1use std::collections::{BTreeMap, BTreeSet, HashMap, HashSet, VecDeque};
4use iter_index::IndexerIterator;
5use vectree::VecTree;
6use lexigram_core::alt::Alternative;
7use lexigram_core::log::LogMsg;
8use lexigram_core::TokenId;
9use crate::grammar::{grtree_to_str, GrTreeExt, LLParsingTable, NTConversion, ProdRuleSet};
10use crate::{columns_to_str, indent_source, AltId, CharLen, NameFixer, NameTransformer, SourceSpacer, StructLibs, SymbolTable, VarId, LL1};
11use crate::fixed_sym_table::{FixedSymTable, SymInfoTable};
12use crate::alt::ruleflag;
13use crate::build::{BuildError, BuildErrorSource, BuildFrom, HasBuildErrorSource, TryBuildFrom};
14use crate::CollectJoin;
15use crate::grammar::origin::{FromPRS, Origin};
16use crate::lexergen::LexigramCrate;
17use crate::log::{BufLog, LogReader, LogStatus, Logger};
18use crate::parser::{OpCode, Parser, Symbol};
19use crate::segments::Segments;
20use crate::segmap::Seg;
21
22pub(crate) mod tests;
23
24pub(crate) fn symbol_to_code(s: &Symbol) -> String {
27 match s {
28 Symbol::Empty => "Symbol::Empty".to_string(),
29 Symbol::T(t) => format!("Symbol::T({t})"),
30 Symbol::NT(nt) => format!("Symbol::NT({nt})"),
31 Symbol::End => "Symbol::End".to_string(),
32 }
33}
34
35#[derive(Clone, Debug, PartialEq)]
38struct ItemInfo {
39 name: String,
40 sym: Symbol, owner: VarId, index: Option<usize> }
44
45#[allow(unused)]
46impl ItemInfo {
47 fn to_str(&self, symbol_table: Option<&SymbolTable>) -> String {
48 format!("{} ({}{}, ◄{})",
49 self.name,
50 self.sym.to_str(symbol_table),
51 if let Some(n) = self.index { format!(", [{n}]") } else { "".to_string() },
52 Symbol::NT(self.owner).to_str(symbol_table))
53 }
54}
55
56pub struct ParserTables {
66 num_nt: usize,
67 num_t: usize,
68 alt_var: Vec<VarId>,
70 alts: Vec<Alternative>,
71 opcodes: Vec<Vec<OpCode>>,
72 init_opcodes: Vec<OpCode>,
73 table: Vec<AltId>,
74 symbol_table: FixedSymTable,
75 start: VarId,
76 include_alts: bool,
77}
78
79impl ParserTables {
80 pub fn new(
81 parsing_table: LLParsingTable,
82 symbol_table: FixedSymTable,
83 opcodes: Vec<Vec<OpCode>>,
84 init_opcodes: Vec<OpCode>,
85 start: VarId,
86 include_alts: bool
87 ) -> Self {
88 assert!(parsing_table.num_nt > start as usize);
89 let num_nt = parsing_table.num_nt;
90 let num_t = parsing_table.num_t;
91 let table = parsing_table.table;
92 let (factor_var, alts): (Vec<_>, Vec<_>) = parsing_table.alts.into_iter().unzip();
93 ParserTables { num_nt, num_t, alt_var: factor_var, alts, opcodes, init_opcodes, table, symbol_table, start, include_alts }
94 }
95
96 pub fn make_parser(&self) -> Parser<'_> {
97 Parser::new(
98 self.num_nt,
99 self.num_t,
100 self.alt_var.as_slice(),
101 if self.include_alts { self.alts.clone() } else { vec![] },
102 self.opcodes.clone(),
103 self.init_opcodes.clone(),
104 self.table.as_slice(),
105 self.symbol_table.clone(),
106 self.start,
107 )
108 }
109}
110
111impl BuildFrom<ParserGen> for ParserTables {
112 fn build_from(parser_gen: ParserGen) -> Self {
115 ParserTables::new(
116 parser_gen.parsing_table,
117 parser_gen.symbol_table.to_fixed_sym_table(),
118 parser_gen.opcodes,
119 parser_gen.init_opcodes,
120 parser_gen.start,
121 parser_gen.include_alts
122 )
123 }
124}
125
126impl TryBuildFrom<ParserGen> for ParserTables {
128 type Error = BuildError;
129
130 fn try_build_from(source: ParserGen) -> Result<Self, Self::Error> {
131 if source.get_log().has_no_errors() {
132 Ok(ParserTables::build_from(source))
133 } else {
134 Err(BuildError::new(source.give_log(), BuildErrorSource::ParserGen))
135 }
136 }
137}
138
139#[derive(Clone, PartialEq, Debug)]
150pub enum NTValue {
151 None,
153 Parents,
155 Default,
157 SetIds(Vec<VarId>),
159 SetNames(Vec<String>),
161}
162
163impl NTValue {
164 pub const DEFAULT: &str = "<default>";
166 pub const PARENTS: &str = "<parents>";
168
169 pub fn is_none(&self) -> bool {
170 matches!(self, NTValue::None)
171 }
172
173 pub fn is_parents(&self) -> bool {
174 matches!(self, NTValue::Parents)
175 }
176
177 pub fn is_default(&self) -> bool {
178 matches!(self, NTValue::Default)
179 }
180
181 pub fn is_ids(&self) -> bool {
182 matches!(self, NTValue::SetIds(_))
183 }
184
185 pub fn is_names(&self) -> bool {
186 matches!(self, NTValue::SetNames(_))
187 }
188}
189
190pub static DEFAULT_LISTENER_NAME: &str = "Parser";
193
194pub type SpanNbr = u16;
195
196fn count_span_nbr(opcode: &[OpCode]) -> SpanNbr {
197 let count = opcode.iter().filter(|op| op.has_span()).count();
198 count.try_into().unwrap_or_else(|_| panic!("# span = {count} > {}", SpanNbr::MAX))
199}
200
201struct SourceInputContext<'a> {
202 parent_has_value : bool,
203 parent_nt : usize,
204 pinfo : &'a LLParsingTable,
205 syns : &'a Vec<VarId>,
206 ambig_op_alts : &'a BTreeMap<AltId, Vec<AltId>>,
207}
208
209struct SourceState<'a> {
210 init_nt_done : &'a mut HashSet<VarId>,
211 span_init : &'a mut HashSet<VarId>,
212 nt_contexts : &'a mut Vec<Option<Vec<AltId>>>,
213 exit_alt_done : &'a mut HashSet<VarId>,
214 exit_fixer : &'a mut NameFixer,
215}
216
217struct WrapperSources {
218 src : Vec<String>,
219 src_listener_decl : Vec<String>,
220 src_skel : Vec<String>,
221 src_types : Vec<String>,
222 src_init : Vec<Vec<String>>,
223 src_exit : Vec<Vec<String>>,
224 src_wrapper_impl : Vec<String>,
225}
226
227#[derive(Debug)]
228pub struct ParserGen {
229 parsing_table: LLParsingTable,
230 symbol_table: SymbolTable,
231 terminal_hooks: Vec<TokenId>,
232 name: String,
233 nt_value: Vec<bool>,
234 nt_parent: Vec<Vec<VarId>>,
236 var_alts: Vec<Vec<AltId>>,
237 origin: Origin<VarId, FromPRS>,
238 item_ops: Vec<Vec<Symbol>>,
239 opcodes: Vec<Vec<OpCode>>,
240 init_opcodes: Vec<OpCode>,
241 nt_name: Vec<(String, String, String)>,
242 alt_info: Vec<Option<(VarId, String)>>,
243 item_info: Vec<Vec<ItemInfo>>,
244 child_repeat_endpoints: HashMap<VarId, Vec<AltId>>,
245 gen_parser: bool,
247 gen_wrapper: bool,
249 indent: usize,
251 types_indent: usize,
253 listener_indent: usize,
255 gen_span_params: bool,
257 gen_token_enums: bool,
258 span_nbrs: Vec<SpanNbr>,
259 span_nbrs_sep_list: HashMap<AltId, SpanNbr>,
261 start: VarId,
262 nt_conversion: HashMap<VarId, NTConversion>,
263 headers: Vec<String>,
264 used_libs: StructLibs,
265 nt_type: HashMap<VarId, String>,
266 log: BufLog,
267 include_alts: bool,
268 lib_crate: LexigramCrate,
269}
270
271impl ParserGen {
272 pub fn build_from_rules<T>(mut rules: ProdRuleSet<T>, name: String) -> Self
278 where
279 ProdRuleSet<LL1>: BuildFrom<ProdRuleSet<T>>,
280 {
281 rules.log.add_note("building parser gen from rules...");
282 let mut ll1_rules = ProdRuleSet::<LL1>::build_from(rules);
283 assert_eq!(ll1_rules.get_log().num_errors(), 0);
284 let parsing_table = ll1_rules.make_parsing_table(true);
285 let num_nt = ll1_rules.get_num_nt();
286 let start = ll1_rules.get_start().unwrap();
287 let mut var_alts = vec![vec![]; num_nt];
288 for (alt_id, (var_id, _)) in parsing_table.alts.iter().index() {
289 var_alts[*var_id as usize].push(alt_id);
290 }
291 let mut nt_parent: Vec<Vec<VarId>> = vec![vec![]; num_nt];
292 for var_id in 0..num_nt {
293 let top_var_id = parsing_table.get_top_parent(var_id as VarId) as usize;
294 nt_parent[top_var_id].push(var_id as VarId);
295 }
296 let ProdRuleSet { symbol_table, nt_conversion, origin, .. } = ll1_rules;
297 let mut builder = ParserGen {
298 parsing_table,
299 symbol_table: symbol_table.expect(stringify!("symbol table is required to create a {}", std::any::type_name::<Self>())),
300 gen_span_params: false,
301 gen_token_enums: false,
302 name,
303 nt_value: vec![false; num_nt],
304 nt_parent,
305 var_alts,
306 origin,
307 terminal_hooks: Vec::new(),
308 item_ops: Vec::new(),
309 opcodes: Vec::new(),
310 init_opcodes: Vec::new(),
311 nt_name: Vec::new(),
312 alt_info: Vec::new(),
313 item_info: Vec::new(),
314 child_repeat_endpoints: HashMap::new(),
315 gen_parser: true,
316 gen_wrapper: true,
317 indent: 0,
318 types_indent: 0,
319 listener_indent: 0,
320 span_nbrs: Vec::new(),
321 span_nbrs_sep_list: HashMap::new(),
322 start,
323 nt_conversion,
324 headers: Vec::new(),
325 used_libs: StructLibs::new(),
326 nt_type: HashMap::new(),
327 log: ll1_rules.log,
328 include_alts: false,
329 lib_crate: LexigramCrate::Core,
330 };
331 builder.make_opcodes();
332 builder.make_span_nbrs();
333 builder
334 }
335
336 pub fn set_name(&mut self, name: String) {
337 self.name = name;
338 }
339
340 pub fn get_name(&self) -> &str {
341 &self.name
342 }
343
344 #[inline]
345 pub fn get_symbol_table(&self) -> Option<&SymbolTable> {
346 Some(&self.symbol_table)
347 }
348
349 #[inline]
350 pub fn get_parsing_table(&self) -> &LLParsingTable {
351 &self.parsing_table
352 }
353
354 #[inline]
355 pub fn set_terminal_hooks(&mut self, terminal_hooks: Vec<TokenId>) {
356 if !terminal_hooks.is_empty() {
357 self.gen_token_enums = true;
358 }
359 self.terminal_hooks = terminal_hooks;
360 self.add_opcode_hooks();
361 }
362
363 #[inline]
364 pub fn add_header<T: Into<String>>(&mut self, header: T) {
365 self.headers.push(header.into());
366 }
367
368 #[inline]
369 pub fn extend_headers<I: IntoIterator<Item=T>, T: Into<String>>(&mut self, headers: I) {
370 self.headers.extend(headers.into_iter().map(|s| s.into()));
371 }
372
373 #[inline]
374 pub fn add_lib<T: Into<String>>(&mut self, lib:T) {
375 self.used_libs.add(lib);
376 }
377
378 #[inline]
379 pub fn extend_libs<I: IntoIterator<Item=T>, T: Into<String>>(&mut self, libs: I) {
380 self.used_libs.extend(libs);
381 }
382
383 #[inline]
384 pub fn add_nt_type<T: Into<String>>(&mut self, org_var: VarId, var_type: T) {
387 let var = self.conv_nt(org_var).unwrap_or_else(|| panic!("var {org_var} doesn't exist"));
388 self.nt_type.insert(var, var_type.into());
389 }
390
391 #[inline]
392 pub fn get_nt_type(&self, v: VarId) -> &str {
393 self.nt_type.get(&v).unwrap().as_str()
394 }
395
396 pub fn set_nt_value(&mut self, nt_value: &NTValue) {
398 let num_nt = self.get_symbol_table().unwrap().get_num_nt() as VarId;
399 let mut stack = vec![nt_value];
400 let mut neg_stack = vec![];
401 while let Some(nt_value) = stack.pop() {
402 match nt_value {
403 NTValue::None => {}
404 NTValue::Parents => {
405 for v in 0..num_nt {
406 if self.get_nt_parent(v).is_none() {
407 self.nt_value[v as usize] = true;
408 }
409 }
410 }
411 NTValue::Default => {
412 for v in 0..num_nt {
413 if self.get_nt_parent(v).is_none() || self.nt_has_all_flags(v, ruleflag::CHILD_REPEAT | ruleflag::L_FORM) {
414 self.nt_value[v as usize] = true;
415 }
416 }
417 }
418 NTValue::SetIds(ids) => {
419 for v in ids {
420 if *v < num_nt {
421 self.nt_value[*v as usize] = true;
422 } else {
423 self.log.add_error(format!("setting value of NT #{v}, which doesn't exist"));
424 }
425 }
426 }
427 NTValue::SetNames(names) => {
428 let name_to_id = self.symbol_table.get_nonterminals().index::<VarId>()
429 .map(|(v, name)| (name.as_str(), v))
430 .collect::<HashMap<&str, VarId>>();
431 for name in names {
432 match name.as_str() {
433 NTValue::DEFAULT => stack.push(&NTValue::Default),
434 NTValue::PARENTS => stack.push(&NTValue::Parents),
435 mut nt_name => {
436 let add = if !nt_name.starts_with('-') {
437 true
438 } else {
439 nt_name = &nt_name[1..];
440 false
441 };
442 if let Some(v) = name_to_id.get(nt_name) {
443 if add {
444 self.nt_value[*v as usize] = true;
445 } else {
446 neg_stack.push(*v);
447 }
448 } else {
449 self.log.add_error(format!("setting value of NT '{name}', which doesn't exist"));
450 }
451 }
452 }
453 }
