Support comments in JSON body (#419)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Gregory Schier
2026-03-05 15:05:09 -08:00
committed by GitHub
parent 242f55b609
commit 267508e533
22 changed files with 954 additions and 161 deletions
+235 -6
View File
@@ -11,6 +11,7 @@ pub fn format_json(text: &str, tab: &str) -> String {
let mut new_json = "".to_string();
let mut depth = 0;
let mut state = FormatState::None;
let mut saw_newline_in_whitespace = false;
loop {
let rest_of_chars = chars.clone();
@@ -61,6 +62,62 @@ pub fn format_json(text: &str, tab: &str) -> String {
continue;
}
// Handle line comments (//)
if current_char == '/' && chars.peek() == Some(&'/') {
chars.next(); // Skip second /
// Collect the rest of the comment until newline
let mut comment = String::from("//");
loop {
match chars.peek() {
Some(&'\n') | None => break,
Some(_) => comment.push(chars.next().unwrap()),
}
}
// Check if the comma handler already added \n + indent
let trimmed = new_json.trim_end_matches(|c: char| c == ' ' || c == '\t');
if trimmed.ends_with(",\n") && !saw_newline_in_whitespace {
// Trailing comment on the same line as comma (e.g. "foo",// comment)
new_json.truncate(trimmed.len() - 1);
new_json.push(' ');
} else if !trimmed.ends_with('\n') && !new_json.is_empty() {
// Trailing comment after a value (no newline before us)
new_json.push(' ');
}
new_json.push_str(&comment);
new_json.push('\n');
new_json.push_str(tab.to_string().repeat(depth).as_str());
saw_newline_in_whitespace = false;
continue;
}
// Handle block comments (/* ... */)
if current_char == '/' && chars.peek() == Some(&'*') {
chars.next(); // Skip *
let mut comment = String::from("/*");
loop {
match chars.next() {
None => break,
Some('*') if chars.peek() == Some(&'/') => {
chars.next(); // Skip /
comment.push_str("*/");
break;
}
Some(c) => comment.push(c),
}
}
// If we're not already on a fresh line, add newline + indent before comment
let trimmed = new_json.trim_end_matches(|c: char| c == ' ' || c == '\t');
if !trimmed.is_empty() && !trimmed.ends_with('\n') {
new_json.push('\n');
new_json.push_str(tab.to_string().repeat(depth).as_str());
}
new_json.push_str(&comment);
// After block comment, add newline + indent for the next content
new_json.push('\n');
new_json.push_str(tab.to_string().repeat(depth).as_str());
continue;
}
match current_char {
',' => {
new_json.push(current_char);
@@ -125,20 +182,37 @@ pub fn format_json(text: &str, tab: &str) -> String {
|| current_char == '\t'
|| current_char == '\r'
{
if current_char == '\n' {
saw_newline_in_whitespace = true;
}
// Don't add these
} else {
saw_newline_in_whitespace = false;
new_json.push(current_char);
}
}
}
}
// Replace only lines containing whitespace with nothing
new_json
.lines()
.filter(|line| !line.trim().is_empty()) // Filter out whitespace-only lines
.collect::<Vec<&str>>() // Collect the non-empty lines into a vector
.join("\n") // Join the lines back into a single string
// Filter out whitespace-only lines, but preserve empty lines inside block comments
let mut result_lines: Vec<&str> = Vec::new();
let mut in_block_comment = false;
for line in new_json.lines() {
if in_block_comment {
result_lines.push(line);
if line.contains("*/") {
in_block_comment = false;
}
} else {
if line.contains("/*") && !line.contains("*/") {
in_block_comment = true;
}
if !line.trim().is_empty() {
result_lines.push(line);
}
}
}
result_lines.iter().map(|line| line.trim_end()).collect::<Vec<&str>>().join("\n")
}
#[cfg(test)]
@@ -297,6 +371,161 @@ mod tests {
r#"
{}
}
"#
.trim()
);
}
#[test]
fn test_line_comment_between_keys() {
assert_eq!(
format_json(
r#"{"foo":"bar",// a comment
"baz":"qux"}"#,
" "
),
r#"
{
"foo": "bar", // a comment
"baz": "qux"
}
"#
.trim()
);
}
#[test]
fn test_line_comment_at_end() {
assert_eq!(
format_json(
r#"{"foo":"bar" // trailing
}"#,
" "
),
r#"
{
"foo": "bar" // trailing
}
"#
.trim()
);
}
#[test]
fn test_block_comment() {
assert_eq!(
format_json(r#"{"foo":"bar",/* comment */"baz":"qux"}"#, " "),
r#"
{
"foo": "bar",
/* comment */
"baz": "qux"
}
"#
.trim()
);
}
#[test]
fn test_comment_in_array() {
assert_eq!(
format_json(
r#"[1,// item comment
2,3]"#,
" "
),
r#"
[
1, // item comment
2,
3
]
"#
.trim()
);
}
#[test]
fn test_comment_only_line() {
assert_eq!(
format_json(
r#"{
// this is a standalone comment
"foo": "bar"
}"#,
" "
),
r#"
{
// this is a standalone comment
"foo": "bar"
}
"#
.trim()
);
}
#[test]
fn test_multiline_block_comment() {
assert_eq!(
format_json(
r#"{
"foo": "bar"
/**
Hello World!
