_filename.unwrap_or_else(|| STDIN.into());
let shown = |file: &OsStr| -> OsString {
if file == STDIN {
stdin_name.clone()
} else {
file.to_owned()
}
};
let files = files
.into_iter()
.map(|file| Ok((settings_for(Path::new(&shown(&file)))?, file)))
.collect::<Result<Vec<_>, String>>()?;
let mut printer = Printer::new(out, args.format, annotate);
let mut counts = Counts::default();
let mut rest = &files[..];
let shares = match schema {
Some(_) => parallel(&files, threads, || false, |(_, file)| schema_share(file)),
None => Vec::new(),
};
let written = loop {
if rest.is_empty() {
let line = (summary && !files.is_empty()).then(|| counts.summary(files.len()));
break line
.map_or(Ok(()), |line| printer.text_line(&line))
.and_then(|()| printer.finish());
}
let offered = match shares.get(files.len() - rest.len()..) {
Some(shares) if !shares.is_empty() => &rest[..chunk_len(shares)],
_ => rest,
};
let texts = read_chunk(offered, threads);
let chunk;
(chunk, rest) = rest.split_at(texts.len());
let jobs: Vec<_> = chunk.iter().zip(&texts).collect();
let parts = (threads / jobs.len()).max(1);
let checked = parallel(
&jobs,
threads,
|| false,
|((settings, file), text)| {
let name = shown(file);
let name = name.to_string_lossy();
text.as_deref()
.map(|text| check(&name, text, settings, schema.as_ref(), parts))
},
);
let outcome =
chunk
.iter()
.zip(checked)
.try_for_each(|((_, file), checked)| match checked {
Ok((report, schema_errors)) => print_file(
&shown(file),
&report,
&schema_errors,
&mut printer,
&mut counts,
),
Err(message) => {
counts.errors += 1;
counts.files += 1;
printer.unreadable(&shown(file), message)
}
});
if outcome.is_err() {
break outcome;
}
};
match written {
Err(e) if e.kind() != io::ErrorKind::BrokenPipe => {
return Err(format!("cannot write output: {e}"));
}
_ => {}
}
let too_many_warnings = args.max_warnings.is_some_and(|max| counts.warnings > max);
Ok(u8::from(counts.errors > 0 || too_many_warnings))
}
fn read_chunk<S: Sync>(files: &[(S, OsString)], threads: usize) -> Vec<Result<String, String>> {
let bytes = AtomicUsize::new(0);
let full = || bytes.load(Ordering::Relaxed) >= CHUNK_BYTES;
parallel(files, threads, full, |(_, file)| {
let text = if file == STDIN {
read_stdin()
} else {
read(Path::new(file))
};
bytes.fetch_add(text.as_ref().map_or(0, String::len), Ordering::Relaxed);
text
})
}
fn schema_share(file: &OsStr) -> usize {
if file == STDIN {
return CHUNK_BYTES;
}
let size = std::fs::metadata(file).map_or(0, |meta| meta.len());
usize::try_from(size)
.unwrap_or(usize::MAX)
.saturating_mul(SCHEMA_BYTES)
.min(CHUNK_BYTES)
}
fn chunk_len(shares: &[usize]) -> usize {
let mut total = 0;
shares
.iter()
.position(|share| {
total += share;
total >= CHUNK_BYTES
})
.map_or(shares.len(), |last| last + 1)
}
fn expand(
args: Vec<OsString>,
resolver: &Arc<Mutex<Resolver>>,
threads: usize,
) -> Result<Vec<OsString>, String> {
let mut files = Vec::with_capacity(args.len());
let dirs = parallel(
&args,
threads,
|| false,
|arg| arg != STDIN && Path::new(arg).is_dir(),
);
for (arg, dir) in args.into_iter().zip(dirs) {
if !dir {
files.push(arg);
continue;
}
let start = files.len();