454 }
455 }
456 }
457 for v in neg_stack {
458 self.nt_value[v as usize] = false;
459 }
460 }
461
462 #[inline]
463 pub fn set_nt_has_value(&mut self, v: VarId, has_value: bool) {
464 self.nt_value[v as usize] = has_value;
465 }
466
467 pub fn set_gen_parser(&mut self, gen_parser: bool) {
469 self.gen_parser = gen_parser;
470 }
471
472 pub fn set_gen_wrapper(&mut self, gen_wrapper: bool) {
474 self.gen_wrapper = gen_wrapper;
475 }
476
477 pub fn set_indent(&mut self, indent: usize) {
479 self.indent = indent;
480 }
481
482 pub fn set_types_indent(&mut self, indent: usize) {
485 self.types_indent = indent;
486 }
487
488 pub fn set_listener_indent(&mut self, indent: usize) {
491 self.listener_indent = indent;
492 }
493
494 pub fn set_indents(&mut self, wrapper: usize, types: usize, listner: usize) {
497 self.indent = wrapper;
498 self.types_indent = types;
499 self.listener_indent = listner;
500 }
501
502 pub fn set_gen_span_params(&mut self, gen_span_params: bool) {
504 self.gen_span_params = gen_span_params;
505 }
506
507 pub fn set_gen_token_enums(&mut self, gen_token_enums: bool) {
528 self.gen_token_enums = gen_token_enums;
529 }
530
531 #[inline]
532 pub fn get_nt_parent(&self, v: VarId) -> Option<VarId> {
533 self.parsing_table.parent[v as usize]
534 }
535
536 pub fn set_include_alts(&mut self, include_alts: bool) {
539 self.include_alts = include_alts;
540 }
541
542 #[inline]
543 pub fn use_full_lib(&mut self, use_full_lib: bool) {
544 self.lib_crate = if use_full_lib { LexigramCrate::Full } else { LexigramCrate::Core };
545 }
546
547 #[inline]
548 pub fn set_crate(&mut self, lcrate: LexigramCrate) {
549 self.lib_crate = lcrate;
550 }
551
552 #[cfg(test)] fn get_original_alt_str(&self, a_id: AltId, symbol_table: Option<&SymbolTable>) -> Option<String> {
554 let (_var, f) = &self.parsing_table.alts[a_id as usize];
555 f.get_origin().and_then(|(o_v, o_id)| {
556 Some(format!(
557 "{} -> {}",
558 Symbol::NT(o_v).to_str(symbol_table),
559 grtree_to_str(self.origin.get_tree(o_v).unwrap(), Some(o_id), None, Some(o_v), symbol_table, false)
560 ))
561 })
562 }
563
564 fn conv_nt(&self, org_var: VarId) -> Option<VarId> {
570 match self.nt_conversion.get(&org_var) {
571 None => if (org_var as usize) < self.parsing_table.num_nt { Some(org_var) } else { None },
572 Some(NTConversion::MovedTo(new)) => Some(*new),
573 Some(NTConversion::Removed) => None
574 }
575 }
576
577 #[allow(unused)]
578 fn nt_has_all_flags(&self, var: VarId, flags: u32) -> bool {
579 self.parsing_table.flags[var as usize] & flags == flags
580 }
581
582 #[allow(unused)]
583 fn nt_has_any_flags(&self, var: VarId, flags: u32) -> bool {
584 self.parsing_table.flags[var as usize] & flags != 0
585 }
586
587 #[allow(unused)]
588 fn sym_has_flags(&self, s: &Symbol, flags: u32) -> bool {
589 if let Symbol::NT(nt) = s { self.nt_has_all_flags(*nt, flags) } else { false }
590 }
591
592 #[allow(unused)]
593 fn sym_has_value(&self, symbol: &Symbol) -> bool {
594 match symbol {
595 Symbol::T(t) => self.symbol_table.is_token_data(*t),
596 Symbol::NT(nt) => self.nt_value[*nt as usize],
597 _ => false
598 }
599 }
600
601 fn full_alt_components(&self, a_id: AltId, emphasis: Option<VarId>) -> (String, String) {
602 const VERBOSE: bool = false;
603 if VERBOSE { println!("full_alt_components(a_id = {a_id}):"); }
604 let &(mut v_a, ref alt) = &self.parsing_table.alts[a_id as usize];
605 while self.parsing_table.flags[v_a as usize] & ruleflag::CHILD_L_FACT != 0 {
606 v_a = *self.parsing_table.parent[v_a as usize].as_ref().unwrap();
607 }
608 let symtab = self.get_symbol_table();
609 if let Some(v_emph) = emphasis {
610 let parent_nt = self.parsing_table.get_top_parent(v_emph);
611 if let Some((t_emph, id_emph)) = self.origin.get(v_emph) {
612 return ((Symbol::NT(parent_nt).to_str(symtab)), grtree_to_str(t_emph, None, Some(id_emph), Some(parent_nt), symtab, true));
613 } else {
614 return (Symbol::NT(parent_nt).to_str(symtab), format!("<VAR {v_emph} NOT FOUND>"));
615 }
616 }
617 if let Some((vo, id)) = alt.get_origin() {
618 let t = self.origin.get_tree(vo).unwrap();
619 let flags = self.parsing_table.flags[v_a as usize];
620 if v_a != vo && flags & ruleflag::CHILD_REPEAT != 0 {
621 (
623 String::new(),
624 format!("`{}` {} in `{} -> {}`",
625 grtree_to_str(t, Some(id), None, Some(vo), symtab, true),
626 if flags & ruleflag::L_FORM != 0 { "iteration" } else { "item" },
627 Symbol::NT(vo).to_str(symtab),
628 grtree_to_str(t, None, Some(id), Some(vo), symtab, true))
629 )
630 } else {
631 let root = Some(id);
632 (Symbol::NT(vo).to_str(symtab), grtree_to_str(t, root, None, Some(vo), symtab, true))
633 }
634 } else {
635 (Symbol::NT(v_a).to_str(symtab), format!("<alt {a_id} NOT FOUND>"))
636 }
637 }
638
639 fn full_alt_str(&self, a_id: AltId, emphasis: Option<VarId>, quote: bool) -> String {
641 let (left, right) = self.full_alt_components(a_id, emphasis);
642 if left.is_empty() {
643 right
644 } else {
645 format!("{q}{left} -> {right}{q}", q = if quote { "`" } else { "" })
646 }
647 }
648
649 fn make_opcodes(&mut self) {
650 const VERBOSE: bool = false;
651 self.log.add_note("- making opcodes...");
652 self.opcodes.clear();
653 self.init_opcodes = vec![OpCode::End, OpCode::NT(self.start)];
654 for (alt_id, (var_id, alt)) in self.parsing_table.alts.iter().index() {
655 if VERBOSE {
656 println!("{alt_id}: {}", alt.to_rule_str(*var_id, self.get_symbol_table(), 0));
657 }
658 let flags = self.parsing_table.flags[*var_id as usize];
659 let stack_sym = Symbol::NT(*var_id);
660 let mut new = self.parsing_table.alts[alt_id as usize].1.iter().filter(|s| !s.is_empty()).rev().cloned().to_vec();
661 if VERBOSE { println!(" - {}", new.iter().map(|s| s.to_str(self.get_symbol_table())).join(" ")); }
662 let mut opcode = Vec::<OpCode>::new();
663 let mut parent = self.parsing_table.parent[*var_id as usize];
664 if flags & ruleflag::CHILD_L_FACT != 0 {
665 while self.nt_has_all_flags(parent.unwrap(), ruleflag::CHILD_L_FACT) {
666 parent = self.parsing_table.parent[parent.unwrap() as usize];
667 }
668 let parent = parent.unwrap();
669 let parent_r_form_right_rec = self.parsing_table.flags[parent as usize] & ruleflag::R_RECURSION != 0 && flags & ruleflag::L_FORM == 0;
673 if VERBOSE {
674 println!(" - child lfact, parent: {}, !parent_r_form_right_rec = !{parent_r_form_right_rec}, match = {}",
675 Symbol::NT(parent).to_str(self.get_symbol_table()),
676 new.first() == Some(&Symbol::NT(parent)));
677 }
678 if new.first() == Some(&Symbol::NT(parent)) && !parent_r_form_right_rec {
679 opcode.push(OpCode::Loop(parent));
680 new.remove(0);
681 }
682 }
683 let parent_lrec_no_lfact = flags & (ruleflag::PARENT_L_RECURSION | ruleflag::PARENT_L_FACTOR) == ruleflag::PARENT_L_RECURSION;
684 if flags & ruleflag::PARENT_L_FACTOR == 0 ||
685 parent_lrec_no_lfact ||
686 new.iter().all(|s| if let Symbol::NT(ch) = s { !self.nt_has_all_flags(*ch, ruleflag::CHILD_L_FACT) } else { true })
687 {
688 opcode.push(OpCode::Exit(alt_id)); }
697 opcode.extend(new.into_iter().map(OpCode::from));
698 let r_form_right_rec = flags & ruleflag::R_RECURSION != 0 && flags & ruleflag::L_FORM == 0;
699 if VERBOSE { println!(" - r_form_right_rec = {r_form_right_rec} = {} || {}",
700 flags & ruleflag::R_RECURSION != 0 && flags & ruleflag::L_FORM == 0,
701 flags & ruleflag::CHILD_L_FACT != 0 && self.parsing_table.flags[parent.unwrap() as usize] & ruleflag::R_RECURSION != 0 && flags & ruleflag::L_FORM == 0); }
702 if opcode.get(1).map(|op| op.matches(stack_sym)).unwrap_or(false) && !r_form_right_rec {
703 opcode.swap(0, 1);
706 if VERBOSE { println!(" - swap 0, 1: {}", opcode.iter().map(|s| s.to_str(self.get_symbol_table())).join(" ")); }
707 } else if parent_lrec_no_lfact {
708 if let Some(OpCode::NT(x)) = opcode.get(1) {
709 if self.nt_has_all_flags(*x, ruleflag::CHILD_L_RECURSION) {
710 opcode.swap(0, 1);
713 if VERBOSE { println!(" - swap 0, 1: {}", opcode.iter().map(|s| s.to_str(self.get_symbol_table())).join(" ")); }
714 }
715 }
716 } else if flags & ruleflag::CHILD_INDEPENDENT_AMBIGUITY != 0 && opcode.len() > 1 {
717 if let Some(OpCode::NT(var_prime)) = opcode.get(1) {
719 let vp = *var_prime; if self.nt_has_all_flags(vp, ruleflag::CHILD_AMBIGUITY) {
721 opcode.swap(0, 1);
722 opcode[0] = OpCode::Loop(vp);
723 if VERBOSE { println!(" - child indep ambig: {}", opcode.iter().map(|s| s.to_str(self.get_symbol_table())).join(" ")); }
724 }
725 }
726 }
727 if flags & ruleflag::CHILD_L_FACT != 0 && opcode.len() >= 2 {
728 if self.nt_has_all_flags(parent.unwrap(), ruleflag::R_RECURSION | ruleflag::L_FORM)
729 && opcode[1] == OpCode::NT(parent.unwrap())
730 {
731 opcode.swap(0, 1);
732 opcode[0] = OpCode::Loop(parent.unwrap());
733 }
734 let fact_top = self.parsing_table.get_top_parent(*var_id);
735 if VERBOSE {
736 println!(" - check for initial exit swap: opcode = [{}], daddy = {}",
737 opcode.iter().map(|s| s.to_str(self.get_symbol_table())).join(" "),
738 Symbol::NT(fact_top).to_str(self.get_symbol_table()));
739 }
740 if self.parsing_table.flags[fact_top as usize] & ruleflag::PARENT_L_RECURSION != 0 &&
741 matches!(opcode[0], OpCode::Exit(_)) &&
742 matches!(opcode[1], OpCode::NT(v) if self.parsing_table.flags[v as usize] & ruleflag::CHILD_L_RECURSION != 0)
743 {
744 if VERBOSE {
745 println!(" swapping for initial exit_{}: {} <-> {}",
746 Symbol::NT(fact_top).to_str(self.get_symbol_table()).to_lowercase(),
747 opcode[0].to_str(self.get_symbol_table()),
748 opcode[1].to_str(self.get_symbol_table())
749 );
750 }
751 opcode.swap(0, 1);
752 }
753 }
754 opcode.iter_mut().for_each(|o| {
755 if let OpCode::NT(v) = o {
756 if v == var_id && !r_form_right_rec {
759 *o = OpCode::Loop(*v)
760 }
761 }
762 });
763 if VERBOSE { println!(" -> {}", opcode.iter().map(|s| s.to_str(self.get_symbol_table())).join(" ")); }
764 self.opcodes.push(opcode);
765 }
766 }
767
768 fn add_opcode_hooks(&mut self) {
808 const VERBOSE: bool = false;
809 self.log.add_note("- adding hooks into opcodes...");
810 let hooks: HashSet<TokenId> = self.terminal_hooks.iter().cloned().collect();
811 let num_nt = self.parsing_table.num_nt;
812 let num_t = self.parsing_table.num_t;
813 let err = self.parsing_table.alts.len() as AltId;
814 if VERBOSE {
815 self.parsing_table.print(self.get_symbol_table(), 0);
816 println!("num_nt = {num_nt}\nnum_t = {num_t}\ntable: {}", self.parsing_table.table.len());
817 }
818 if VERBOSE { println!("hooks: {}", self.terminal_hooks.iter().map(|t| self.symbol_table.get_t_name(*t)).join(", ")); }
819 let deps: HashSet<VarId> = (0..num_nt as VarId)
820 .filter(|&nt| hooks.iter().any(|&t| self.parsing_table.table[nt as usize * num_t + t as usize] < err))
821 .collect();
822 if VERBOSE { println!("deps = {deps:?} = {}", deps.iter().map(|nt| self.symbol_table.get_nt_name(*nt)).join(", ")); }
823
824 if deps.contains(&self.start) {
826 self.init_opcodes = vec![OpCode::End, OpCode::NT(self.start), OpCode::Hook];
827 }
828 let mut changed = false;
829 for opcodes in self.opcodes.iter_mut() {
830 let mut new = vec![];
831 let n = opcodes.len();
832 for op in &opcodes[..n - 1] {
833 new.push(*op);
834 match op {
835 OpCode::T(t) if hooks.contains(t) => {
836 new.push(OpCode::Hook);
837 }
838 OpCode::NT(nt) | OpCode::Loop(nt) if deps.contains(nt) => {
839 new.push(OpCode::Hook);
840 }
841 _ => {}
842 }
843 }
844 if new.len() + 1 > n {
845 new.push(opcodes[n - 1]);
846 *opcodes = new;
847 changed = true;
848 }
849 }
850 if VERBOSE && changed {
851 println!("new opcodes:");
