use regex::Regex;
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct Sentence {
pub line: usize,
pub text: String,
pub raw_text: String,
pub end_mark: Option<char>,
pub line_byte_start: usize,
}
pub fn numbered_lines(text: &str) -> Vec<(usize, &str)> {
text.split('\n')
.enumerate()
.map(|(i, line)| (i + 1, line))
.collect()
}
fn heading_prefix(line: &str) -> Option<(&str, usize)> {
let trimmed = line.trim_start();
let level = trimmed.bytes().take_while(|byte| *byte == b'#').count();
let valid_prefix = (1..=6).contains(&level)
&& trimmed
.as_bytes()
.get(level)
.is_none_or(u8::is_ascii_whitespace);
valid_prefix.then_some((trimmed, level))
}
pub fn is_heading(line: &str) -> bool {
heading_prefix(line).is_some()
}
pub fn heading(line: &str) -> Option<(usize, String)> {
let (trimmed, level) = heading_prefix(line)?;
Some((
level,
trimmed[level..]
.trim()
.trim_end_matches('#')
.trim()
.to_owned(),
))
}
pub fn is_list_item(line: &str) -> bool {
let trimmed = line.trim_start();
if ["- ", "* ", "+ "]
.iter()
.any(|prefix| trimmed.starts_with(prefix))
{
return true;
}
let digits = trimmed.bytes().take_while(u8::is_ascii_digit).count();
digits > 0
&& matches!(trimmed.as_bytes().get(digits), Some(b'.' | b')'))
&& trimmed
.as_bytes()
.get(digits + 1)
.is_some_and(u8::is_ascii_whitespace)
}
pub fn mask_html_comments(text: &str) -> String {
let mut output = String::with_capacity(text.len());
let mut rest = text;
let mut in_comment = false;
while !rest.is_empty() {
if in_comment {
if let Some(end) = rest.find("-->") {
for ch in rest[..end + 3].chars() {
if ch == '\n' {
output.push('\n');
} else {
output.push_str(&" ".repeat(ch.len_utf8()));
}
}
rest = &rest[end + 3..];
in_comment = false;
} else {
for ch in rest.chars() {
if ch == '\n' {
output.push('\n');
} else {
output.push_str(&" ".repeat(ch.len_utf8()));
}
}
break;
}
} else if let Some(start) = rest.find("<!--") {
output.push_str(&rest[..start]);
rest = &rest[start..];
in_comment = true;
} else {
output.push_str(rest);
break;
}
}
output
}
#[derive(Clone, Copy, Debug, Default)]
pub struct MaskStats {
pub reference_lines: usize,
pub code_annotation_lines: usize,
}
pub fn is_reference_line(line: &str) -> bool {
let trimmed = line.trim_start();
let Some(rest) = trimmed.strip_prefix('[') else {
return false;
};
let Some(close) = rest.find(']') else {
return false;
};
let label = &rest[..close];
let after = &rest[close + 1..];
let numeric = !label.is_empty() && label.bytes().all(|byte| byte.is_ascii_digit());
let footnote = label.starts_with('^') && label.len() > 1;
(numeric && after.starts_with([' ', '\t', ':'])) || (footnote && after.starts_with(':'))
}
pub fn is_code_annotation_line(line: &str) -> bool {
let bytes = line.trim_start().as_bytes();
bytes.len() >= 3 && bytes[0] == b'#' && bytes[1].is_ascii_uppercase() && bytes[2] == b' '
}
fn is_reference_heading(title: &str) -> bool {
matches!(
title.trim().to_ascii_lowercase().as_str(),
"参考文献" | "引用文献" | "references" | "bibliography"
)
}
pub fn mask_markdown_structure(text: &str) -> String {
mask_markdown(text, false).0
}
pub fn mask_markdown_structure_with_stats(text: &str) -> (String, MaskStats) {
mask_markdown(text, false)
}
pub fn mask_markdown_structure_preserving_headings(text: &str) -> String {
mask_markdown(text, true).0
}
fn mask_markdown(text: &str, preserve_headings: bool) -> (String, MaskStats) {
let text = mask_html_comments(text);
let inline_code = Regex::new(r"``[^\n]*?``|`[^`\n]+`").expect("valid inline-code regex");
let inline_html =
Regex::new(r"</?[A-Za-z][A-Za-z0-9-]*(?:\s[^>\n]*)?/?>").expect("valid HTML-tag regex");
let link_url = Regex::new(r"(\]\()([^)]*)(\))").expect("valid Markdown-link regex");
let embed_citation =
Regex::new(r"^\[https?://[^]\n]+:embed:cite\]$").expect("valid embed-citation regex");
let mut masked = Vec::new();
let mut stats = MaskStats::default();
let mut fence: Option<(char, usize)> = None;
let mut front_matter = false;