let above = std::fs::canonicalize(&arg).ok();
if let Some(problem) = std::iter::once(Path::new(&arg))
.chain(above.iter().flat_map(|dir| dir.ancestors()))
.find_map(unsafe_ignore_file)
{
return Err(problem);
}
let skipped = Arc::new(Mutex::new(None));
let skip = Arc::clone(&skipped);
let resolver = Arc::clone(resolver);
let walk = ignore::WalkBuilder::new(&arg)
.filter_entry(move |entry| {
if entry.file_type().is_some_and(|t| t.is_dir())
&& let Some(problem) = unsafe_ignore_file(entry.path())
{
skip.lock()
.unwrap_or_else(std::sync::PoisonError::into_inner)
.get_or_insert(problem);
return false;
}
let settings = lock(&resolver).for_file(entry.path());
!settings.is_ok_and(|settings| settings.excluded(entry.path()))
})
.build();
for entry in walk {
let entry = entry.map_err(|e| output::printable(&format!("cannot walk: {e}")))?;
if entry.file_type().is_some_and(|t| t.is_file())
&& Dialect::for_path(entry.path()).is_some()
{
files.push(entry.into_path().into_os_string());
}
}
if let Some(problem) = skipped
.lock()
.unwrap_or_else(std::sync::PoisonError::into_inner)
.take()
{
return Err(problem);
}
files[start..].sort_unstable();
}
Ok(files)
}
fn lock(resolver: &Mutex<Resolver>) -> std::sync::MutexGuard<'_, Resolver> {
resolver
.lock()
.unwrap_or_else(std::sync::PoisonError::into_inner)
}
#[derive(Default)]
struct Counts {
errors: usize,
warnings: usize,
files: usize,
}
impl Counts {
fn summary(&self, checked: usize) -> String {
let n = |count: usize, what: &str| {
format!("{count} {what}{}", if count == 1 { "" } else { "s" })
};
if self.errors + self.warnings == 0 {
return format!("no problems in {}", n(checked, "file"));
}
let problems = match (self.errors, self.warnings) {
(e, 0) => n(e, "error"),
(0, w) => n(w, "warning"),
(e, w) => format!("{} and {}", n(e, "error"), n(w, "warning")),
};
format!("{problems} in {} of {}", self.files, n(checked, "file"))
}
}
fn parallel<T: Sync, U: Send>(
items: &[T],
threads: usize,
stop: impl Fn() -> bool + Sync,
f: impl Fn(&T) -> U + Sync,
) -> Vec<U> {
let next = AtomicUsize::new(0);
let work = || {
let mut done = Vec::new();
loop {
if stop() && next.load(Ordering::Relaxed) > 0 {
break done;
}
let i = next.fetch_add(1, Ordering::Relaxed);
let Some(item) = items.get(i) else { break done };
done.push((i, f(item)));
}
};
let mut done = std::thread::scope(|scope| {
let helpers: Vec<_> = (1..threads.min(items.len()))
.filter_map(|_| {
std::thread::Builder::new()
.stack_size(STACK)
.spawn_scoped(scope, work)
.ok()
})
.collect();
let mut done = work();
for helper in helpers {
done.extend(
helper
.join()
.unwrap_or_else(|panic| std::panic::resume_unwind(panic)),
);
}
done
});
done.sort_unstable_by_key(|&(i, _)| i);
done.into_iter().map(|(_, result)| result).collect()
}
fn split(text: &str, dialect: Dialect, parts: usize) -> Vec<std::ops::Range<usize>> {
let split = dialect == Dialect::JsonLines && parts > 1 && text.len() >= SPLIT_BYTES;
jsonvet::line_parts(text, if split { parts } else { 1 })
}
fn check<'t>(
name: &str,
text: &'t str,
settings: &Settings,
schema: Option<&Schema>,
parts: usize,
) -> (Report<'t>, Vec<(usize, String, bool)>) {
let config = settings.config_for(name);