852 let mut cols = vec![];
853 let tbl = self.get_symbol_table();
854 for (i, (opcodes, (nt, alt))) in self.opcodes.iter().zip(&self.parsing_table.alts).enumerate() {
855 cols.push(vec![
856 i.to_string(),
857 format!("{} -> ", Symbol::NT(*nt).to_str(tbl)),
858 alt.to_str(tbl),
859 opcodes.iter().map(|op| op.to_str(tbl)).join(" "),
860 ]);
861 }
862 println!("{}", indent_source(vec![columns_to_str(cols, None)], 4))
863 }
864 }
865
866 fn make_span_nbrs(&mut self) {
867 self.log.add_note("- making spans...");
868 let mut span_nbrs = vec![0 as SpanNbr; self.parsing_table.alts.len()];
869 for (alt_id, (var_id, _)) in self.parsing_table.alts.iter().enumerate() {
870 let opcode = &self.opcodes[alt_id];
871 let mut span_nbr = span_nbrs[alt_id] + count_span_nbr(opcode);
872 if self.nt_has_any_flags(*var_id, ruleflag::CHILD_REPEAT | ruleflag::CHILD_L_RECURSION) ||
873 self.nt_has_all_flags(*var_id, ruleflag::R_RECURSION | ruleflag::L_FORM) {
874 span_nbr += 1;
876 }
877 if matches!(opcode.first(), Some(OpCode::NT(nt)) if nt != var_id && self.parsing_table.flags[*nt as usize] & ruleflag::CHILD_L_RECURSION != 0) {
878 span_nbr -= 1;
880 }
881 if self.nt_has_all_flags(*var_id, ruleflag::PARENT_L_FACTOR) {
885 if let Some(OpCode::NT(nt)) = opcode.first() {
886 span_nbr -= 1;
887 for a_id in self.var_alts[*nt as usize].iter() {
888 span_nbrs[*a_id as usize] += span_nbr;
889 }
891 span_nbr = 0;
893 }
894 }
895 span_nbrs[alt_id] = span_nbr;
896 }
897 self.span_nbrs = span_nbrs;
898 }
899
900 fn get_group_alts(&self, g: &[VarId]) -> Vec<(VarId, AltId)> {
901 g.iter().flat_map(|c|
902 self.var_alts[*c as usize].iter().map(|a| (*c, *a))
903 ).collect::<Vec<_>>()
904 }
905
906 fn gather_alts(&self, nt: VarId) -> Vec<AltId> {
910 const VERBOSE: bool = false;
911 let mut alt = vec![];
912 let mut explore = VecDeque::<VarId>::new();
913 explore.push_back(nt);
914 while !explore.is_empty() {
915 let var = explore.pop_front().unwrap();
916 if VERBOSE { println!("{var}: alt = {} | explore = {} | alts: {}",
917 alt.iter().join(", "), explore.iter().join(", "),
918 &self.var_alts[var as usize].iter().join(", ")); }
919 for a in &self.var_alts[var as usize] {
920 let (_, alter) = &self.parsing_table.alts[*a as usize];
921 if let Some(Symbol::NT(last)) = alter.symbols().last() {
922 if self.nt_has_all_flags(*last, ruleflag::CHILD_L_FACT) {
923 explore.push_back(*last);
925 continue;
926 }
927 }
928 alt.push(*a);
929 }
930 if VERBOSE { println!(" => alt = {} | explore = {}", alt.iter().join(", "), explore.iter().join(", ")); }
931 }
932 alt
933 }
934
935 fn calc_nt_value(&mut self) {
936 const VERBOSE: bool = false;
937 self.log.add_note("- calculating nonterminals' value...");
938 for g in self.nt_parent.iter().filter(|va| !va.is_empty()) {
940 let group = self.get_group_alts(g);
942 let mut re_evaluate = true;
943 let g_top = g[0];
944 let is_ambig = self.nt_has_all_flags(g_top, ruleflag::PARENT_AMBIGUITY);
945 while re_evaluate {
946 re_evaluate = false;
947 let mut nt_used = HashSet::<VarId>::new();
948 if VERBOSE {
949 let ids = group.iter().map(|(v, _)| *v).collect::<BTreeSet<VarId>>();
950 println!("parent: {}, NT with value: {}",
951 Symbol::NT(g[0]).to_str(self.get_symbol_table()),
952 ids.into_iter().filter_map(|v|
953 if self.nt_value[v as usize] { Some(Symbol::NT(v as VarId).to_str(self.get_symbol_table())) } else { None }
954 ).join(", "));
955 }
956 for (var_id, alt_id) in &group {
957 let mut has_value = false;
959 for s in &self.opcodes[*alt_id as usize] {
960 match s {
961 OpCode::T(t) =>
962 has_value |= self.symbol_table.is_token_data(*t),
963 OpCode::NT(nt) => {
964 let is_ambig_top = is_ambig && self.get_nt_parent(*nt) == Some(g_top)
965 && !self.nt_has_any_flags(*nt, ruleflag::CHILD_L_RECURSION | ruleflag::CHILD_REPEAT);
966 let var = if is_ambig_top { g_top } else { *nt };
967 nt_used.insert(var);
968 has_value |= self.nt_value[var as usize]
969 },
970 _ => {}
971 }
972 }
973 if has_value && self.parsing_table.parent[*var_id as usize].is_some() {
975 let mut child_nt = *var_id as usize;
977 while self.parsing_table.flags[child_nt] & ruleflag::CHILD_REPEAT == 0 {
978 if let Some(parent) = self.parsing_table.parent[child_nt] {
979 child_nt = parent as usize;
980 } else {
981 break;
982 }
983 }
984 if self.parsing_table.flags[child_nt] & (ruleflag::CHILD_REPEAT | ruleflag::L_FORM) == ruleflag::CHILD_REPEAT {
987 if VERBOSE && !self.nt_value[child_nt] {
988 print!(" | {} is now valued {}",
989 Symbol::NT(child_nt as VarId).to_str(self.get_symbol_table()),
990 if nt_used.contains(&(child_nt as VarId)) { "and was used before" } else { "but wasn't used before" }
991 );
992 }
993 re_evaluate |= !self.nt_value[child_nt] && nt_used.contains(&(child_nt as VarId));
994 self.nt_value[child_nt] = true;
995 }
996 }
997 }
998 }
999 }
1000 }
1001
1002 pub(crate) fn make_item_ops(&mut self) {
1003 const VERBOSE: bool = false;
1004 self.calc_nt_value();
1005 self.log.add_note("- making item ops...");
1006 let info = &self.parsing_table;
1007 let mut items = vec![Vec::<Symbol>::new(); self.parsing_table.alts.len()];
1008 if VERBOSE {
1009 println!("Groups:");
1010 for g in self.nt_parent.iter().filter(|va| !va.is_empty()) {
1011 let group = self.get_group_alts(g);
1012 let ids = group.iter().map(|(v, _)| *v).collect::<BTreeSet<VarId>>();
1013 println!("{}: {}, alts {}",
1014 Symbol::NT(g[0]).to_str(self.get_symbol_table()),
1015 ids.iter().map(|v| Symbol::NT(*v).to_str(self.get_symbol_table())).join(", "),
1016 group.iter().map(|(_, a)| a.to_string()).join(", ")
1017 );
1018 }
1019 }
1020 let mut alts_to_revisit = HashSet::<AltId>::new();
1021 for g in self.nt_parent.iter().filter(|va| !va.is_empty()) {
1023 let group = self.get_group_alts(g);
1025 let g_top = g[0];
1026 let is_ambig = self.nt_has_all_flags(g_top, ruleflag::PARENT_AMBIGUITY);
1027 if VERBOSE {
1028 let ids = group.iter().map(|(v, _)| *v).collect::<BTreeSet<VarId>>();
1029 println!("parent: {}, NT with value: {}",
1030 Symbol::NT(g[0]).to_str(self.get_symbol_table()),
1031 ids.into_iter().filter_map(|v|
1032 if self.nt_value[v as usize] { Some(Symbol::NT(v as VarId).to_str(self.get_symbol_table())) } else { None }
1033 ).join(", "));
1034 }
1035 let g_top_has_value = self.nt_value[g_top as usize];
1036 for (var_id, alt_id) in &group {
1037 let ambig_loop_value = g_top_has_value && is_ambig && self.nt_has_all_flags(*var_id, ruleflag::CHILD_L_RECURSION);
1038 items[*alt_id as usize] = if ambig_loop_value { vec![Symbol::NT(g_top)] } else { vec![] };
1039 }
1040 for (var_id, alt_id) in &group {
1041 let opcode = &self.opcodes[*alt_id as usize];
1042 let (_, alt) = &info.alts[*alt_id as usize];
1043 if VERBOSE {
1044 print!("- {alt_id}: {} -> {} [{}]",
1045 Symbol::NT(*var_id).to_str(self.get_symbol_table()),
1046 alt.to_str(self.get_symbol_table()),
1047 opcode.iter().map(|op| op.to_str(self.get_symbol_table())).join(" "));
1048 }
1049 let flags = info.flags[*var_id as usize];
1050
1051 let mut has_sep_list_child_without_value = false;
1055 let mut values = self.opcodes[*alt_id as usize].iter().rev()
1056 .filter_map(|s| {
1057 let sym_maybe = match s {
1058 OpCode::T(t) => Some(Symbol::T(*t)),
1059 OpCode::NT(nt) => {
1060 let is_ambig_top = is_ambig && self.get_nt_parent(*nt) == Some(g_top)
1061 && !self.nt_has_any_flags(*nt, ruleflag::CHILD_L_RECURSION | ruleflag::CHILD_REPEAT);
1062 let var = if is_ambig_top { g_top } else { *nt };
1063 Some(Symbol::NT(var))
1064 },
1065 _ => {
1066 if VERBOSE { print!(" | {} dropped", s.to_str(self.get_symbol_table())); }
1067 None
1068 }
1069 };
1070 sym_maybe.and_then(|s| {
1071 const REP_MASK: u32 = ruleflag::CHILD_REPEAT | ruleflag::REPEAT_PLUS | ruleflag::L_FORM;
1072 const CHILD_STAR: u32 = ruleflag::CHILD_REPEAT | ruleflag::L_FORM;
1073 let has_value = self.sym_has_value(&s);
1074 if has_value
1075 || matches!(s, Symbol::NT(v) if v != *var_id && self.parsing_table.flags[v as usize] & REP_MASK == CHILD_STAR)
1078 {
1079 if !has_value {
1080 has_sep_list_child_without_value = true;
1081 }
1082 Some(s)
1083 } else {
1084 None
1085 }
1086 })
1087 }).to_vec();
1088 if has_sep_list_child_without_value {
1090 alts_to_revisit.insert(*alt_id);
1092 }
1093 let parent_is_rrec_lfact = !is_ambig && self.nt_has_all_flags(g[0], ruleflag::R_RECURSION | ruleflag::PARENT_L_FACTOR);
1095 if parent_is_rrec_lfact {
1096 if flags & ruleflag::CHILD_L_FACT != 0 && self.nt_has_all_flags(g[0], ruleflag::L_FORM) {
1097 assert!(!self.nt_has_all_flags(*var_id, ruleflag::CHILD_L_FACT | ruleflag::L_FORM), "this was useful after all");
1098 if VERBOSE { print!(" child_rrec_lform_lfact"); }
1099 items[*alt_id as usize].insert(0, Symbol::NT(g[0]));
1100 }
1101 } else {
1102 let sym_maybe = if flags & ruleflag::CHILD_REPEAT != 0 && (self.nt_value[*var_id as usize] || flags & ruleflag::L_FORM != 0) {
1103 Some(Symbol::NT(*var_id))
1104 } else if !is_ambig && flags & ruleflag::CHILD_L_RECURSION != 0 {
1105 let parent = info.parent[*var_id as usize].unwrap();
1106 Some(Symbol::NT(parent))
1107 } else if !is_ambig && flags & (ruleflag::R_RECURSION | ruleflag::L_FORM) == ruleflag::R_RECURSION | ruleflag::L_FORM {
1108 Some(Symbol::NT(*var_id))
1109 } else {
1110 None
1111 };
1112 if let Some(s) = sym_maybe {
1113 if self.sym_has_value(&s) {
1114 if VERBOSE { print!(" | loop => {}", s.to_str(self.get_symbol_table())); }
1115 values.insert(0, s);
1116 }
1117 }
1118 }
1119 if VERBOSE {
1120 println!(" ==> [{}] + [{}]",
1121 items[*alt_id as usize].iter().map(|s| s.to_str(self.get_symbol_table())).join(" "),
1122 values.iter().map(|s| s.to_str(self.get_symbol_table())).join(" "));
1123 }
1124 if let Some(OpCode::NT(nt)) = opcode.first() {
1125 let backup = if matches!(values.last(), Some(Symbol::NT(x)) if x == nt) {
1127 Some(values.pop().unwrap())
1128 } else {
1129 None
1130 };
1131 if nt != var_id && self.nt_has_all_flags(*nt, ruleflag::CHILD_L_RECURSION) {
1132 if VERBOSE { println!(" CHILD_L_RECURSION"); }
1133 items[*alt_id as usize].extend(values);
1135 continue;
1136 }
1137 if flags & ruleflag::PARENT_L_FACTOR != 0 {
1138 if VERBOSE {
1139 println!(" PARENT_L_FACTOR: moving {} to child {}",
1140 values.iter().map(|s| s.to_str(self.get_symbol_table())).join(" "),
1141 Symbol::NT(*nt).to_str(self.get_symbol_table()));
1142 }
1143 let pre = &mut items[*alt_id as usize];
1145 if !pre.is_empty() {
1146 values.splice(0..0, std::mem::take(pre));
1148 }
1149 for a_id in self.var_alts[*nt as usize].iter() {
1150 items[*a_id as usize].extend(values.clone());
1151 }
1157 continue;
1158 }
1159 if let Some(sym) = backup {
1160 values.push(sym);
1161 }
1162 }
1163 items[*alt_id as usize].extend(values);
1164 } }
1166
1167 self.check_sep_list(&mut items);
1170
1171 for alt_id in alts_to_revisit {
1173 items[alt_id as usize].retain(|s| self.sym_has_value(s));
1174 }
1175 self.item_ops = items;
1176
1177 self.log.add_note(
1178 format!(
1179 "NT with value: {}",
1180 self.nt_value.iter().index()
1181 .filter(|&(_, val)| *val)
1182 .map(|(var, _)| Symbol::NT(var).to_str(self.get_symbol_table()))
1183 .join(", ")));
1184 }
1185
1186 fn check_sep_list(&mut self, items: &mut [Vec<Symbol>]) {
1190 const VERBOSE: bool = false;
1228 if VERBOSE {
1229 let log = std::mem::take(&mut self.log);
1230 self.item_ops = items.iter().cloned().to_vec();
1231 self.log_nt_info();
1232 self.log_alt_info();
1233 println!("{}", self.log);
1234 self.item_ops.clear();
1235 self.log = log;
1236 }
