1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316
//! A simple and reasonably fast Rust implementation of the
//! [Segmentation Rules eXchange 2.0 standard](https://www.unicode.org/uli/pas/srx/srx20.html)
//! for text segmentation. `srx` is *not* fully compliant with the standard.
//!
//! This crate is intended for segmentation of plaintext so markup information (`<formathandle>` and `segmentsubflows`)
//! is ignored.
//!
//! Not complying with the SRX spec, overlapping matches of the same `<rule>` are not found which could
//! lead to different behavior in a few edge cases.
//!
//! ## Example
//!
//! ```
//! use std::{fs, str::FromStr};
//! use srx::SRX;
//!
//! let srx = SRX::from_str(&fs::read_to_string("data/segment.srx").unwrap())?;
//! let english_rules = srx.language_rules("en");
//!
//! assert_eq!(
//! english_rules.split("e.g. U.K. and Mr. do not split. SRX is a rule-based format.").collect::<Vec<_>>(),
//! vec!["e.g. U.K. and Mr. do not split. ", "SRX is a rule-based format."]
//! );
//! # Ok::<(), srx::Error>(())
//! ```
//!
//! ## Features
//!
//! - `serde`: Serde serialization and deserialization support for [SRX].
//! - `from_xml`: [SRX::from_reader] method and [std::str::FromStr] implementation to load from an XML file in SRX format.
//!
//! ## A note on regular expressions
//!
//! This crate uses the [`regex` crate](https://github.com/rust-lang/regex) for parsing and executing
//! regular expressions. The `regex` crate is mostly compatible with the
//! [regular expression standard](https://www.unicode.org/uli/pas/srx/srx20.html#Intro_RegExp) from the SRX specification.
//! However, some metacharacters such as `\Q` and `\E` are not supported.
//!
//! To still be able to use files containing unsupported rules and to parse useful SRX files
//! such as
//! [`segment.srx` from LanguageTool](https://github.com/languagetool-org/languagetool/blob/master/languagetool-core/src/main/resources/org/languagetool/resource/segment.srx)
//! which does not comply with the standard by e. g. using look-ahead and look-behind, `srx`
//! ignores `<rule>` elements with invalid regular expressions and provides information about
//! them via the [SRX::errors] function.
#![cfg_attr(docsrs, feature(doc_cfg))] // see https://stackoverflow.com/a/61417700
#[cfg(feature = "serde")]
extern crate serde_crate as serde;
#[cfg(feature = "serde")]
use serde::{Deserialize, Serialize};
use std::{collections::HashMap, ops::Range};
use regex::Regex;
#[cfg(feature = "from_xml")]
mod from_xml;
#[cfg(feature = "from_xml")]
mod utils;
#[cfg(feature = "from_xml")]
pub use from_xml::Error;
/// Newtype denoting a language (`languagerulename` attribute in SRX).
#[cfg_attr(
feature = "serde",
derive(Serialize, Deserialize),
serde(crate = "serde_crate")
)]
#[derive(Debug, Clone, Eq, PartialEq, Hash, Ord, PartialOrd)]
pub struct Language(pub String);
/// A single SRX rule. In SRX, consists of one `before_break` and one `after_break` Regex.
/// For efficiency this crate compiles these regexes into one regex of the form `before_break(after_break)`
/// and uses the start of the first capture group as the split index.
#[cfg_attr(
feature = "serde",
derive(Serialize, Deserialize),
serde(crate = "serde_crate")
)]
#[derive(Debug, Clone)]
#[non_exhaustive]
struct Rule {
#[cfg_attr(feature = "serde", serde(with = "serde_regex"))]
regex: Regex,
do_break: bool,
}
impl Rule {
/// Gets all byte indices in the text at which this rule matches.
/// Contrary to the SRX 2.0 spec this does not find overlapping matches.
fn match_indices<'a>(&'a self, text: &'a str) -> impl Iterator<Item = usize> + 'a {
self.regex.captures_iter(text).filter_map(|x| {
// generally it is guaranteed that a regex has
// at least one match, but be lenient about
// errors in the srx xml files and drop those without
x.get(1).map(|x| x.start())
})
}
/// Whether this rule breaks or prevents breaking.
fn do_break(&self) -> bool {
self.do_break
}
}
/// An ordered set of rules.
/// Rules are executed in order.
/// Once a rule matches on an index, no other rule can match at the same index.
/// Each rule either breaks (i. e. splits the text at this index) or prevents breaking.
