use crate::colors::COLOR_HAIR;
use crate::pipeline::{splice_created, Optimizer, ZHT_SYNONYM_OPTIMIZER};
use crate::postag::POSTAG;
use crate::segment::Segment;
use crate::text::char_len;
use crate::word::Word;
use once_cell::sync::Lazy;
use regex::Regex;
const CLOSE_P: &[&str] = &["】", "」", "》", "』", "］", "’", "”", "〉"];
const SEP_P: &[&str] = &["、", ",", "…"];
fn hex_and_any(p: u32, flags: &[u32]) -> bool {
flags.iter().any(|f| p & f != 0)
}
fn get_synonym(seg: &Segment, w: &str, mut nw: String) -> String {
if let Some(s) = seg.synonym.get(w) {
nw = s.to_string();
}
if let Some(s) = seg.synonym.get(&nw) {
nw = s.to_string();
}
nw
}
fn replace_li(s: &str) -> String {
static RE: Lazy<Regex> = Lazy::new(|| Regex::new(r"(.)里|里(.)").unwrap());
RE.replace_all(s, |caps: ®ex::Captures| {
let a = caps.get(1).map(|m| m.as_str()).unwrap_or("");
let b = caps.get(2).map(|m| m.as_str()).unwrap_or("");
format!("{a}裡{b}")
})
.into_owned()
}
fn replace_hou(s: &str) -> String {
static RE: Lazy<Regex> = Lazy::new(|| Regex::new(r"(.)后|后(.)").unwrap());
RE.replace_all(s, |caps: ®ex::Captures| {
let a = caps.get(1).map(|m| m.as_str()).unwrap_or("");
let b = caps.get(2).map(|m| m.as_str()).unwrap_or("");
format!("{a}後{b}")
})
.into_owned()
}
fn replace_shen(s: &str) -> String {
static RE: Lazy<Regex> = Lazy::new(|| Regex::new(r"蔘(.)").unwrap());
RE.replace_all(s, "參$1").into_owned()
}
pub struct ZhtSynonymOptimizer;
impl Optimizer for ZhtSynonymOptimizer {
fn name(&self) -> &'static str {
ZHT_SYNONYM_OPTIMIZER
}
fn do_optimize(&self, mut words: Vec<Word>, seg: &Segment) -> Vec<Word> {
let mut i = 0;
while i < words.len() {
if seg.blacklist_synonym.contains(&words[i].w) {
i += 1;
continue;
}
let w0 = if i > 0 { Some(words[i - 1].clone()) } else { None };
let w2 = words.get(i + 1).cloned();
let w1_len = char_len(&words[i].w);
let mut bool_ok = false;
let mut new_p: Option<u32> = None;
if w1_len == 1 {
let w = words[i].w.clone();
if w == "里" {
if w0.as_ref().map(|p| p.w.ends_with('的') || p.w == "和").unwrap_or(false) {
} else if w0.as_ref().map(|p| CLOSE_P.contains(&p.w.as_str())).unwrap_or(false)
|| w0.as_ref().map(|p| {
hex_and_any(p.pos(), &[POSTAG::D_N, POSTAG::D_S, POSTAG::D_F, POSTAG::D_T, POSTAG::D_V])
}).unwrap_or(false)
{
words[i].ow = Some(std::mem::replace(&mut words[i].w, "裡".into()));
bool_ok = true;
}
} else if w == "后" {
if w0.as_ref().map(|p| p.w == "和").unwrap_or(false) {
} else if w0.as_ref().map(|p| CLOSE_P.contains(&p.w.as_str())).unwrap_or(false)
|| w0.as_ref().map(|p| p.w == "腰").unwrap_or(false)
|| w0.as_ref().map(|p| {
p.p.is_some()
&& p.pos() != 0
&& hex_and_any(
p.pos(),
&[
POSTAG::D_V,
POSTAG::D_S,
POSTAG::D_T,
POSTAG::D_N,
POSTAG::D_MQ,
POSTAG::A_M,
POSTAG::D_F,
POSTAG::D_D,
POSTAG::D_R,
],
)
}).unwrap_or(false)
|| w2.as_ref().map(|n| n.p.is_some() && hex_and_any(n.pos(), &[POSTAG::D_V])).unwrap_or(false)
|| (w2.is_some()
&& w0.is_some()
&& w0.as_ref().unwrap().pos_falsy()
&& w2.as_ref().map(|n| n.p.is_some() && hex_and_any(n.pos(), &[POSTAG::D_D])).unwrap_or(false))
|| (w2.is_some()
&& w0.as_ref().map(|p| p.pos_falsy()).unwrap_or(true)
&& w2.as_ref().map(|n| SEP_P.contains(&n.w.as_str())).unwrap_or(false))
{
words[i].ow = Some(std::mem::replace(&mut words[i].w, "後".into()));
bool_ok = true;
}
} else if w == "发" || w == "發" {