1237 self.log.add_note("- determining sep_list nonterminals...");
1238 if VERBOSE { println!("check_sep_list:"); }
1239 for (top_nt, g) in self.nt_parent.iter().enumerate().filter(|va| !va.1.is_empty()) {
1241 let candidate_children = g.iter()
1243 .filter_map(|&var| {
1244 let alts = &self.var_alts[var as usize];
1245 let flags = self.parsing_table.flags[var as usize];
1246 if alts.len() == 2 && flags & (ruleflag::CHILD_REPEAT | ruleflag::REPEAT_PLUS) == ruleflag::CHILD_REPEAT {
1248 Some((var, alts[0] as usize, flags))
1249 } else {
1250 None
1251 }
1252 })
1253 .to_vec(); for &(c_var, c_alt_id, _c_flags) in &candidate_children {
1255 let has_value = self.nt_value[c_var as usize];
1256 let skip_loop_nt = if has_value { 1 } else { 0 }; let mut pattern = items[c_alt_id].iter().skip(skip_loop_nt).cloned().to_vec();
1258 if VERBOSE {
1259 println!(
1260 "? {} {c_alt_id}: pattern = {}",
1261 Symbol::NT(c_var).to_str(self.get_symbol_table()),
1262 pattern.iter().map(|s| s.to_str(self.get_symbol_table())).join(" ")); }
1263 if !pattern.is_empty() {
1264 let pattern_len = pattern.len();
1265 let pattern_copy = pattern.clone();
1266 let c_sym = Symbol::NT(c_var);
1267 let (p_var, _p_alt_id, p_alt, mut p_pos) = self.nt_parent[top_nt].iter()
1269 .flat_map(|&p_var| &self.var_alts[p_var as usize])
1270 .filter_map(|&p_alt_id| {
1271 let (p_var, p_alt) = &self.parsing_table.alts[p_alt_id as usize];
1272 if *p_var != c_var {
1273 p_alt.v.iter().position(|s| s == &c_sym).map(|p_pos| (*p_var, p_alt_id as usize, p_alt, p_pos))
1274 } else {
1275 None
1276 }
1277 })
1278 .next()
1279 .unwrap_or_else(|| panic!("NT {c_var} alt {c_alt_id} should have a parent's alt that includes it"));
1280 if p_pos > 0 {
1281 p_pos -= 1; let c_alt = &self.parsing_table.alts[c_alt_id].1.v;
1284 let mut c_pos = c_alt.len() - 2; let p_pos0 = p_pos;
1286 let mut span_nbr = 0;
1287 while !pattern.is_empty() {
1288 if p_alt[p_pos] == c_alt[c_pos] {
1289 span_nbr += 1;
1290 if self.sym_has_value(&c_alt[c_pos]) {
1291 pattern.pop();
1292 }
1293 if c_pos == 0 || p_pos == 0 {
1294 break;
1295 }
1296 c_pos -= 1;
1297 p_pos -= 1;
1298 } else {
1299 break;
1300 }
1301 }
1302 if pattern.is_empty() {
1303 let exit_alts = self.gather_alts(p_var);
1304 let mut found_pos = vec![];
1310 let all_match = exit_alts.into_iter().all(|a| {
1311 let a_items = &items[a as usize];
1312 if let Some(p) = a_items.iter().position(|s| *s == c_sym) {
1313 if p >= pattern_len && a_items[p - pattern_len..p] == pattern_copy {
1315 found_pos.push((a as usize, p));
1316 true
1317 } else {
1318 false
1320 }
1321 } else {
1322 true
1323 }
1324 });
1325 if all_match {
1326 if VERBOSE {
1327 println!("- match:");
1328 println!(" c[{c_alt_id}]: {} items: {}",
1329 c_alt.iter().map(|s| s.to_str_quote(self.get_symbol_table())).join(" "),
1330 items[c_alt_id].iter().map(|s| s.to_str_quote(self.get_symbol_table())).join(" "));
1331 }
1332 for (p_alt_id, pos) in found_pos {
1333 if VERBOSE {
1334 println!(" p[{p_alt_id}]: {} items: {}",
1335 p_alt.iter().map(|s| s.to_str_quote(self.get_symbol_table())).join(" "),
1336 items[p_alt_id].iter().map(|s| s.to_str_quote(self.get_symbol_table())).join(" "));
1337 println!(
1338 " c_alt_id = {c_alt_id}, p_alt_id = {p_alt_id}, p_pos0 = {p_pos0}, span_nbr = {span_nbr}, pos = {pos} => remove [{}..{}]",
1339 pos - pattern_len, pos);
1340 }
1341 self.span_nbrs[p_alt_id] -= span_nbr as SpanNbr;
1342 self.span_nbrs_sep_list.insert(c_alt_id as AltId, span_nbr as SpanNbr);
1343 items[p_alt_id].drain(pos - pattern_len..pos);
1344 if VERBOSE {
1345 println!(" => p items: {}", items[p_alt_id].iter().map(|s| s.to_str_quote(self.get_symbol_table())).join(" "));
1346 }
1347 self.parsing_table.flags[c_var as usize] |= ruleflag::SEP_LIST;
1348 }
1349 }
1350 }
1351 }
1352 }
1353 }
1354 }
1355 }
1356
1357 fn sort_alt_ids(&self, top_nt: VarId, alts: &[AltId]) -> Vec<AltId> {
1358 const VERBOSE: bool = false;
1359 if VERBOSE {
1360 println!(" sorting {} alts {alts:?}", Symbol::NT(top_nt).to_str(self.get_symbol_table()));
1361 for &a_id in alts {
1362 let &(_nt, ref alt) = &self.parsing_table.alts[a_id as usize];
1363 if let Some((v, id)) = alt.origin {
1364 let tree = &self.origin.trees[v as usize];
1365 println!(" [{a_id}] id = {},{id} -> {} <-> {}",
1366 Symbol::NT(v).to_str(self.get_symbol_table()),
1367 crate::grammar::grtree_to_str_ansi(tree, None, Some(id), Some(v), self.get_symbol_table(), false),
1368 tree.to_str_index(None, self.get_symbol_table())
1369 );
1370 assert_eq!(v, top_nt, "v = {}, top_nt = {}", Symbol::NT(v).to_str(self.get_symbol_table()), Symbol::NT(top_nt).to_str(self.get_symbol_table()));
1371 }
1372 }
1373 }
1374 let mut sorted = vec![];
1375 let mut ids = alts.iter().filter_map(|&alt_id| self.parsing_table.alts[alt_id as usize].1.origin.map(|(_var, id)| (id, alt_id)))
1376 .collect::<HashMap<_, _>>();
1377 let tree = &self.origin.trees[top_nt as usize];
1378 for node in tree.iter_post_depth() {
1379 if let Some((_, alt_id)) = ids.remove_entry(&node.index) {
1380 sorted.push(alt_id);
1381 }
1382 }
1383 if VERBOSE { println!(" -> {sorted:?}"); }
1384 sorted
1385 }
1386
1387 fn get_type_info(&mut self) {
1444 const VERBOSE: bool = false;
1445
1446 self.log.add_note("- determining item_info...");
1447 let pinfo = &self.parsing_table;
1448 let mut nt_upper_fixer = NameFixer::new();
1449 let mut nt_lower_fixer = NameFixer::new();
1450 let mut nt_plower_fixer = NameFixer::new_empty(); let nt_name: Vec<(String, String, String)> = (0..pinfo.num_nt).map(|v| {
1452 let name = self.symbol_table.get_nt_name(v as VarId);
1453 let nu = nt_upper_fixer.get_unique_name(name.to_camelcase());
1454 let nl = nt_lower_fixer.get_unique_name(nu.to_underscore_lowercase());
1455 let npl = nt_plower_fixer.get_unique_name(nu.to_underscore_lowercase());
1456 (nu, nl, npl)
1457 }).to_vec();
1458
1459 let mut alt_info: Vec<Option<(VarId, String)>> = vec![None; pinfo.alts.len()];
1460 let mut nt_repeat = HashMap::<VarId, Vec<ItemInfo>>::new();
1461 let mut item_info: Vec<Vec<ItemInfo>> = vec![vec![]; pinfo.alts.len()];
1462 let mut child_repeat_endpoints = HashMap::<VarId, Vec<AltId>>::new();
1463 for group in self.nt_parent.iter().filter(|vf| !vf.is_empty()) {
1464 let is_ambig = self.nt_has_any_flags(group[0], ruleflag::PARENT_AMBIGUITY);
1465 let mut is_ambig_1st_child = is_ambig;
1466 let mut alt_info_to_sort = HashMap::<VarId, Vec<AltId>>::new();
1467 for var in group {
1468 let nt = *var as usize;
1469 let nt_flags = pinfo.flags[nt];
1470 if is_ambig && (nt_flags & ruleflag::PARENT_L_RECURSION != 0 || (nt_flags & ruleflag::CHILD_L_RECURSION != 0 && !is_ambig_1st_child)) {
1471 continue;
1472 }
1473 if nt_flags & (ruleflag::CHILD_REPEAT | ruleflag::L_FORM) == ruleflag::CHILD_REPEAT {
1474 let is_plus = nt_flags & ruleflag::REPEAT_PLUS != 0;
1477 let mut endpoints = self.gather_alts(*var);
1478 if VERBOSE { println!("** {} endpoints: {endpoints:?} ", Symbol::NT(*var).to_str(self.get_symbol_table())); }
1479 if is_plus {
1480 endpoints = endpoints.chunks(2).map(|slice| slice[0]).to_vec();
1483 } else {
1484 endpoints.retain(|e| !pinfo.alts[*e as usize].1.is_sym_empty());
1486 }
1487 assert!(!endpoints.is_empty());
1488 let endpoints = self.sort_alt_ids(group[0], &endpoints);
1489 child_repeat_endpoints.insert(*var, endpoints);
1490 }
1491 for &alt_id in &self.var_alts[nt] {
1492 let i = alt_id as usize;
1493 if is_ambig_1st_child && pinfo.alts[i].1.is_sym_empty() {
1494 continue;
1495 }
1496 let item_ops = &self.item_ops[alt_id as usize];
1497 let mut indices = HashMap::<Symbol, (String, Option<usize>)>::new();
1502 let mut fixer = NameFixer::new();
1503 let mut owner = pinfo.alts[i].0;
1504 while let Some(parent) = pinfo.parent[owner as usize] {
1505 if pinfo.flags[owner as usize] & ruleflag::CHILD_REPEAT != 0 {
1506 break;
1511 }
1512 owner = parent;
1513 }
1514 let is_nt_child_repeat = pinfo.flags[owner as usize] & ruleflag::CHILD_REPEAT != 0;
1515 for s in item_ops {
1516 if let Some((_, c)) = indices.get_mut(s) {
1517 *c = Some(0);
1518 } else {
1519 let name = if let Symbol::NT(vs) = s {
1520 let flag = pinfo.flags[*vs as usize];
1521 if flag & ruleflag::CHILD_REPEAT != 0 {
1522 let inside_alt_id = self.var_alts[*vs as usize][0];
1523 let inside_alt = &pinfo.alts[inside_alt_id as usize].1;
1524 if false {
1525 let mut plus_name = inside_alt.symbols()[0].to_str(self.get_symbol_table()).to_underscore_lowercase();
1527 plus_name.push_str(if flag & ruleflag::REPEAT_PLUS != 0 { "_plus" } else { "_star" });
1528 plus_name
1529 } else if is_nt_child_repeat && indices.is_empty() {
1530 if flag & ruleflag::REPEAT_PLUS != 0 { "plus_acc".to_string() } else { "star_acc".to_string() }
1532 } else {
1533 if flag & ruleflag::REPEAT_PLUS != 0 { "plus".to_string() } else { "star".to_string() }
1535 }
1536 } else {
1537 nt_name[*vs as usize].clone().1
1538 }
1539 } else {
1540 s.to_str(self.get_symbol_table()).to_lowercase()
1541 };
1542 indices.insert(*s, (fixer.get_unique_name(name), None));
1543 }
1544 }
1545
1546 let has_lfact_child = nt_flags & ruleflag::PARENT_L_FACTOR != 0 &&
1549 pinfo.alts[i].1.symbols().iter().any(|s| matches!(s, &Symbol::NT(c) if pinfo.flags[c as usize] & ruleflag::CHILD_L_FACT != 0));
1550
1551 let is_hidden_repeat_child = pinfo.flags[owner as usize] & (ruleflag::CHILD_REPEAT | ruleflag::L_FORM) == ruleflag::CHILD_REPEAT;
1554
1555 let is_alt_sym_empty = self.is_alt_sym_empty(alt_id);
1557
1558 let is_duplicate = i > 0 && self.nt_has_all_flags(owner, ruleflag::CHILD_REPEAT | ruleflag::REPEAT_PLUS | ruleflag::L_FORM) &&
1561 is_alt_sym_empty;
1562
1563 let is_last_empty_iteration = (nt_flags & ruleflag::CHILD_L_RECURSION != 0
1564 || self.nt_has_all_flags(*var, ruleflag::CHILD_REPEAT | ruleflag::L_FORM)) && is_alt_sym_empty;
1565
1566 let has_context = !has_lfact_child && !is_hidden_repeat_child && !is_duplicate && !is_last_empty_iteration;
1567 if VERBOSE {
1568 println!("NT {nt}, alt {alt_id}: has_lfact_child = {has_lfact_child}, is_hidden_repeat_child = {is_hidden_repeat_child}, \
1569 is_duplicate = {is_duplicate}, is_last_empty_iteration = {is_last_empty_iteration} => has_context = {has_context}");
1570 }
1571 if has_context {
1572 alt_info_to_sort.entry(owner)
1573 .and_modify(|v| v.push(alt_id))
1574 .or_insert_with(|| vec![alt_id]);
1575 }
1576 let has_owner_value = self.nt_value[owner as usize];
1577 item_info[i] = if item_ops.is_empty() && nt_flags & ruleflag::CHILD_L_RECURSION != 0 {
1578 if has_owner_value {
1580 vec![ItemInfo {
1581 name: nt_name[owner as usize].1.clone(),
1582 sym: Symbol::NT(owner),
1583 owner,
1584 index: None,
1585 }]
1586 } else {
1587 vec![]
1588 }
1589 } else {
1590 let is_rrec_lform = self.nt_has_all_flags(owner, ruleflag::R_RECURSION | ruleflag::L_FORM);
1591 let skip = if (is_nt_child_repeat || is_rrec_lform) && has_owner_value { 1 } else { 0 };
1592 let mut infos = item_ops.iter()
1593 .skip(skip)
1594 .map(|s| {
1595 let index = if let Some((_, Some(index))) = indices.get_mut(s) {
1596 let idx = *index;
1597 *index += 1;
1598 Some(idx)
1599 } else {
1600 None
1601 };
1602 ItemInfo {
1603 name: indices[s].0.clone(),
1604 sym: *s,
1605 owner,
1606 index,
1607 }
1608 }).to_vec();
1609 if self.nt_has_all_flags(owner, ruleflag::CHILD_REPEAT | ruleflag::REPEAT_PLUS | ruleflag::L_FORM) {
1610 let last_name = fixer.get_unique_name("last_iteration".to_string());
1612 infos.push(ItemInfo {
1613 name: last_name,
1614 sym: Symbol::Empty, owner,
1616 index: None,
1617 });