let ranges = split(text, config.dialect(), parts);
if ranges.len() > 1 {
let config = match schema {
Some(_) => config.max_depth(SCHEMA_MAX_DEPTH),
None => config,
};
let lines = parallel(
&ranges,
parts,
|| false,
|lines| jsonvet::parse_lines(text, lines.clone(), &config, schema.is_some()),
);
let (report, events) = jsonvet::join_lines(text, lines);
let errors = schema
.zip(events)
.map_or_else(Vec::new, |(schema, events)| schema.errors(&events));
return (report, errors);
}
match schema {
Some(schema) => {
let (report, events) = jsonvet::parse(text, &config.max_depth(SCHEMA_MAX_DEPTH));
let errors = events.map_or_else(Vec::new, |events| schema.errors(&events));
(report, errors)
}
None => (jsonvet::check(text, &config), Vec::new()),
}
}
fn print_file(
file: &OsStr,
report: &Report<'_>,
schema_errors: &[(usize, String, bool)],
printer: &mut Printer<impl Write>,
counts: &mut Counts,
) -> io::Result<()> {
let before = counts.errors + counts.warnings;
for problem in report.problems() {
match problem.severity() {
Severity::Error => counts.errors += 1,
Severity::Warning => counts.warnings += 1,
_ => {}
}
printer.problem(file, report, problem)?;
}
for (offset, message, summary) in schema_errors {
counts.errors += usize::from(!summary);
printer.schema_error(file, report, *offset, message, *summary)?;
}
counts.files += usize::from(counts.errors + counts.warnings > before);
Ok(())
}
fn read(path: &Path) -> Result<String, String> {
let read = std::fs::metadata(path).and_then(|meta| {
if meta.is_file() {
std::fs::read(path).map(Some)
} else {
Ok(None)
}
});
match read {
Ok(Some(bytes)) => decode(bytes),
Ok(None) => Err("not a regular file".to_owned()),
Err(e) if e.kind() == io::ErrorKind::NotFound => Err("no such file".to_owned()),
Err(e) => Err(format!("cannot read: {e}")),
}
}
fn read_stdin() -> Result<String, String> {
let mut bytes = Vec::new();
match io::stdin().lock().read_to_end(&mut bytes) {
Ok(_) => decode(bytes),
Err(e) => Err(format!("cannot read: {e}")),
}
}
fn decode(bytes: Vec<u8>) -> Result<String, String> {
let opens_string = |b: u8| b == b'"' || b == b'\'';
let encoding = match *bytes.as_slice() {
[0xFF, 0xFE, 0, 0, ..] => Some("UTF-32 (LE)"),
[0, 0, 0xFE, 0xFF, ..] => Some("UTF-32 (BE)"),
[0xFF, 0xFE, ..] => Some("UTF-16 (LE)"),
[0xFE, 0xFF, ..] => Some("UTF-16 (BE)"),
[0, 0, 0, a, ..] if a != 0 => Some("UTF-32 (BE)"),
[0, a, 0, b, ..] if a != 0 && b != 0 => Some("UTF-16 (BE)"),
[a, 0, 0, 0, ..] if a != 0 && !opens_string(a) => Some("UTF-32 (LE)"),
[a, 0, b, 0, ..] if a != 0 && b != 0 && !opens_string(a) => Some("UTF-16 (LE)"),
_ => None,
};
if let Some(encoding) = encoding {
return Err(format!("file is {encoding}; jsonvet reads UTF-8"));
}
String::from_utf8(bytes)
.map_err(|e| format!("not valid UTF-8 (byte {})", e.utf8_error().valid_up_to()))
}
pub(crate) const MAX_CONFIG_BYTES: u64 = 16 << 20;
pub(crate) const MAX_SCHEMA_BYTES: u64 = 64 << 20;
pub(crate) fn read_small(path: &Path, max: u64) -> io::Result<String> {
if !std::fs::metadata(path)?.is_file() {
return Err(io::Error::other("not a regular file"));
}
let mut text = String::new();
std::fs::File::open(path)?