#[cfg_attr(
feature = "serde",
derive(Serialize, Deserialize),
serde(crate = "serde_crate")
)]
#[derive(Debug, Clone, Default)]
pub struct Rules {
rules: Vec<Rule>,
}
impl Rules {
/// Obtain the ranges for text segments. Guaranteed to be at character bounds.
pub fn split_ranges(&self, text: &str) -> Vec<Range<usize>> {
let mut segments = Vec::new();
// TODO use a proper tri-state enum here
let mut masked_bytes: Vec<Option<bool>> = vec![None; text.len()];
'outer: for rule in &self.rules {
for byte_index in rule.match_indices(text) {
if byte_index >= text.len() {
continue 'outer;
}
if masked_bytes[byte_index].is_none() {
masked_bytes[byte_index] = Some(rule.do_break());
}
}
}
let mut prev_byte_pos = 0;
// Iterate over characters, we don't want no half characters in the output ranges
for (byte_pos, _c) in text.char_indices() {
if let Some(true) = masked_bytes[byte_pos] {
segments.push(prev_byte_pos..byte_pos);
prev_byte_pos = byte_pos;
}
}
// Deal with the trailing element, which is by definition
// not required to be suffixed by a gap char.
if text[prev_byte_pos..].chars().next().is_some() {
segments.push(prev_byte_pos..text.len());
}
segments
}
/// Split text into segments.
pub fn split<'a, 'b>(&self, text: &'a str) -> impl Iterator<Item = &'a str> + 'b
where
'a: 'b,
{
self.split_ranges(text)
.into_iter()
.map(move |range| &text[range])
}
pub fn is_empty(&self) -> bool {
self.rules.is_empty()
}
pub fn len(&self) -> usize {
self.rules.len()
}
}
/// An entry of the `<maprules>` element.
/// Associates a regex with a [Language].
#[cfg_attr(
feature = "serde",
derive(Serialize, Deserialize),
serde(crate = "serde_crate")
)]
#[derive(Debug, Clone)]
struct LanguageRegex {
#[cfg_attr(feature = "serde", serde(with = "serde_regex"))]
regex: Regex,
language: Language,
}
/// The SRX root.
/// Does not execute rules on is own.
#[cfg_attr(
feature = "serde",
derive(Serialize, Deserialize),
serde(crate = "serde_crate")
)]
#[derive(Debug, Clone)]
pub struct SRX {
cascade: bool,
map: Vec<LanguageRegex>,
rules: HashMap<Language, Vec<Rule>>,
errors: HashMap<Language, Vec<String>>,
}
impl SRX {
/// Gets the rules for a language code by
/// - aggregating rules from all [Language]s with a matching `<languagepattern>` (if the SRX is set to be cascading)
/// - finding the first matching `<languagepattern>` (if the SRX is set to be not cascading)
///
/// Result should be cached instead of calling this repeatedly as it clones the rules.
pub fn language_rules<S: AsRef<str>>(&self, lang_code: S) -> Rules {
let mut rules = Vec::new();
for item in &self.map {
if item.regex.is_match(lang_code.as_ref()) {
rules.extend(self.rules.get(&item.language).expect("languagerulename in <languagemap> must have a corresponding entry in <languagerules>").iter().cloned());
if !self.cascade {
break;
}
}
}
Rules { rules }
}
/// Maps [Language]s to a vector of string representations of errors which occured during parsing regular expressions for this language.
pub fn errors(&self) -> &HashMap<Language, Vec<String>> {
&self.errors
}
}
#[cfg(test)]
mod tests {
use super::*;
use quickcheck_macros::quickcheck;
use std::{fs, str::FromStr};
#[quickcheck]
fn length_invariant(text: String) {
let rules =
SRX::from_str(&fs::read_to_string("data/example.srx").expect("example file exists"))
.expect("example file is valid")
.language_rules("en");
assert_eq!(
text.len(),
rules.split(&text).fold(0, |acc, x| acc + x.len())
);
}
#[test]
fn match_indices_correct() {
let rule = Rule::new(Some("abc"), Some("d+fg"), true).expect("test rule is valid");
assert_eq!(
rule.match_indices("abcddfgxxx").collect::<Vec<_>>(),
vec![3_usize]
);
}
#[test]
fn example_splits_correct() {
let rules =
SRX::from_str(&fs::read_to_string("data/example.srx").expect("example file exists"))
.expect("example file is valid")
.language_rules("en");
// example from the spec: https://www.unicode.org/uli/pas/srx/srx20.html#AppExample
let text =
"The U.K. Prime Minister, Mr. Blair, was seen out with his family today. He is well.";
assert_eq!(
rules.split(text).collect::<Vec<_>>(),
vec![
"The U.K. Prime Minister, Mr. Blair, was seen out with his family today.",
" He is well."
]
);
}
#[test]
fn example_splits_correct_multi_emoji() {
let rules =
SRX::from_str(&fs::read_to_string("data/segment.srx").expect("example file exists"))
.expect("example file is valid")
.language_rules("en");
let text = "e.g. U.K. and Mr. do not split. SRX is a 👒🍏🍱-based format 🐱";
assert_eq!(
rules.split(text).collect::<Vec<_>>(),
vec![
"e.g. U.K. and Mr. do not split. ",
"SRX is a 👒🍏🍱-based format 🐱"
]
);
}
#[test]
fn ignores_last_match_index() {
let rules =
SRX::from_str(&fs::read_to_string("data/segment.srx").expect("example file exists"))
.expect("example file is valid")
.language_rules("en");
let _ = rules.split("Hello! ").collect::<Vec<_>>();
}
#[test]
fn errors_reported() {
let srx =
SRX::from_str(&fs::read_to_string("data/segment.srx").expect("segment file exists"))
.expect("segment file is valid");
assert!(!srx.errors().is_empty());
assert_eq!(srx.errors().values().flatten().count(), 49);
}
}