1618 };
1619 if is_nt_child_repeat && !infos.is_empty() && !nt_repeat.contains_key(&owner) {
1620 nt_repeat.insert(owner, infos.clone());
1621 }
1622 infos
1623 }
1624 } if is_ambig && nt_flags & ruleflag::CHILD_L_RECURSION != 0 {
1626 is_ambig_1st_child = false;
1627 }
1628 } if VERBOSE { println!("alt_info_to_sort = {alt_info_to_sort:?}"); }
1630 for (owner, alts) in alt_info_to_sort {
1631 for (num, alt) in self.sort_alt_ids(group[0], &alts).into_iter().index_start(1) {
1632 alt_info[alt as usize] = Some((owner, format!("V{num}")));
1633 }
1634 }
1635 } if VERBOSE {
1638 println!("NT names: {}", nt_name.iter()
1639 .map(|(u, l, pl)| format!("{u}/{l}/{pl}"))
1640 .join(", "));
1641 println!("alt info:");
1642 for (alt_id, alt_names) in alt_info.iter().enumerate() {
1643 if let Some((v, name)) = alt_names {
1644 println!("- alt {alt_id}, NT {v} {}, Ctx name: {name}", Symbol::NT(*v).to_str(self.get_symbol_table()));
1645 }
1646 }
1647 println!();
1648 println!("nt_name: {nt_name:?}");
1649 println!("alt_info: {alt_info:?}");
1650 println!("item_info:");
1651 for (i, item) in item_info.iter().enumerate().filter(|(_, item)| !item.is_empty()) {
1652 println!("- {i}: {{ {} }}", item.iter()
1653 .map(|ii| format!("{}{} ({})", ii.name, ii.index.map(|i| format!("[{i}]")).unwrap_or(String::new()), ii.sym.to_str(self.get_symbol_table())))
1654 .join(", "));
1655 }
1656 println!("item_info: {item_info:?}");
1657 println!("child_repeat_endpoints: {child_repeat_endpoints:?}");
1658 }
1659 self.nt_name = nt_name;
1660 self.alt_info = alt_info;
1661 self.item_info = item_info;
1662 self.child_repeat_endpoints = child_repeat_endpoints;
1663 }
1664
1665 pub fn gen_source_code(&mut self) -> (String, String, String) {
1670 self.log.add_note("generating source code...");
1671 if !self.log.has_no_errors() {
1672 return (String::new(), String::new(), String::new());
1673 }
1674 let mut parts = vec![];
1679 if !self.headers.is_empty() {
1680 parts.push(self.headers.clone());
1681 }
1682 let mut tmp_parts = if self.gen_parser {
1683 vec![self.source_build_parser()]
1684 } else {
1685 vec![]
1686 };
1687 let (src_types, src_listener) = if self.gen_wrapper {
1688 self.make_item_ops();
1689 let (src_wrapper, src_types, src_listener) = self.source_wrapper();
1690 tmp_parts.push(src_wrapper);
1691 (
1692 indent_source(vec![src_types], self.types_indent),
1693 indent_source(vec![src_listener], self.listener_indent)
1694 )
1695 } else {
1696 (String::new(), String::new())
1697 };
1698 self.log_nt_info();
1699 self.log_alt_info();
1700 parts.push(self.source_use());
1701 parts.extend(tmp_parts);
1702 (indent_source(parts, self.indent), src_types, src_listener)
1704 }
1705
1706 pub fn try_gen_source_code(mut self) -> Result<(BufLog, String, String, String), BuildError> {
1707 let (src, src_types, src_listener) = self.gen_source_code();
1708 if self.log.has_no_errors() {
1709 Ok((self.give_log(), src, src_types, src_listener))
1710 } else {
1711 Err(BuildError::new(self.give_log(), BuildErrorSource::ParserGen))
1712 }
1713 }
1714
1715 fn source_use(&self) -> Vec<String> {
1716 self.used_libs.gen_source_code()
1717 }
1718
1719 fn source_build_parser(&mut self) -> Vec<String> {
1720 static BASE_PARSER_LIBS: [&str; 5] = [
1721 "::VarId",
1722 "::AltId",
1723 "::parser::OpCode",
1724 "::parser::Parser",
1725 "::fixed_sym_table::FixedSymTable",
1726 ];
1727 static ALT_PARSER_LIBS: [&str; 2] = [
1728 "::alt::Alternative",
1729 "::parser::Symbol",
1730 ];
1731
1732 self.log.add_note("generating build_parser source...");
1733 let num_nt = self.symbol_table.get_num_nt();
1734 let num_t = self.symbol_table.get_num_t();
1735 self.used_libs.extend(BASE_PARSER_LIBS.into_iter().map(|s| format!("{}{s}", self.lib_crate)));
1736 self.log.add_note(format!("- creating symbol tables: {num_t} terminals, {num_nt} nonterminals"));
1737 let mut src = vec![
1738 format!("const PARSER_NUM_T: usize = {num_t};"),
1739 format!("const PARSER_NUM_NT: usize = {num_nt};"),
1740 format!("static SYMBOLS_T: [(&str, Option<&str>); PARSER_NUM_T] = [{}];",
1741 self.symbol_table.get_terminals().map(|(s, os)|
1742 format!("(\"{s}\", {})", os.as_ref().map(|s| format!("Some({s:?})")).unwrap_or("None".to_string()))).join(", ")),
1743 format!("static SYMBOLS_NT: [&str; PARSER_NUM_NT] = [{}];",
1744 self.symbol_table.get_nonterminals().map(|s| format!("{s:?}")).join(", ")),
1745 format!("static ALT_VAR: [VarId; {}] = [{}];",
1746 self.parsing_table.alts.len(),
1747 self.parsing_table.alts.iter().map(|(v, _)| format!("{v}")).join(", ")),
1748 ];
1749 if self.include_alts {
1750 self.used_libs.extend(ALT_PARSER_LIBS.into_iter().map(|s| format!("{}{s}", self.lib_crate)));
1751 src.push(format!("static ALTERNATIVES: [&[Symbol]; {}] = [{}];",
1752 self.parsing_table.alts.len(),
1753 self.parsing_table.alts.iter().map(|(_, f)| format!("&[{}]", f.iter().map(symbol_to_code).join(", "))).join(", ")));
1754 }
1755 self.log.add_note(format!("- creating parsing tables: {} items, {} opcodes", self.parsing_table.table.len(), self.opcodes.len()));
1756 src.extend(vec![
1757 format!(
1758 "static PARSING_TABLE: [AltId; {}] = [{}];",
1759 self.parsing_table.table.len(),
1760 self.parsing_table.table.iter().map(|v| format!("{v}")).join(", ")),
1761 format!(
1762 "static OPCODES: [&[OpCode]; {}] = [{}];",
1763 self.opcodes.len(),
1764 self.opcodes.iter().map(|strip| format!("&[{}]", strip.iter().map(|op| format!("OpCode::{op:?}")).join(", "))).join(", ")),
1765 format!(
1766 "static INIT_OPCODES: [OpCode; {}] = [{}];",
1767 self.init_opcodes.len(),
1768 self.init_opcodes.iter().map(|op| format!("OpCode::{op:?}")).join(", ")),
1769 format!("static START_SYMBOL: VarId = {};\n", self.start),
1770 ]);
1771 if self.gen_token_enums {
1772 src.add_space();
1773 src.push("#[derive(Clone, Copy, PartialEq, Debug)]".to_string());
1774 src.push("#[repr(u16)]".to_string());
1775 src.push("pub enum Term {".to_string());
1776 let cols = self.symbol_table.get_terminals().enumerate()
1777 .map(|(t, (s, s_opt))| vec![
1778 format!(" #[doc = \"{}\"]", if let Some(so) = s_opt { format!("'{so}'") } else { "(variable)".to_string() }),
1781 format!("{s} = {t},", )])
1782 .to_vec();
1783 src.extend(columns_to_str(cols, Some(vec![16, 0])));
1784 src.push("}\n".to_string());
1785 src.push("#[derive(Clone, Copy, PartialEq, Debug)]".to_string());
1786 src.push("#[repr(u16)]".to_string());
1787 src.push("pub enum NTerm {".to_string());
1788 let cols = self.symbol_table.get_nonterminals().index()
1789 .map(|(t, s)| vec![
1790 format!(
1791 " #[doc = \"`{s}`{}\"]",
1792 if let Some(p) = self.get_nt_parent(t) {
1793 format!(", parent: `{}`", Symbol::NT(p).to_str(self.get_symbol_table()))
1794 } else {
1795 String::new()
1796 }),
1797 format!("{} = {t},", s.to_camelcase())])
1798 .to_vec();
1799 src.extend(columns_to_str(cols, Some(vec![16, 0])));
1800 src.push("}\n".to_string());
1801 src.push("pub fn get_term_name(t: TokenId) -> (&'static str, Option<&'static str>) {".to_string());
1802 src.push(" SYMBOLS_T[t as usize]".to_string());
1803 src.push("}\n".to_string());
1804 }
1805 src.extend(vec![
1806 "pub fn build_parser() -> Parser<'static> {{".to_string(),
1807 " let symbol_table = FixedSymTable::new(".to_string(),
1808 " SYMBOLS_T.into_iter().map(|(s, os)| (s.to_string(), os.map(|s| s.to_string()))).collect(),".to_string(),
1809 " SYMBOLS_NT.into_iter().map(|s| s.to_string()).collect()".to_string(),
1810 " );".to_string(),
1811 " Parser::new(".to_string(),
1812 " PARSER_NUM_NT, PARSER_NUM_T + 1,".to_string(),
1813 " &ALT_VAR,".to_string(),
1814 if self.include_alts {
1815 " ALTERNATIVES.into_iter().map(|s| Alternative::new(s.to_vec())).collect(),".to_string()
1816 } else {
1817 " Vec::new(),".to_string()
1818 },
1819 " OPCODES.into_iter().map(|strip| strip.to_vec()).collect(),".to_string(),
1820 " INIT_OPCODES.to_vec(),".to_string(),
1821 " &PARSING_TABLE,".to_string(),
1822 " symbol_table,".to_string(),
1823 " START_SYMBOL".to_string(),
1824 " )".to_string(),
1825 "}}".to_string(),
1826 ]);
1827 src
1828 }
1829
1830 fn get_info_type(&self, infos: &[ItemInfo], info: &ItemInfo) -> String {
1831 let type_name_base = match info.sym {
1832 Symbol::T(_) => "String".to_string(),
1833 Symbol::NT(vs) => self.get_nt_type(vs).to_string(),
1834 Symbol::Empty => "bool".to_string(),
1835 _ => panic!("unexpected symbol {}", info.sym)
1836 };
1837 if info.index.is_some() {
1838 let nbr = infos.iter()
1839 .map(|nfo| if nfo.sym == info.sym { nfo.index.unwrap() } else { 0 })
1840 .max().unwrap() + 1;
1841 format!("[{type_name_base}; {nbr}]")
1842 } else {
1843 type_name_base
1844 }
1845 }
1846
1847 fn source_infos(&self, infos: &[ItemInfo], add_pub: bool, add_type: bool) -> String {
1849 let pub_str = if add_pub { "pub " } else { "" };
1850 infos.iter()
1851 .filter_map(|info| {
1852 if info.index.is_none() || info.index == Some(0) {
1853 let type_name = if add_type {
1854 format!(": {}", self.get_info_type(infos, info))
1855 } else {
1856 String::new()
1857 };
1858 Some(format!("{pub_str}{}{type_name}", info.name))
1859 } else {
1860 None
1861 }
1862 }).join(", ")
1863 }
1864
1865 fn is_alt_sym_empty(&self, a_id: AltId) -> bool {
1866 self.parsing_table.alts[a_id as usize].1.is_sym_empty()
1867 }
1868
1869 fn make_match_choices(&self, alts: &[AltId], name: &str, flags: u32, no_method: bool, force_id: Option<AltId>) -> (bool, Vec<String>) {
1871 assert!(!alts.is_empty(), "alts cannot be empty");
1872 let discarded = if !no_method && flags & (ruleflag::CHILD_REPEAT | ruleflag::REPEAT_PLUS | ruleflag::L_FORM) == ruleflag::CHILD_REPEAT { 1 } else { 0 };
1877
1878 let is_plus_no_lform = flags & (ruleflag::CHILD_REPEAT | ruleflag::REPEAT_PLUS | ruleflag::L_FORM) == (ruleflag::CHILD_REPEAT | ruleflag::REPEAT_PLUS);
1882 let is_alt_id_threshold = if is_plus_no_lform { 2 } else { 1 };
1883 let is_alt_id = force_id.is_none() && alts.len() - discarded > is_alt_id_threshold;
1884
1885 let mut choices = Vec::<String>::new();
1886 let force_id_str = force_id.map(|f| f.to_string()).unwrap_or_default();
1887 if alts.len() - discarded == 1 {
1888 if no_method {
1889 choices.push(format!(" {} => {{}}", alts[0]));
1890 } else {
1891 choices.push(format!(" {} => self.{name}({force_id_str}),", alts[0]));
1892 }
1893 } else {
1894 let last = alts.len() - 1 - discarded;
1895 choices.extend((0..last).map(|i| format!(" {} |", alts[i])));
1896 if no_method {
1897 choices.push(format!(" {} => {{}}", alts[last]));
1898 } else {
1899 choices.push(format!(" {} => self.{name}({}{force_id_str}),",
1900 alts[last],
1901 if is_alt_id { "alt_id" } else { "" }));
1902 }
1903 }
1904 if discarded == 1 {
1905 choices.push(format!(" {} => {{}}", alts.last().unwrap()));
1906 }
1907 (is_alt_id, choices)
1908 }
1909
1910 fn gen_match_item<F: FnOnce() -> String>(&self, common: String, span_only: F) -> String {
1914 if self.gen_span_params {
1915 let span_code = span_only();
1916 format!("({span_code}, {common})")
1917 } else {
1918 common
1919 }
1920 }
1921
1922 fn get_var_param(item: &ItemInfo, indices: &HashMap<Symbol, Vec<String>>, non_indices: &mut Vec<String>) -> Option<String> {
1923 if let Some(index) = item.index {
1924 if index == 0 {
1925 Some(format!("{}: [{}]", item.name, indices[&item.sym].iter().rev().join(", ")))
1926 } else {
1927 None
1928 }
1929 } else {
1930 let name = non_indices.pop().unwrap();
1931 if name == item.name {
1932 Some(name)
1933 } else {
1934 Some(format!("{}: {name}", item.name))