.take(max + 1)
.read_to_string(&mut text)?;
if text.len() as u64 > max {
return Err(io::Error::other(format!("larger than {} MB", max >> 20)));
}
Ok(text)
}
fn unsafe_ignore_file(dir: &Path) -> Option<String> {
[".gitignore", ".ignore"].into_iter().find_map(|name| {
let path = dir.join(name);
let meta = std::fs::metadata(&path).ok()?;
let problem = if !meta.is_file() {
"not a regular file".to_owned()
} else if meta.len() > MAX_CONFIG_BYTES {
format!("larger than {} MB", MAX_CONFIG_BYTES >> 20)
} else {
return None;
};
Some(output::printable(&format!(
"cannot walk: {}: {problem}",
path.display()
)))
})
}
fn print(mut out: impl Write, text: &str) -> Result<u8, String> {
match writeln!(out, "{text}").and_then(|()| out.flush()) {
Err(e) if e.kind() != io::ErrorKind::BrokenPipe => Err(format!("cannot write output: {e}")),
_ => Ok(0),
}
}
#[cfg(test)]
mod tests {
use std::fmt::Write as _;
use std::sync::atomic::{AtomicUsize, Ordering};
use jsonvet::Dialect;
use super::{
CHUNK_BYTES, MAX_CONFIG_BYTES, SCHEMA_BYTES, SPLIT_BYTES, Schema, Settings, check,
chunk_len, parallel, read_small, schema_share, split, unsafe_ignore_file,
};
#[test]
fn small_files_are_read_up_to_their_limit_and_no_further() {
let dir = std::env::temp_dir().join(format!("jsonvet-small-{}", std::process::id()));
std::fs::create_dir_all(&dir).expect("a temporary directory");
let path = dir.join("f");
std::fs::write(&path, "0123456789").expect("written");
assert_eq!(read_small(&path, 10).expect("ten bytes"), "0123456789");
assert!(read_small(&path, 9).is_err());
let sized = |len: u64| {
std::fs::File::create(dir.join(".gitignore"))
.and_then(|file| file.set_len(len))
.expect("a large file");
};
sized(MAX_CONFIG_BYTES);
assert_eq!(unsafe_ignore_file(&dir), None);
sized(MAX_CONFIG_BYTES + 1);
assert!(unsafe_ignore_file(&dir).is_some_and(|e| e.ends_with("larger than 16 MB")));
let _ = std::fs::remove_dir_all(&dir);
}
#[test]
fn only_json_lines_of_the_split_size_is_split_one_part_per_thread() {
let line = "[1]\n".repeat(SPLIT_BYTES / 4);
assert_eq!(split(&line, Dialect::JsonLines, 8).len(), 8);
assert_eq!(split(&line, Dialect::JsonLines, 1).len(), 1);
assert_eq!(split(&line, Dialect::Json, 8).len(), 1);
assert_eq!(split(&line[4..], Dialect::JsonLines, 8).len(), 1);
}
#[test]
fn a_large_json_lines_file_checked_in_parts_reports_what_one_pass_does() {
let pad = "x".repeat(500);
let mut text = String::from("\u{FEFF}");
for i in 0..4_000 {
let _ = match i % 7 {
0 => writeln!(text, "{{\"a\": {i}, \"a\": 1}}"),
1 => write!(text, "{{\"a\": \"{i}\", \"p\": \"{pad}\"}}\r\n"),
_ => writeln!(text, "{{\"a\": {i}, \"p\": \"{pad}\"}}"),
};
}
assert!(text.len() > SPLIT_BYTES);
let deep = format!("{}{}\n", "[".repeat(600), "]".repeat(600));
let texts = [
text.clone(),
format!("{text}{{\"a\": 1\n{text}"),
format!("{text}{deep}{text}"),
];
let dir = std::env::temp_dir().join(format!("jsonvet-split-{}", std::process::id()));
std::fs::create_dir_all(&dir).expect("a temporary directory");
let path = dir.join("schema.json");
std::fs::write(&path, r#"{"properties": {"a": {"type": "integer"}}}"#).expect("written");
let settings = Settings::default();
let schema = Schema::load(&path, &settings).expect("a schema");
let _ = std::fs::remove_dir_all(&dir);
for text in &texts {
for schema in [None, Some(&schema)] {
let (whole, whole_errors) = check("x.jsonl", text, &settings, schema, 1);
let (split, split_errors) = check("x.jsonl", text, &settings, schema, 8);
assert_ne!(whole.problems(), &[]);
assert_eq!(split.problems(), whole.problems());
assert_eq!(split_errors, whole_errors);
}
}
let (_, errors) = check("x.jsonl", &texts[0], &settings, Some(&schema), 8);
assert_eq!(errors.len(), 572);