Hi there
*/
}"#,
" "
),
r#"
{
"foo": "bar"
/**
Hello World!
Hi there
*/
}
"#
.trim()
);
}
// NOTE: trailing whitespace on output lines is trimmed by the formatter.
// We can't easily add a test for this because raw string literals get
// trailing whitespace stripped by the editor/linter.
#[test]
fn test_comment_inside_string_ignored() {
assert_eq!(
format_json(r#"{"foo":"// not a comment","bar":"/* also not */"}"#, " "),
r#"
{
"foo": "// not a comment",
"bar": "/* also not */"
}
"#
.trim()
);
}
#[test]
fn test_comment_on_line_after_comma() {
assert_eq!(
format_json(
r#"{
"a": "aaa",
// "b": "bbb"
}"#,
" "
),
r#"
{
"a": "aaa",
// "b": "bbb"
}
"#
.trim()
);
+1
View File
@@ -1,6 +1,7 @@
pub mod error;
pub mod escape;
pub mod format_json;
pub mod strip_json_comments;
pub mod parser;
pub mod renderer;
pub mod wasm;
@@ -0,0 +1,318 @@
/// Strips JSON comments only if the result is valid JSON. If stripping comments
/// produces invalid JSON, the original text is returned unchanged.
pub fn maybe_strip_json_comments(text: &str) -> String {
let stripped = strip_json_comments(text);
if serde_json::from_str::<serde_json::Value>(&stripped).is_ok() {
stripped
} else {
text.to_string()
}
}
/// Strips comments from JSONC, preserving the original formatting as much as possible.
///
/// - Trailing comments on a line are removed (along with preceding whitespace)
/// - Whole-line comments are removed, including the line itself
/// - Block comments are removed, including any lines that become empty
/// - Comments inside strings and template tags are left alone
pub fn strip_json_comments(text: &str) -> String {
let mut chars = text.chars().peekable();
let mut result = String::with_capacity(text.len());
let mut in_string = false;
let mut in_template_tag = false;
loop {
let current_char = match chars.next() {
None => break,
Some(c) => c,
};
// Handle JSON strings
if in_string {
result.push(current_char);
match current_char {
'"' => in_string = false,
'\\' => {
if let Some(c) = chars.next() {
result.push(c);
}
}
_ => {}
}
continue;
}
// Handle template tags
if in_template_tag {
result.push(current_char);
if current_char == ']' && chars.peek() == Some(&'}') {
result.push(chars.next().unwrap());
in_template_tag = false;
}
continue;
}
// Check for template tag start
if current_char == '$' && chars.peek() == Some(&'{') {
let mut lookahead = chars.clone();
lookahead.next(); // skip {
if lookahead.peek() == Some(&'[') {
in_template_tag = true;
result.push(current_char);
result.push(chars.next().unwrap()); // {
result.push(chars.next().unwrap()); // [
continue;
}
}
// Check for line comment
if current_char == '/' && chars.peek() == Some(&'/') {
chars.next(); // skip second /
// Consume until newline
loop {
match chars.peek() {
Some(&'\n') | None => break,
Some(_) => {
chars.next();
}
}
}
// Trim trailing whitespace that preceded the comment
let trimmed_len = result.trim_end_matches(|c: char| c == ' ' || c == '\t').len();
result.truncate(trimmed_len);
continue;
}
// Check for block comment
if current_char == '/' && chars.peek() == Some(&'*') {
chars.next(); // skip *
// Consume until */
loop {
match chars.next() {
None => break,
Some('*') if chars.peek() == Some(&'/') => {
chars.next(); // skip /
break;
}
Some(_) => {}
}
}
// Trim trailing whitespace that preceded the comment
let trimmed_len = result.trim_end_matches(|c: char| c == ' ' || c == '\t').len();
result.truncate(trimmed_len);
// Skip whitespace/newline after the block comment if the next line is content
// (this handles the case where the block comment is on its own line)
continue;
}
if current_char == '"' {
in_string = true;
}
result.push(current_char);
}
// Remove lines that are now empty (were comment-only lines)
let result = result
.lines()
.filter(|line| !line.trim().is_empty())
.collect::<Vec<&str>>()
.join("\n");
// Remove trailing commas before } or ]
strip_trailing_commas(&result)
}
/// Removes trailing commas before closing braces/brackets, respecting strings.