if let Some(prev) = &w0 {
if COLOR_HAIR.contains(&prev.w) {
let nw = get_synonym(seg, &w, "髮".into());
if nw != w {
words[i].ow = Some(std::mem::replace(&mut words[i].w, nw));
new_p = Some(POSTAG::D_N);
bool_ok = true;
}
}
}
if !bool_ok && words[i].w == "发" && w2.as_ref().map(|n| n.w == "的").unwrap_or(false) {
words[i].ow = Some(std::mem::replace(&mut words[i].w, "發".into()));
bool_ok = true;
}
if !bool_ok
&& words[i].w == "发"
&& w0.as_ref().map(|p| p.pos() & POSTAG::D_R != 0).unwrap_or(false)
&& w2.as_ref().map(|n| n.pos() & POSTAG::D_R != 0).unwrap_or(false)
{
words[i].ow = Some(std::mem::replace(&mut words[i].w, "發".into()));
bool_ok = true;
}
if !bool_ok
&& words[i].w == "发"
&& w2.as_ref().map(|n| n.w == "那麼" || n.w == "那么").unwrap_or(false)
{
words[i].ow = Some(std::mem::replace(&mut words[i].w, "發".into()));
bool_ok = true;
}
} else if w == "于" {
let start_ok = w0.as_ref().map(|p| p.pos() & POSTAG::D_W != 0).unwrap_or(true);
if start_ok
&& w2.as_ref().map(|n| {
n.p.is_some()
&& n.pos() != 0
&& hex_and_any(
n.pos(),
&[
POSTAG::D_N,
POSTAG::D_V,
POSTAG::D_R,
POSTAG::D_D,
POSTAG::D_T,
POSTAG::A_NR,
POSTAG::D_S,
POSTAG::D_F,
],
)
}).unwrap_or(false)
{
words[i].ow = Some(std::mem::replace(&mut words[i].w, "於".into()));
new_p = Some(POSTAG::D_P);
words[i].p = new_p;
bool_ok = true;
} else if let (Some(prev), Some(next)) = (&w0, &w2) {
let cond = (hex_and_any(prev.pos(), &[POSTAG::D_V, POSTAG::D_R, POSTAG::D_A, POSTAG::D_T, POSTAG::D_F])
&& hex_and_any(next.pos(), &[POSTAG::D_N, POSTAG::D_V, POSTAG::D_R, POSTAG::D_S, POSTAG::A_NX, POSTAG::D_F, POSTAG::D_W]))
|| (hex_and_any(prev.pos(), &[POSTAG::D_N]) && hex_and_any(next.pos(), &[POSTAG::D_N]))
|| (hex_and_any(prev.pos(), &[POSTAG::D_V, POSTAG::D_N])
&& hex_and_any(next.pos(), &[POSTAG::D_F, POSTAG::D_T, POSTAG::A_NR, POSTAG::D_R, POSTAG::D_S, POSTAG::D_W]))
|| (hex_and_any(prev.pos(), &[POSTAG::A_NS, POSTAG::D_T, POSTAG::D_C])
&& hex_and_any(next.pos(), &[POSTAG::A_NS, POSTAG::D_T]))
|| (hex_and_any(prev.pos(), &[POSTAG::D_D]) && hex_and_any(next.pos(), &[POSTAG::D_N]))
|| (hex_and_any(prev.pos(), &[POSTAG::A_NR])
&& hex_and_any(next.pos(), &[POSTAG::A_NS, POSTAG::A_NT, POSTAG::D_S, POSTAG::D_N, POSTAG::D_V]))
|| (hex_and_any(prev.pos(), &[POSTAG::D_V]) && hex_and_any(next.pos(), &[POSTAG::D_W]))
|| (hex_and_any(prev.pos(), &[POSTAG::D_D]) && hex_and_any(next.pos(), &[POSTAG::D_V]))
|| (hex_and_any(prev.pos(), &[POSTAG::D_V]) && hex_and_any(next.pos(), &[POSTAG::D_D]))
|| (hex_and_any(prev.pos(), &[POSTAG::D_N]) && hex_and_any(next.pos(), &[POSTAG::D_V]))
|| (hex_and_any(prev.pos(), &[POSTAG::D_D]) && hex_and_any(next.pos(), &[POSTAG::D_F]));
if cond {
words[i].ow = Some(std::mem::replace(&mut words[i].w, "於".into()));
new_p = Some(POSTAG::D_P);
words[i].p = new_p;
bool_ok = true;
} else if let Some(w3) = words.get(i + 2) {
if prev.pos() & POSTAG::D_V != 0
&& next.pos() & POSTAG::D_D != 0
&& w3.pos() & POSTAG::D_V != 0
{
words[i].ow = Some(std::mem::replace(&mut words[i].w, "於".into()));
new_p = Some(POSTAG::D_P);
words[i].p = new_p;
bool_ok = true;
}
}
}
if !bool_ok && w2.as_ref().map(|n| n.pos() & POSTAG::D_T != 0).unwrap_or(false) {
words[i].ow = Some(std::mem::replace(&mut words[i].w, "於".into()));
new_p = Some(POSTAG::D_P);
words[i].p = new_p;
bool_ok = true;
}
} else if w == "么" {