}
#[test]
fn parallel_keeps_the_order_and_stops_after_a_prefix() {
let items: Vec<usize> = (0..1000).collect();
for threads in [1, 3, 16] {
let squares = parallel(&items, threads, || false, |i| i * i);
assert_eq!(squares, items.iter().map(|i| i * i).collect::<Vec<_>>());
assert_eq!(parallel(&items, threads, || true, |&i| i), [0]);
let done = AtomicUsize::new(0);
let some = parallel(
&items,
threads,
|| done.load(Ordering::Relaxed) >= 100,
|&i| {
done.fetch_add(1, Ordering::Relaxed);
i
},
);
assert!((100..100 + threads).contains(&some.len()), "{}", some.len());
assert_eq!(some, (0..some.len()).collect::<Vec<_>>());
}
assert_eq!(
parallel(&[] as &[u8], 4, || false, |&b| b),
Vec::<u8>::new()
);
}
#[test]
fn a_chunk_ends_with_the_file_that_fills_it() {
assert_eq!(chunk_len(&[]), 0);
assert_eq!(chunk_len(&[1, 2, 3]), 3);
assert_eq!(chunk_len(&[CHUNK_BYTES, 1]), 1);
assert_eq!(chunk_len(&[CHUNK_BYTES - 1, 1, 1]), 2);
assert_eq!(chunk_len(&[CHUNK_BYTES - 2, 1, 1]), 3);
assert_eq!(chunk_len(&[0, CHUNK_BYTES - 1, 0, 2, 5]), 4);
}
#[test]
fn a_file_counts_its_size_twenty_times_with_a_schema() {
let dir = std::env::temp_dir().join(format!("jsonvet-share-{}", std::process::id()));
std::fs::create_dir_all(&dir).expect("a temp dir");
let path = dir.join("a.json");
std::fs::write(&path, " ".repeat(1000)).expect("write");
let path = path.as_os_str();
assert_eq!(schema_share(path), 1000 * SCHEMA_BYTES);
std::fs::write(path, " ".repeat(CHUNK_BYTES / SCHEMA_BYTES + 1)).expect("write");
assert_eq!(schema_share(path), CHUNK_BYTES, "capped");
assert_eq!(schema_share(dir.join("gone.json").as_os_str()), 0);
assert_eq!(schema_share(std::ffi::OsStr::new("-")), CHUNK_BYTES);
let _ = std::fs::remove_dir_all(&dir);
}
#[test]
fn the_summary_counts_problems_and_files() {
let dir = std::env::temp_dir().join(format!("jsonvet-summary-{}", std::process::id()));
std::fs::create_dir_all(&dir).expect("a temp dir");
let dup = dir.join("dup.json");
let ok = dir.join("ok.json");
std::fs::write(&dup, "{\"a\": 1, \"a\": 2}").expect("write");
std::fs::write(&ok, "{}").expect("write");
let run = |summary: bool, files: &[&std::path::Path]| {
let mut out = Vec::new();
let args = ["--no-config".into()]
.into_iter()
.chain(files.iter().map(|f| f.as_os_str().to_owned()));
let code = super::run_with(args, false, summary, &mut out).expect("a check");
(code, String::from_utf8(out).expect("UTF-8"))
};
let (code, out) = run(true, &[&dup, &ok]);
assert_eq!(code, 1);
assert!(out.ends_with("\n1 error in 1 of 2 files\n"), "{out}");
let gone = dir.join("gone.json");
let (_, out) = run(true, &[&dup, &gone, &dup, &ok]);
assert!(out.ends_with("\n3 errors in 3 of 4 files\n"), "{out}");
assert_eq!(run(true, &[&ok]).1, "no problems in 1 file\n");
assert!(!run(false, &[&dup]).1.contains(" in 1 of 1 file"));
assert_eq!(run(true, &[]).1, "");
for format in ["sarif", "json"] {
let args = ["--no-config", "--format", format]
.map(std::ffi::OsString::from)
.into_iter()
.chain([dup.clone().into_os_string()]);
let mut out = Vec::new();
super::run_with(args, false, true, &mut out).expect("a check");
let out = String::from_utf8(out).expect("UTF-8");
assert!(
serde_json::from_str::<serde_json::Value>(&out).is_ok(),
"{out}"
);
}
let _ = std::fs::remove_dir_all(&dir);
}
#[test]
fn a_panic_on_the_big_stack_carries_on_in_the_caller() {
static RUNS: AtomicUsize = AtomicUsize::new(0);
let panic = std::panic::catch_unwind(|| {
super::on_big_stack(|| -> u8 {
RUNS.fetch_add(1, Ordering::Relaxed);
panic!("on the big stack")
})
})
.expect_err("the panic carries on");
assert_eq!(panic.downcast_ref::<&str>(), Some(&"on the big stack"));
assert_eq!(RUNS.load(Ordering::Relaxed), 1);
}
}