fn strip_trailing_commas(text: &str) -> String {
let mut result = String::with_capacity(text.len());
let chars: Vec<char> = text.chars().collect();
let mut i = 0;
let mut in_string = false;
while i < chars.len() {
let ch = chars[i];
if in_string {
result.push(ch);
match ch {
'"' => in_string = false,
'\\' => {
i += 1;
if i < chars.len() {
result.push(chars[i]);
}
}
_ => {}
}
i += 1;
continue;
}
if ch == '"' {
in_string = true;
result.push(ch);
i += 1;
continue;
}
if ch == ',' {
// Look ahead past whitespace/newlines for } or ]
let mut j = i + 1;
while j < chars.len() && chars[j].is_whitespace() {
j += 1;
}
if j < chars.len() && (chars[j] == '}' || chars[j] == ']') {
// Skip the comma
i += 1;
continue;
}
}
result.push(ch);
i += 1;
}
result
}
#[cfg(test)]
mod tests {
use crate::strip_json_comments::strip_json_comments;
#[test]
fn test_no_comments() {
let input = r#"{
"foo": "bar",
"baz": 123
}"#;
assert_eq!(strip_json_comments(input), input);
}
#[test]
fn test_trailing_line_comment() {
assert_eq!(
strip_json_comments(r#"{
"foo": "bar", // this is a comment
"baz": 123
}"#),
r#"{
"foo": "bar",
"baz": 123
}"#
);
}
#[test]
fn test_whole_line_comment() {
assert_eq!(
strip_json_comments(r#"{
// this is a comment
"foo": "bar"
}"#),
r#"{
"foo": "bar"
}"#
);
}
#[test]
fn test_inline_block_comment() {
assert_eq!(
strip_json_comments(r#"{
"foo": /* a comment */ "bar"
}"#),
r#"{
"foo": "bar"
}"#
);
}
#[test]
fn test_whole_line_block_comment() {
assert_eq!(
strip_json_comments(r#"{
/* a comment */
"foo": "bar"
}"#),
r#"{
"foo": "bar"
}"#
);
}
#[test]
fn test_multiline_block_comment() {
assert_eq!(
strip_json_comments(r#"{
/**
* Hello World!
*/
"foo": "bar"
}"#),
r#"{
"foo": "bar"
}"#
);
}
#[test]
fn test_comment_inside_string_preserved() {
let input = r#"{
"foo": "// not a comment",
"bar": "/* also not */"
}"#;
assert_eq!(strip_json_comments(input), input);
}
#[test]
fn test_comment_inside_template_tag_preserved() {
let input = r#"{
"foo": ${[ fn("// hi", "/* hey */") ]}
}"#;
assert_eq!(strip_json_comments(input), input);
}
#[test]
fn test_multiple_comments() {
assert_eq!(
strip_json_comments(r#"{
// first comment
"foo": "bar", // trailing
/* block */
"baz": 123
}"#),
r#"{
"foo": "bar",
"baz": 123
}"#
);
}
#[test]
fn test_trailing_comma_after_comment_removed() {
assert_eq!(
strip_json_comments(r#"{
"a": "aaa",
// "b": "bbb"
}"#),
r#"{
"a": "aaa"
}"#
);
}
#[test]
fn test_trailing_comma_in_array() {
assert_eq!(
strip_json_comments(r#"[1, 2, /* 3 */]"#),
r#"[1, 2]"#
);
}
#[test]
fn test_comma_inside_string_preserved() {
let input = r#"{"a": "hello,}"#;
assert_eq!(strip_json_comments(input), input);
}
}