let mut reference_section_level = None;
for (index, line) in text.split('\n').enumerate() {
let trimmed = line.trim_start();
if index == 0 && line.trim() == "---" {
front_matter = true;
masked.push(String::new());
continue;
}
if front_matter {
masked.push(String::new());
if line.trim() == "---" {
front_matter = false;
}
continue;
}
let fence_run = trimmed
.chars()
.next()
.filter(|ch| *ch == '`' || *ch == '~')
.map(|ch| {
(
ch,
trimmed
.chars()
.take_while(|candidate| candidate == &ch)
.count(),
)
});
if let Some((open_char, open_len)) = fence {
masked.push(String::new());
if fence_run.is_some_and(|(ch, len)| ch == open_char && len >= open_len) {
fence = None;
}
continue;
}
if let Some((ch, len)) = fence_run.filter(|(_, len)| *len >= 3) {
fence = Some((ch, len));
masked.push(String::new());
continue;
}
let current_heading = heading(line);
if let Some((level, title)) = ¤t_heading {
if is_reference_heading(title) {
reference_section_level = Some(*level);
} else if reference_section_level.is_some_and(|reference| *level <= reference) {
reference_section_level = None;
}
}
if reference_section_level.is_some() && current_heading.is_none() {
if !trimmed.is_empty() {
stats.reference_lines += 1;
}
masked.push(String::new());
continue;
}
let is_table = (trimmed.starts_with('|') && trimmed.matches('|').count() >= 2)
|| (trimmed.contains('|')
&& trimmed
.chars()
.all(|ch| ch.is_whitespace() || matches!(ch, '|' | '-' | ':')));
if (!preserve_headings && is_heading(line))
|| is_list_item(line)
|| trimmed.starts_with('>')
|| is_table
|| embed_citation.is_match(trimmed)
{
masked.push(String::new());
continue;
}
if is_reference_line(line) {
stats.reference_lines += 1;
masked.push(String::new());
continue;
}
if is_code_annotation_line(line) {
stats.code_annotation_lines += 1;
masked.push(String::new());
continue;
}
let no_code = inline_code.replace_all(line, |captures: ®ex::Captures<'_>| {
" ".repeat(captures[0].len())
});
let no_html = inline_html.replace_all(&no_code, |captures: ®ex::Captures<'_>| {
" ".repeat(captures[0].len())
});
let no_urls = link_url.replace_all(&no_html, |captures: ®ex::Captures<'_>| {
format!(
"{}{}{}",
&captures[1],
" ".repeat(captures[2].len()),
&captures[3]
)
});
masked.push(no_urls.into_owned());
}
(masked.join("\n"), stats)
}
pub fn sentences(text: &str) -> Vec<Sentence> {
sentences_with_raw(text, text)
}
pub fn sentences_with_raw(text: &str, raw_text: &str) -> Vec<Sentence> {
let mut output = Vec::new();
let raw_lines = raw_text.split('\n').collect::<Vec<_>>();
for (line_no, line) in numbered_lines(text) {
let raw_line = raw_lines.get(line_no - 1).copied().unwrap_or(line);
let mut start = 0;
for (byte, ch) in line.char_indices() {
if matches!(ch, '。' | '！' | '？' | '!' | '?') {
let end = byte;
let segment = &line[start..end];
let sentence = segment.trim();
if !sentence.is_empty() {
let leading = segment.len() - segment.trim_start().len();
output.push(Sentence {
line: line_no,
text: sentence.to_owned(),
raw_text: raw_line
.get(start..end)
.unwrap_or(sentence)
.trim()
.to_owned(),
end_mark: Some(ch),
line_byte_start: start + leading,
});
}
start = byte + ch.len_utf8();
}
}
let segment = &line[start..];
let tail = segment.trim();
if !tail.is_empty() {
let leading = segment.len() - segment.trim_start().len();
output.push(Sentence {
line: line_no,
text: tail.to_owned(),
raw_text: raw_line.get(start..).unwrap_or(tail).trim().to_owned(),
end_mark: None,
line_byte_start: start + leading,
});
}
}
output
}
pub fn excerpt_around(
line: &str,
byte_start: usize,
byte_len: usize,
context_chars: usize,
) -> String {
let char_start = line[..byte_start].chars().count();
let match_chars = line[byte_start..byte_start + byte_len].chars().count();
let chars = line.chars().collect::<Vec<_>>();
let from = char_start.saturating_sub(context_chars);
let to = (char_start + match_chars + context_chars).min(chars.len());
chars[from..to].iter().collect()
}