1935 }
1936 }
1937 }
1938
1939 fn get_var_params(item_info: &[ItemInfo], skip: usize, indices: &HashMap<Symbol, Vec<String>>, non_indices: &mut Vec<String>) -> String {
1940 item_info.iter().skip(skip).filter_map(|item| {
1941 Self::get_var_param(item, indices, non_indices)
1942 }).join(", ")
1943 }
1944
1945 fn source_lets(infos: &[ItemInfo], nt_name: &[(String, String, String)], indent: &str, last_alt_id_maybe: Option<AltId>) -> (Vec<String>, String) {
1946 let mut src_let = vec![];
1947 let mut var_fixer = NameFixer::new();
1948 let mut indices = HashMap::<Symbol, Vec<String>>::new();
1949 let mut non_indices = Vec::<String>::new();
1950 for item in infos.iter().rev() {
1951 let varname = if let Some(index) = item.index {
1952 let name = var_fixer.get_unique_name(format!("{}_{}", item.name, index + 1));
1953 indices.entry(item.sym).and_modify(|v| v.push(name.clone())).or_insert(vec![name.clone()]);
1954 name
1955 } else {
1956 let name = item.name.clone();
1957 non_indices.push(name.clone());
1958 name
1959 };
1960 if item.sym.is_empty() {
1961 src_let.push(format!("{indent}let {varname} = alt_id == {};", last_alt_id_maybe.unwrap()));
1962 } else if let Symbol::NT(v) = item.sym {
1963 src_let.push(format!("{indent}let {varname} = self.stack.pop().unwrap().get_{}();", nt_name[v as usize].2));
1964 } else {
1965 src_let.push(format!("{indent}let {varname} = self.stack_t.pop().unwrap();"));
1966 }
1967 }
1968 let src_struct = Self::get_var_params(infos, 0, &indices, &mut non_indices);
1969 (src_let, src_struct)
1970 }
1971
1972 fn source_update_span(n: &str) -> Vec<String> {
1973 vec![
1974 format!(" let spans = self.stack_span.drain(self.stack_span.len() - {n} ..).collect::<Vec<_>>();"),
1975 " self.stack_span.push(spans.iter().fold(PosSpan::empty(), |acc, sp| acc + sp));".to_string(),
1976 ]
1977 }
1978
1979 fn source_child_repeat_lets(
1984 &self,
1985 endpoints: &[AltId],
1986 item_info: &[Vec<ItemInfo>],
1987 is_plus: bool,
1988 nt_name: &[(String, String, String)],
1989 fn_name: &str,
1990 nu: &str,
1991 is_init: bool,
1992 ) -> (Vec<String>, String)
1993 {
1994 let mut src_val = vec![];
2002 let val_name = if endpoints.len() > 1 {
2003 src_val.push(format!(" let {} = match alt_id {{", self.gen_match_item("val".to_string(), || "n".to_string())));
2005 for (i, &a_id) in endpoints.iter().index_start(1) {
2006 let infos = &item_info[a_id as usize];
2007 src_val.push(format!(" {a_id}{} => {{", if is_plus { format!(" | {}", a_id + 1) } else { String::new() }));
2008 let (src_let, src_struct) = Self::source_lets(infos, nt_name, " ", None);
2009 src_val.extend(src_let);
2010 let return_value = self.gen_match_item(
2011 format!("Syn{nu}Item::V{i} {{ {} }}", src_struct),
2012 || self.span_nbrs[a_id as usize].to_string());
2013 src_val.push(format!(" {return_value}"));
2014 src_val.push(" }".to_string());
2015 }
2016 src_val.push(format!(" _ => panic!(\"unexpected alt id {{alt_id}} in fn {fn_name}\"),"));
2017 src_val.push(" };".to_string());
2018 if self.gen_span_params {
2019 src_val.extend(Self::source_update_span("n"));
2020 }
2021 "val".to_string()
2022 } else {
2023 let a_id = endpoints[0];
2025 if self.gen_span_params {
2026 let span_nbr = if is_init {
2027 *self.span_nbrs_sep_list.get(&a_id).unwrap()
2028 } else {
2029 self.span_nbrs[a_id as usize]
2030 };
2031 src_val.extend(Self::source_update_span(&span_nbr.to_string()));
2032 }
2033 let infos = &item_info[a_id as usize];
2034 let (src_let, src_struct) = Self::source_lets(infos, nt_name, " ", None);
2035 src_val.extend(src_let);
2036 if infos.len() == 1 {
2037 infos[0].name.clone()
2039 } else {
2040 src_val.push(format!(" let val = Syn{nu}Item {{ {} }};", src_struct));
2042 "val".to_string()
2043 }
2044 };
2045 (src_val, val_name)
2046 }
2047
2048 fn source_wrapper(&mut self) -> (Vec<String>, Vec<String>, Vec<String>) {
2053 const VERBOSE: bool = false;
2054 const MATCH_COMMENTS_SHOW_DESCRIPTIVE_ALTS: bool = false;
2055
2056 static PARSER_LIBS: [&str; 8] = [
2057 "::VarId", "::parser::Call", "::parser::ListenerWrapper",
2058 "::AltId", "::log::Logger", "::TokenId", "::lexer::PosSpan",
2059 "::parser::Terminate"
2060 ];
2061
2062 self.log.add_note("generating wrapper source...");
2063 self.used_libs.extend(PARSER_LIBS.into_iter().map(|s| format!("{}{s}", self.lib_crate)));
2064 if self.gen_span_params {
2065 self.used_libs.add(format!("{}::lexer::PosSpan", self.lib_crate));
2066 }
2067
2068 self.get_type_info();
2069 let pinfo = &self.parsing_table;
2070
2071 for (v, name) in self.nt_name.iter().enumerate().filter(|(v, _)| self.nt_value[*v]) {
2073 let v = v as VarId;
2074 self.nt_type.entry(v).or_insert_with(|| format!("Syn{}", name.0));
2075 }
2076
2077 let mut src = vec![];
2078
2079 let mut nt_contexts = self.source_wrapper_ctx::<VERBOSE>(&mut src);
2081
2082 let (src_types, syns) = self.source_wrapper_types::<VERBOSE>(&mut src);
2084
2085 let mut exit_fixer = NameFixer::new();
2087 let mut span_init = HashSet::<VarId>::new();
2088 let src_skel = vec![
2089 format!("// {:-<80}", ""),
2090 format!("// Template for the user implementation of {}Listener", self.name),
2091 String::new(),
2092 "struct Listener {".to_string(),
2093 " log: BufLog,".to_string(),
2094 "}".to_string(),
2095 String::new(),
2096 "#[allow(unused)]".to_string(),
2097 format!("impl {}Listener for Listener {{", self.name),
2098 " fn get_log_mut(&mut self) -> &mut impl Logger {".to_string(),
2099 " &mut self.log".to_string(),
2100 " }".to_string(),
2101 String::new(),
2102 ];
2103 let mut sources = WrapperSources {
2104 src,
2105 src_listener_decl: vec![],
2106 src_skel,
2107 src_types,
2108 src_init: vec![],
2109 src_exit: vec![],
2110 src_wrapper_impl: vec![],
2111 };
2112
2113 for group in self.nt_parent.iter().filter(|vf| !vf.is_empty()) {
2115 let parent_nt = group[0] as usize;
2116 let parent_flags = self.parsing_table.flags[parent_nt];
2117 let parent_has_value = self.nt_value[parent_nt];
2118 let mut exit_alt_done = HashSet::<VarId>::new();
2119 let mut init_nt_done = HashSet::<VarId>::new();
2120 if VERBOSE { println!("- GROUP {}, parent has {}value, parent flags: {}",
2121 group.iter().map(|v| Symbol::NT(*v).to_str(self.get_symbol_table())).join(", "),
2122 if parent_has_value { "" } else { "no " },
2123 ruleflag::to_string(parent_flags).join(" | ")); }
2124 let is_ambig = parent_flags & ruleflag::PARENT_AMBIGUITY != 0;
2125 let ambig_children = if is_ambig {
2126 group.iter().filter(|&v| self.nt_has_any_flags(*v, ruleflag::CHILD_L_RECURSION)).cloned().to_vec()
2127 } else {
2128 Vec::new()
2129 };
2130 let mut ambig_op_alts = BTreeMap::<AltId, Vec<AltId>>::new();
2131 for (id, f) in ambig_children.iter() .flat_map(|v| self.gather_alts(*v))
2133 .filter_map(|f| self.parsing_table.alts[f as usize].1.get_ambig_alt_id().map(|id| (id, f)))
2134 {
2135 ambig_op_alts.entry(id).or_default().push(f);
2136 }
2137 if VERBOSE && is_ambig {
2138 println!("- ambig children vars: {}", ambig_children.iter().map(|v| Symbol::NT(*v).to_str(self.get_symbol_table())).join(", "));
2139 println!(" ambig op alts: {ambig_op_alts:?}");
2140 }
2141
2142 let in_ctx = SourceInputContext {
2144 parent_has_value,
2145 parent_nt,
2146 pinfo,
2147 syns: &syns,
2148 ambig_op_alts: &ambig_op_alts,
2149 };
2150 let mut state = SourceState {
2151 init_nt_done: &mut init_nt_done,
2152 span_init: &mut span_init,
2153 nt_contexts: &mut nt_contexts,
2154 exit_alt_done: &mut exit_alt_done,
2155 exit_fixer: &mut exit_fixer,
2156 };
2157
2158 for var in group {
2159 let nt = *var as usize;
2160 let flags = self.parsing_table.flags[nt];
2161 let is_ambig_1st_child = is_ambig && flags & ruleflag::CHILD_L_RECURSION != 0 && ambig_children.first() == Some(var);
2163 let is_ambig_redundant = is_ambig && flags & ruleflag::L_RECURSION != 0 && !is_ambig_1st_child;
2166 let has_value = self.nt_value[nt];
2167
2168 self.source_wrapper_init::<VERBOSE>(
2170 &in_ctx,
2171 *var,
2172 flags,
2173 has_value,
2174 is_ambig_1st_child,
2175 &mut state,
2176 &mut sources
2177 );
2178
2179 self.source_wrapper_exit::<VERBOSE>(
2181 &in_ctx,
2182 *var,
2183 flags,
2184 has_value,
2185 is_ambig_1st_child,
2186 is_ambig_redundant,
2187 &mut state,
2188 &mut sources
2189 );
2190 }
2191
2192 for a in group.iter().flat_map(|v| &self.var_alts[*v as usize]).filter(|a| !exit_alt_done.contains(a)) {
2194 let is_called = self.opcodes[*a as usize].contains(&OpCode::Exit(*a));
2195 let (v, alt) = &self.parsing_table.alts[*a as usize];
2196 let alt_str = if MATCH_COMMENTS_SHOW_DESCRIPTIVE_ALTS {
2197 self.full_alt_str(*a, None, false)
2198 } else {
2199 alt.to_rule_str(*v, self.get_symbol_table(), self.parsing_table.flags[*v as usize])
2200 };
2201 let comment = format!("// {alt_str} ({})", if is_called { "not used" } else { "never called" });
2202 if is_called {
2203 sources.src_exit.push(vec![format!(" {a} => {{}}"), comment]);
2204 } else {
2205 sources.src_exit.push(vec![format!(" /* {a} */"), comment]);
2206 }
2207 }
2208 let mut seg_init = Segments::from_iter(
2210 group.iter()
2211 .filter_map(|&v| if !init_nt_done.contains(&v) { Some(Seg(v as u32, v as u32)) } else { None })
2212 );
2213 seg_init.normalize();
2214 for seg in seg_init {
2215 let Seg(a, b) = seg;
2216 if a == b {
2217 sources.src_init.push(vec![format!(" {a} => {{}}"), format!("// {}", Symbol::NT(a as VarId).to_str(self.get_symbol_table()))]);
2218 } else {
2219 sources.src_init.push(vec![
2220 format!(" {a}{}{b} => {{}}", if b == a + 1 { " | " } else { " ..= " }),
2221 format!("// {}", (a..=b).map(|v| Symbol::NT(v as VarId).to_str(self.get_symbol_table())).join(", "))
2222 ]);
2223 }
2224 }
2225 }
2226
2227 self.source_wrapper_finalize(span_init, sources)
2228 }
2229
2230 fn source_wrapper_ctx<const VERBOSE: bool>(&self, src: &mut Vec<String>) -> Vec<Option<Vec<AltId>>> {
2232 let mut nt_contexts: Vec<Option<Vec<AltId>>> = vec![None; self.parsing_table.num_nt];
2233 for group in self.nt_parent.iter().filter(|vf| !vf.is_empty()) {
2234 let mut group_names = HashMap::<VarId, Vec<AltId>>::new();
2235 for nt in group {
2237 for &alt_id in &self.var_alts[*nt as usize] {
2238 if let Some((owner, _name)) = &self.alt_info[alt_id as usize] {
2239 group_names.entry(*owner)
2240 .and_modify(|v| v.push(alt_id))
2241 .or_insert_with(|| vec![alt_id]);
2242 }
2243 }
2244 }
2245 if VERBOSE {
2246 println!("group {}", group.iter().map(|nt| Symbol::NT(*nt).to_str(self.get_symbol_table())).join(" "));
2247 }
2248 for &nt in group {
2249 if let Some(alts) = group_names.get(&nt) {
2250 let flags = self.parsing_table.flags[nt as usize];
2251 if VERBOSE {
2252 print!("- {}: flags {}", Symbol::NT(nt).to_str(self.get_symbol_table()), ruleflag::to_string(flags).join(" "));
2253 if let Some(gn) = group_names.get(&nt) {
2254 println!(", alts = {}", gn.iter().map(|a| a.to_string()).join(", "));
2255 let sorted = self.sort_alt_ids(group[0], gn);
2256 println!(" sorted alts: {sorted:?}");
2257 } else {
2258 println!();
2259 }
2260 }
2261 if flags & (ruleflag::SEP_LIST | ruleflag::L_FORM) == ruleflag::SEP_LIST | ruleflag::L_FORM {
2262 src.push("#[derive(Debug)]".to_string());
2263 src.push(format!("pub enum InitCtx{} {{", self.nt_name[nt as usize].0));
2264 let a_id = self.var_alts[nt as usize][0];
2265 let comment = format!(
2266 "value of `{}` before {}",
2267 self.item_ops[a_id as usize][1..].iter().map(|s| s.to_str(self.get_symbol_table())).join(" "),
2268 self.full_alt_components(a_id, None).1
2269 );