if w2.as_ref().map(|n| n.pos() & POSTAG::D_W != 0).unwrap_or(true) {
words[i].ow = Some(std::mem::replace(&mut words[i].w, "麼".into()));
bool_ok = true;
}
} else if w == "余" {
if w2.as_ref().map(|n| n.w == "力").unwrap_or(false)
&& words.get(i + 2).map(|n| n.pos() & POSTAG::D_W != 0).unwrap_or(false)
{
let nw = format!("{}{}", words[i].w, words[i + 1].w);
let (p, f) = seg
.table
.table
.get(&nw)
.map(|e| (e.p, Some(e.f)))
.unwrap_or((0x101000, None));
let mut tok = Word::new(nw).with_p(p);
tok.f = f;
splice_created(&mut words, i, 2, tok, seg);
continue;
}
}
} else if w1_len > 1 {
let w = words[i].w.clone();
if let Some(caps) = Regex::new(r"^(.+)[发發]$").unwrap().captures(&w) {
let c = caps.get(1).unwrap().as_str().to_string();
if COLOR_HAIR.contains(&c) {
let nw = get_synonym(seg, &w, format!("{c}髮"));
if nw != w {
words[i].ow = Some(std::mem::replace(&mut words[i].w, nw));
bool_ok = true;
}
} else if w == format!("{c}发") && words[i].pos() & POSTAG::D_MQ != 0 {
words[i].ow = Some(std::mem::replace(&mut words[i].w, format!("{c}發")));
bool_ok = true;
} else if w == format!("{c}发")
&& (w0.is_none() || w0.as_ref().map(|p| p.pos() == POSTAG::D_W).unwrap_or(false))
{
let nw = format!("{c}髮");
if let Some(ow) = seg.table.table.get(&nw) {
if ow.s {
words[i].ow = Some(std::mem::replace(&mut words[i].w, nw));
new_p = Some(ow.p);
bool_ok = true;
}
}
}
} else if words[i].pos() & POSTAG::D_MQ != 0
&& Regex::new(r"^(.+)余$").unwrap().is_match(&w)
{
let nw = Regex::new(r"^(.+)余$").unwrap().replace(&w, "${1}餘").into_owned();
words[i].ow = Some(std::mem::replace(&mut words[i].w, nw));
bool_ok = true;
} else if words[i].pos() & POSTAG::D_MQ != 0
&& w.starts_with('几')
&& words[i].m.as_ref().map(|m| m.len() > 1).unwrap_or(false)
{
let nw = w.replacen('几', "幾", 1);
words[i].ow = Some(std::mem::replace(&mut words[i].w, nw));
bool_ok = true;
} else if words[i].pos() & POSTAG::BAD != 0 {
let mut nw = replace_li(&w);
nw = replace_hou(&nw);
nw = replace_shen(&nw);
nw = get_synonym(seg, &w, nw);
if nw != w {
words[i].ow = Some(std::mem::replace(&mut words[i].w, nw));
bool_ok = true;
}
} else if words[i].pos() & POSTAG::D_F != 0 {
let mut nw = replace_li(&w);
nw = replace_hou(&nw);
nw = get_synonym(seg, &w, nw);
if nw != w {
words[i].ow = Some(std::mem::replace(&mut words[i].w, nw));
bool_ok = true;
}
} else if words[i].pos() & POSTAG::D_S != 0 {
let nw = Regex::new(r"(.)里$").unwrap().replace(&w, "${1}裡").into_owned();
let nw = get_synonym(seg, &w, nw);
if nw != w {
words[i].ow = Some(std::mem::replace(&mut words[i].w, nw));
bool_ok = true;
}
} else if words[i].pos() & POSTAG::D_T != 0 || words[i].pos() & POSTAG::D_V != 0 {
let nw = replace_hou(&w);
let nw = get_synonym(seg, &w, nw);
if nw != w {
words[i].op = words[i].op.or(words[i].p);
words[i].ow = Some(std::mem::replace(&mut words[i].w, nw));
bool_ok = true;
}
}
}
if bool_ok {
if let Some(ow_s) = words[i].ow.clone() {
if ow_s != words[i].w {
if let Some(e) = seg.table.table.get(&words[i].w) {
if let Some(np) = new_p {
words[i].op = words[i].op.or(Some(e.p));
words[i].p = Some(np);
} else if e.p != words[i].pos() {
words[i].op = words[i].op.or(words[i].p);
words[i].p = Some(e.p);
}
if Some(e.s) != words[i].s {
words[i].os = words[i].os.or(words[i].s).or(Some(false));
words[i].s = Some(e.s);
}
}
}
}
}
i += 1;
}
words
}
}