2270 let ctx_content = self.source_infos(&self.item_info[a_id as usize], false, true);
2271 src.push(format!(" /// {comment}"));
2272 let a_name = &self.alt_info[a_id as usize].as_ref().unwrap().1;
2273 let ctx_item = if ctx_content.is_empty() {
2274 if VERBOSE { println!(" {a_name},"); }
2275 format!(" {a_name},", )
2276 } else {
2277 if VERBOSE { println!(" {a_name} {{ {ctx_content} }},"); }
2278 format!(" {a_name} {{ {ctx_content} }},", )
2279 };
2280 src.push(ctx_item);
2281 src.push("}".to_string());
2282 }
2283 src.push("#[derive(Debug)]".to_string());
2284 src.push(format!("pub enum Ctx{} {{", self.nt_name[nt as usize].0));
2285 if VERBOSE { println!(" context Ctx{}:", self.nt_name[nt as usize].0); }
2286 let alts = self.sort_alt_ids(group[0], alts);
2287 nt_contexts[nt as usize] = Some(alts.clone());
2288 for a_id in alts {
2289 let comment = self.full_alt_str(a_id, None, true);
2290 src.push(format!(" /// {comment}"));
2291 if VERBOSE { println!(" /// {comment}"); }
2292 let ctx_content = self.source_infos(&self.item_info[a_id as usize], false, true);
2293 let a_name = &self.alt_info[a_id as usize].as_ref().unwrap().1;
2294 let ctx_item = if ctx_content.is_empty() {
2295 if VERBOSE { println!(" {a_name},"); }
2296 format!(" {a_name},", )
2297 } else {
2298 if VERBOSE { println!(" {a_name} {{ {ctx_content} }},"); }
2299 format!(" {a_name} {{ {ctx_content} }},", )
2300 };
2301 src.push(ctx_item);
2302 }
2303 src.push("}".to_string());
2304 }
2305 }
2306 }
2307 nt_contexts
2308 }
2309
2310 fn source_wrapper_types<const VERBOSE: bool>(&self, src: &mut Vec<String>) -> (Vec<String>, Vec<VarId>) {
2312 static TYPE_DERIVE: &str = "#[derive(Debug, PartialEq)]";
2313
2314 let mut src_types = vec![
2315 format!("// {:-<80}", ""),
2316 "// Template for the user-defined types:".to_string(),
2317 ];
2318 src.add_space();
2319 let mut syns = Vec::<VarId>::new(); for (v, names) in self.nt_name.iter().enumerate().filter(|(v, _)| self.nt_value[*v]) {
2321 let v = v as VarId;
2322 let (nu, _nl, _npl) = names;
2323 let nt_type = self.get_nt_type(v);
2324 if self.nt_has_all_flags(v, ruleflag::CHILD_REPEAT) {
2325 let is_lform = self.nt_has_all_flags(v, ruleflag::L_FORM);
2326 let first_alt = self.var_alts[v as usize][0];
2327 let (t, var_oid) = self.origin.get(v).unwrap();
2328 if is_lform {
2329 let astr = format!("/// User-defined type for {}", self.full_alt_str(first_alt, None, true));
2330 src_types.push(String::new());
2331 src_types.push(astr.clone());
2332 src_types.push(TYPE_DERIVE.to_string());
2333 src_types.push(format!("pub struct {}();", self.get_nt_type(v)));
2334 } else {
2335 let top_parent = self.parsing_table.get_top_parent(v);
2336 src.push(format!("/// Computed `{}` array in `{} -> {}`",
2337 grtree_to_str(t, Some(var_oid), None, Some(top_parent), self.get_symbol_table(), true),
2338 Symbol::NT(top_parent).to_str(self.get_symbol_table()),
2339 grtree_to_str(t, None, Some(var_oid), Some(top_parent), self.get_symbol_table(), true),
2340 ));
2341 let endpoints = self.child_repeat_endpoints.get(&v).unwrap();
2342 if endpoints.len() > 1 {
2343 src.push("#[derive(Debug, PartialEq)]".to_string());
2345 src.push(format!("pub struct {nt_type}(pub Vec<Syn{nu}Item>);"));
2346 src.push("#[derive(Debug, PartialEq)]".to_string());
2347 src.push(format!("pub enum Syn{nu}Item {{"));
2348 for (i, &a_id) in endpoints.iter().index_start(1) {
2349 src.push(format!(" /// {}", self.full_alt_str(a_id, None, true)));
2350 src.push(format!(" V{i} {{ {} }},", self.source_infos(&self.item_info[a_id as usize], false, true)));
2351 }
2352 src.push("}".to_string());
2353 } else {
2354 let a_id = endpoints[0];
2356 let infos = &self.item_info[a_id as usize];
2357 if infos.len() == 1 {
2358 let type_name = self.get_info_type(infos, &infos[0]);
2360 src.push("#[derive(Debug, PartialEq)]".to_string());
2361 src.push(format!("pub struct {nt_type}(pub Vec<{type_name}>);", ));
2362 } else {
2363 src.push("#[derive(Debug, PartialEq)]".to_string());
2365 src.push(format!("pub struct {nt_type}(pub Vec<Syn{nu}Item>);"));
2366 src.push(format!("/// {}", self.full_alt_str(first_alt, None, false)));
2367 src.push("#[derive(Debug, PartialEq)]".to_string());
2368 src.push(format!("pub struct Syn{nu}Item {{ {} }}", self.source_infos(infos, true, true)));
2369 }
2370 }
2371 }
2372 } else {
2373 src_types.push(String::new());
2374 src_types.push(format!("/// User-defined type for `{}`", Symbol::NT(v).to_str(self.get_symbol_table())));
2375 src_types.push(TYPE_DERIVE.to_string());
2376 src_types.push(format!("pub struct {}();", self.get_nt_type(v)));
2377 }
2378 syns.push(v);
2379 }
2380 if !self.nt_value[self.start as usize] {
2381 let nu = &self.nt_name[self.start as usize].0;
2382 src.push(format!("/// Top non-terminal {nu} (has no value)"));
2383 src.push("#[derive(Debug, PartialEq)]".to_string());
2384 src.push(format!("pub struct Syn{nu}();"))
2385 }
2386
2387 if VERBOSE { println!("syns = {syns:?}"); }
2389 src.add_space();
2390 src.push("#[derive(Debug)]".to_string());
2392 src.push(format!("enum EnumSynValue {{ {} }}",
2393 syns.iter().map(|v| format!("{}({})", self.nt_name[*v as usize].0, self.get_nt_type(*v))).join(", ")));
2394 if !syns.is_empty() {
2395 src.add_space();
2397 src.push("impl EnumSynValue {".to_string());
2398 for v in &syns {
2399 let (nu, _, npl) = &self.nt_name[*v as usize];
2400 let nt_type = self.get_nt_type(*v);
2401 src.push(format!(" fn get_{npl}(self) -> {nt_type} {{"));
2402 if syns.len() == 1 {
2403 src.push(format!(" let EnumSynValue::{nu}(val) = self;"));
2404 src.push(" val".to_string());
2405 } else {
2406 src.push(format!(" if let EnumSynValue::{nu}(val) = self {{ val }} else {{ panic!() }}"));
2407 }
2408 src.push(" }".to_string());
2409 }
2410 src.push("}".to_string());
2411 }
2412 (src_types, syns)
2413 }
2414
2415 fn source_wrapper_init<const VERBOSE: bool>(
2417 &self,
2418 ctx : &SourceInputContext,
2419 var : VarId,
2420 flags : u32,
2421 has_value : bool,
2422 is_ambig_1st_child : bool,
2423 state : &mut SourceState,
2424 sources : &mut WrapperSources
2425 ) {
2426 let &SourceInputContext { ambig_op_alts, .. } = ctx;
2427 let SourceState { init_nt_done, span_init, .. } = state;
2428 let WrapperSources { src_listener_decl, src_skel, src_init, src_wrapper_impl, .. } = sources;
2429 let nt = var as usize;
2430 let sym_nt = Symbol::NT(var);
2431 let nt_comment = format!("// {}", sym_nt.to_str(self.get_symbol_table()));
2432 let is_sep_list = flags & ruleflag::SEP_LIST != 0;
2433 let is_lform = flags & ruleflag::L_FORM != 0;
2434 let is_rrec_lform = is_lform && flags & ruleflag::R_RECURSION != 0;
2435 let is_plus = flags & ruleflag::REPEAT_PLUS != 0;
2436 let (nu, nl, npl) = &self.nt_name[nt];
2437 if VERBOSE { println!(" - VAR {}, has {}value, flags: {}",
2438 sym_nt.to_str(self.get_symbol_table()),
2439 if has_value { "" } else { "no " },
2440 ruleflag::to_string(flags).join(" | ")); }
2441
2442 let mut has_skel_init = false;
2443 let init_fn_name = format!("init_{npl}");
2444 if self.parsing_table.parent[nt].is_none() {
2445 init_nt_done.insert(var);
2446 if is_rrec_lform {
2447 span_init.insert(var);
2448 }
2449 if is_rrec_lform && has_value {
2450 src_wrapper_impl.push(String::new());
2451 src_listener_decl.push(format!(" fn {init_fn_name}(&mut self) -> {};", self.get_nt_type(nt as VarId)));
2452 src_skel.push(format!(" fn {init_fn_name}(&mut self) -> {} {{", self.get_nt_type(nt as VarId)));
2453 has_skel_init = true;
2454 src_init.push(vec![format!(" {nt} => self.init_{nl}(),"), nt_comment]);
2455 src_wrapper_impl.push(format!(" fn {init_fn_name}(&mut self) {{"));
2456 src_wrapper_impl.push(format!(" let val = self.listener.init_{nl}();"));
2457 src_wrapper_impl.push(format!(" self.stack.push(EnumSynValue::{nu}(val));"));
2458 src_wrapper_impl.push(" }".to_string());
2459 } else {
2460 src_listener_decl.push(format!(" fn {init_fn_name}(&mut self) {{}}"));
2461 src_init.push(vec![format!(" {nt} => self.listener.{init_fn_name}(),"), nt_comment]);
2462 }
2463 } else if flags & ruleflag::CHILD_REPEAT != 0 {
2464 if !is_sep_list {
2465 span_init.insert(var);
2466 }
2467 if has_value || is_sep_list {
2468 init_nt_done.insert(var);
2469 src_wrapper_impl.push(String::new());
2470 src_init.push(vec![format!(" {nt} => self.{init_fn_name}(),"), nt_comment]);
2471 src_wrapper_impl.push(format!(" fn {init_fn_name}(&mut self) {{"));
2472 if is_lform {
2473 if is_sep_list {
2474 let all_exit_alts = if is_ambig_1st_child {
2475 ambig_op_alts.values().rev().map(|v| v[0]).to_vec()
2476 } else {
2477 self.gather_alts(nt as VarId)
2478 };
2479 let exit_alts = all_exit_alts.into_iter()
2480 .filter(|f|
2481 (flags & ruleflag::CHILD_L_RECURSION == 0
2482 && flags & (ruleflag::CHILD_REPEAT_LFORM | ruleflag::REPEAT_PLUS) != ruleflag::CHILD_REPEAT_LFORM)
2483 || !self.is_alt_sym_empty(*f)
2484 );
2485 let (mut last_alt_ids, exit_info_alts): (Vec<AltId>, Vec<AltId>) = exit_alts.into_iter()
2486 .partition(|i| self.alt_info[*i as usize].is_none());
2487 let last_alt_id_maybe = if last_alt_ids.is_empty() { None } else { Some(last_alt_ids.remove(0)) };
2488 let a = exit_info_alts[0];
2489 let indent = " ";
2490 let (src_let, ctx_params) = Self::source_lets(&self.item_info[a as usize], &self.nt_name, indent, last_alt_id_maybe);
2491 src_wrapper_impl.extend(src_let);
2492 let ctx = if ctx_params.is_empty() {
2493 format!("InitCtx{nu}::{}", self.alt_info[a as usize].as_ref().unwrap().1)
2494 } else {
2495 format!("InitCtx{nu}::{} {{ {ctx_params} }}", self.alt_info[a as usize].as_ref().unwrap().1)
2496 };
2497 src_wrapper_impl.push(format!(" let ctx = {ctx};"));
2498 if self.gen_span_params {
2499 src_wrapper_impl.extend(Self::source_update_span(&self.span_nbrs_sep_list[&a].to_string()));
2500 }
2501 src_wrapper_impl.push(format!(
2502 " {}self.listener.{init_fn_name}(ctx{});",
2503 if has_value { "let val = " } else { "" },
2504 if self.gen_span_params { ", spans" } else { "" }));
2505 let ret = if has_value {
2506 format!("-> {};", self.get_nt_type(nt as VarId))
2507 } else {
2508 src_listener_decl.push(" #[allow(unused_variables)]".to_string());
2509 "{}".to_string()
2510 };
2511 src_listener_decl.push(format!(
2512 " fn {init_fn_name}(&mut self, ctx: InitCtx{nu}{}) {ret}",
2513 if self.gen_span_params { ", spans: Vec<PosSpan>" } else { "" }));
2514
2515 let ret = if has_value { format!(" -> {}", self.get_nt_type(nt as VarId)) } else { String::new() };
2517 src_skel.push(format!(
2518 " fn {init_fn_name}(&mut self, ctx: InitCtx{nu}{}){ret} {{",
2519 if self.gen_span_params { ", spans: Vec<PosSpan>" } else { "" }));
2520 let a_id = self.var_alts[nt][0];
2521 let a_info = &self.item_info[a_id as usize];
2522 if !a_info.is_empty() {
2523 let comment = format!(
2524 "value of `{}` before {}",
2525 self.item_ops[a_id as usize][1..].iter().map(|s| s.to_str(self.get_symbol_table())).join(" "),
2526 self.full_alt_components(a_id, None).1
2527 );
2528 let ctx_content = a_info.iter().map(|i| i.name.clone()).join(", ");
2529 let a_name = &self.alt_info[a_id as usize].as_ref().unwrap().1;
2530 src_skel.push(format!(" // {comment}"));
2531 src_skel.push(format!(" let InitCtx{nu}::{a_name} {{ {ctx_content} }} = ctx;"));
2532 }
2533 has_skel_init = true;
2534 } else {
2535 src_wrapper_impl.push(format!(" let val = self.listener.{init_fn_name}();"));
2536 src_listener_decl.push(format!(" fn {init_fn_name}(&mut self) -> {};", self.get_nt_type(nt as VarId)));
2537 src_skel.push(format!(" fn {init_fn_name}(&mut self) -> {} {{", self.get_nt_type(nt as VarId)));
2538 has_skel_init = true;
2539 }
2540 if has_value {
2541 src_wrapper_impl.push(format!(" self.stack.push(EnumSynValue::{nu}(val));"));
2542 }
2543 } else if is_sep_list {
2544 let endpoints = self.child_repeat_endpoints.get(&var).unwrap();
2547 let (src_val, val_name) = self.source_child_repeat_lets(endpoints, &self.item_info, is_plus, &self.nt_name, &init_fn_name, nu, true);
2548 src_wrapper_impl.extend(src_val);
2549 src_wrapper_impl.push(format!(" self.stack.push(EnumSynValue::{nu}(Syn{nu}(vec![{val_name}])));"));
2550 } else {
2551 src_wrapper_impl.push(format!(" let val = Syn{nu}(Vec::new());"));
2552 src_wrapper_impl.push(format!(" self.stack.push(EnumSynValue::{nu}(val));"));
2553 }
2554 src_wrapper_impl.push(" }".to_string());
2555 } else if is_lform {
2556 init_nt_done.insert(var);
2557 src_init.push(vec![format!(" {nt} => self.listener.{init_fn_name}(),"), nt_comment]);
2558 src_listener_decl.push(format!(" fn {init_fn_name}(&mut self) {{}}"));
2559 } else {
2560 }
2562 } else {
2563 }
2565 if has_skel_init {
2566 if has_value {
2567 src_skel.push(format!(" {}()", self.get_nt_type(nt as VarId)));
2568 }
2569 src_skel.push(" }".to_string());
2570 src_skel.push(String::new());
2571 }
2572 }
2573
2574 fn source_wrapper_exit<const VERBOSE: bool>(
2576 &self,
2577 ctx : &SourceInputContext,
2578 var : VarId,
2579 flags : u32,
2580 has_value : bool,
2581 is_ambig_1st_child : bool,
2582 is_ambig_redundant : bool,
2583 state : &mut SourceState,
2584 sources : &mut WrapperSources
2585 ) {
2586 const MATCH_COMMENTS_SHOW_DESCRIPTIVE_ALTS: bool = false;
2587
2588 let &SourceInputContext {
2589 parent_has_value, parent_nt, pinfo, syns, ambig_op_alts
2590 } = ctx;
2591 let SourceState { nt_contexts, exit_alt_done, exit_fixer, .. } = state;
2592 let WrapperSources { src_listener_decl, src_skel, src_exit, src_wrapper_impl, .. } = sources;
2593 let nt = var as usize;
2594 let is_plus = flags & ruleflag::REPEAT_PLUS != 0;
2595 let is_parent = nt == parent_nt;
2596 let is_child_repeat_lform = self.nt_has_all_flags(var, ruleflag::CHILD_REPEAT_LFORM);
2597 let (nu, _nl, npl) = &self.nt_name[nt];
2598
2599 if !is_ambig_redundant && flags & ruleflag::CHILD_L_FACT == 0 {
2601 let mut has_skel_exit = false;
2602 let mut has_skel_exit_return = false;
2603 let (pnu, _pnl, pnpl) = &self.nt_name[parent_nt];
2604 if VERBOSE { println!(" {nu} (parent {pnu})"); }
2605 let no_method = !has_value && flags & ruleflag::CHILD_REPEAT_LFORM == ruleflag::CHILD_REPEAT;
2606 let is_rrec_lform = self.nt_has_all_flags(var, ruleflag::R_RECURSION | ruleflag::L_FORM);
2607 let (fnpl, fnu, fnt, f_valued) = if is_ambig_1st_child {
2608 (pnpl, pnu, parent_nt, parent_has_value) } else {
2610 (npl, nu, nt, has_value)
2611 };
2612 if is_parent || (is_child_repeat_lform && !no_method) || is_ambig_1st_child {
2613 let extra_param = if self.gen_span_params { ", spans: Vec<PosSpan>" } else { "" };
2614 if f_valued {
2615 let nt_type = self.get_nt_type(fnt as VarId);
2616 if is_rrec_lform || (is_child_repeat_lform) {
2617 src_listener_decl.push(format!(" fn exit_{fnpl}(&mut self, acc: &mut {nt_type}, ctx: Ctx{fnu}{extra_param});"));
2618 src_skel.push(format!(" fn exit_{fnpl}(&mut self, acc: &mut {nt_type}, ctx: Ctx{fnu}{extra_param}) {{"));
2619 } else {
2620 src_listener_decl.push(format!(" fn exit_{fnpl}(&mut self, ctx: Ctx{fnu}{extra_param}) -> {nt_type};"));
2621 src_skel.push(format!(" fn exit_{fnpl}(&mut self, ctx: Ctx{fnu}{extra_param}) -> {nt_type} {{"));
2622 has_skel_exit_return = true;
2623 }
2624 } else {
2625 src_listener_decl.push(" #[allow(unused_variables)]".to_string());
2626 src_listener_decl.push(format!(" fn exit_{fnpl}(&mut self, ctx: Ctx{fnu}{extra_param}) {{}}"));
2627 src_skel.push(format!(" fn exit_{fnpl}(&mut self, ctx: Ctx{fnu}{extra_param}) {{"));
2628 }
2629 has_skel_exit = true;
2630 }
2631 let all_exit_alts = if is_ambig_1st_child {
2632 ambig_op_alts.values().rev().map(|v| v[0]).to_vec()
2633 } else {
2634 self.gather_alts(nt as VarId)
2635 };
2636 let (last_it_alts, exit_alts) = all_exit_alts.into_iter()
2637 .partition::<Vec<_>, _>(|f|
2638 (flags & ruleflag::CHILD_L_RECURSION != 0
2639 || flags & (ruleflag::CHILD_REPEAT_LFORM | ruleflag::REPEAT_PLUS) == ruleflag::CHILD_REPEAT_LFORM)
2640 && self.is_alt_sym_empty(*f));
2641 if VERBOSE {
2642 println!(" no_method: {no_method}, exit alts: {}", exit_alts.iter().join(", "));
2643 if !last_it_alts.is_empty() {
2644 println!(" last_it_alts: {}", last_it_alts.iter().join(", "));
2645 }
2646 }
2647
2648 if has_skel_exit {
2650 if let Some(alts) = &nt_contexts[fnt] {
2651 let mut skel_ctx = vec![];
2652 for &a_id in alts {
2653 if let Some((_, variant)) = self.alt_info[a_id as usize].as_ref() {
2654 let comment = self.full_alt_str(a_id, None, false);
2655 let fields = self.source_infos(&self.item_info[a_id as usize], false, false);
2656 let ctx_content = if fields.is_empty() {
2657 String::new()
2658 } else {
2659 format!(" {{ {fields} }}")
2660 };
2661 skel_ctx.push((comment, variant, ctx_content));
2662 }
2663 }
2664 match skel_ctx.len() {
2665 0 => {}
2666 1 => {
2667 let (comment, variant, ctx_content) = skel_ctx.pop().unwrap();
2668 src_skel.push(format!(" // {comment}"));
2669 src_skel.push(format!(" let Ctx{fnu}::{variant}{ctx_content} = ctx;"));
2670 }
2671 _ => {
2672 src_skel.push(" match ctx {".to_string());
2673 for (comment, variant, ctx_content) in skel_ctx {
2674 src_skel.push(format!(" // {comment}"));
2675 src_skel.push(format!(" Ctx{fnu}::{variant}{ctx_content} => {{}}"));
2676 }
2677 src_skel.push(" }".to_string());
2678 }
2679 }
2680 if has_skel_exit_return {
2681 src_skel.push(format!(" {}()", self.get_nt_type(fnt as VarId)));
2682 }
2683 src_skel.push(" }".to_string());
2684 src_skel.push(String::new());
2685 } else {
2686 panic!("no alts for NT {fnpl} [{fnt}]");
2687 }
2688 }
2689
2690 for f in &exit_alts {
2691 exit_alt_done.insert(*f);
2692 }
2693 let inter_or_exit_name = if flags & ruleflag::PARENT_L_RECURSION != 0 { format!("inter_{npl}") } else { format!("exit_{npl}") };
2694 let fn_name = exit_fixer.get_unique_name(inter_or_exit_name.clone());
2695 let (is_alt_id, choices) = self.make_match_choices(&exit_alts, &fn_name, flags, no_method, None);
2696 if VERBOSE { println!(" choices: {}", choices.iter().map(|s| s.trim()).join(" ")); }
2697 let comments = exit_alts.iter().map(|f| {
2698 let (v, pf) = &self.parsing_table.alts[*f as usize];
2699 if MATCH_COMMENTS_SHOW_DESCRIPTIVE_ALTS {
2700 format!("// {}", self.full_alt_str(*f, None, false))
2701 } else {
2702 format!("// {}", pf.to_rule_str(*v, self.get_symbol_table(), self.parsing_table.flags[*v as usize]))
2703 }
2704 }).to_vec();
2705 src_exit.extend(choices.into_iter().zip(comments).map(|(a, b)| vec![a, b]));
2706 if is_ambig_1st_child {
2707 for (a_id, dup_alts) in ambig_op_alts.values().rev().filter_map(|v| if v.len() > 1 { v.split_first() } else { None }) {
2708 let (_, choices) = self.make_match_choices(dup_alts, &fn_name, 0, no_method, Some(*a_id));
2711 let comments = dup_alts.iter()
2712 .map(|a| {
2713 let (v, alt) = &pinfo.alts[*a as usize];
2714 format!("// {} (duplicate of {a_id})", alt.to_rule_str(*v, self.get_symbol_table(), 0))
2715 }).to_vec();
2716 src_exit.extend(choices.into_iter().zip(comments).map(|(a, b)| vec![a, b]));
2717 for a in dup_alts {
2718 exit_alt_done.insert(*a);
2719 }
2720 }
2721 }
2722 if !no_method {
2723 src_wrapper_impl.push(String::new());
2724 src_wrapper_impl.push(format!(" fn {fn_name}(&mut self{}) {{", if is_alt_id { ", alt_id: AltId" } else { "" }));
2725 }
2726 if flags & ruleflag::CHILD_REPEAT_LFORM == ruleflag::CHILD_REPEAT {
2727 if has_value {
2728 let endpoints = self.child_repeat_endpoints.get(&var).unwrap();
2729 let (src_val, val_name) = self.source_child_repeat_lets(endpoints, &self.item_info, is_plus, &self.nt_name, &fn_name, nu, false);
2730 src_wrapper_impl.extend(src_val);
2731 let vec_name = if is_plus { "plus_acc" } else { "star_acc" };
2732 src_wrapper_impl.push(format!(" let Some(EnumSynValue::{nu}(Syn{nu}({vec_name}))) = self.stack.last_mut() else {{"));
2733 src_wrapper_impl.push(format!(" panic!(\"expected Syn{nu} item on wrapper stack\");"));
2734 src_wrapper_impl.push(" };".to_string());
2735 src_wrapper_impl.push(format!(" {vec_name}.push({val_name});"));
2736 }
2737 } else {
2738 assert!(!no_method, "no_method is not expected here (only used in +* with no lform)");
2739 let (mut last_alt_ids, exit_info_alts): (Vec<AltId>, Vec<AltId>) = exit_alts.into_iter()
2740 .partition(|i| self.alt_info[*i as usize].is_none());
2741 let fnu = if is_child_repeat_lform { nu } else { pnu }; let fnpl = if is_child_repeat_lform { npl } else { pnpl }; let a_has_value = if is_child_repeat_lform { has_value } else { parent_has_value };
2744 let is_single = exit_info_alts.len() == 1;
2745 let indent = if is_single { " " } else { " " };
2746 if !is_single {
2747 if self.gen_span_params {
2748 src_wrapper_impl.push(" let (n, ctx) = match alt_id {".to_string());
2749 } else {
2750 src_wrapper_impl.push(" let ctx = match alt_id {".to_string());
2751 }
2752 }
2753 if VERBOSE { println!(" exit_alts -> {exit_info_alts:?}, last_alt_id -> {last_alt_ids:?}"); }
2754 let spans_param = if self.gen_span_params { ", spans" } else { "" };
2755 for a in exit_info_alts {
2756 if VERBOSE {
2757 println!(" - ALTERNATIVE {a}: {} -> {}",
2758 Symbol::NT(var).to_str(self.get_symbol_table()),
2759 self.parsing_table.alts[a as usize].1.to_str(self.get_symbol_table()));
2760 }
2761 let last_alt_id_maybe = if last_alt_ids.is_empty() { None } else { Some(last_alt_ids.remove(0)) };
2762 if !is_single {
2763 let last_alt_choice = if let Some(last_alt_id) = last_alt_id_maybe { format!(" | {last_alt_id}") } else { String::new() };
2764 src_wrapper_impl.push(format!(" {a}{last_alt_choice} => {{", ));
2765 }
2766 let (src_let, ctx_params) = Self::source_lets(&self.item_info[a as usize], &self.nt_name, indent, last_alt_id_maybe);
2767 src_wrapper_impl.extend(src_let);
2768 let ctx = if ctx_params.is_empty() {
2769 format!("Ctx{fnu}::{}", self.alt_info[a as usize].as_ref().unwrap().1)
2770 } else {
2771 format!("Ctx{fnu}::{} {{ {ctx_params} }}", self.alt_info[a as usize].as_ref().unwrap().1)
2772 };
2773 if is_single {
2774 src_wrapper_impl.push(format!(" let ctx = {ctx};"));
2775 if self.gen_span_params {
2776 src_wrapper_impl.extend(Self::source_update_span(&self.span_nbrs[a as usize].to_string()));
2777
2778 }
2779 } else {
2780 let ctx_value = self.gen_match_item(ctx, || self.span_nbrs[a as usize].to_string());
2781 src_wrapper_impl.push(format!("{indent}{ctx_value}"));
2782 src_wrapper_impl.push(" }".to_string());
2783 }
2784 }
2785 if !is_single {
2786 src_wrapper_impl.push(format!(" _ => panic!(\"unexpected alt id {{alt_id}} in fn {fn_name}\")"));
2787 src_wrapper_impl.push(" };".to_string());
2788 if self.gen_span_params {
2789 src_wrapper_impl.extend(Self::source_update_span("n"));
2790 }
2791 }
2792 if (is_rrec_lform | is_child_repeat_lform) && f_valued {
2793 src_wrapper_impl.push(
2794 format!(" let Some(EnumSynValue::{fnu}(acc)) = self.stack.last_mut() else {{ panic!() }};"));
2795 src_wrapper_impl.push(
2796 format!(" self.listener.exit_{fnpl}(acc, ctx{spans_param});"));
2797 } else {
2798 src_wrapper_impl.push(format!(
2799 " {}self.listener.exit_{fnpl}(ctx{spans_param});",
2800 if a_has_value { "let val = " } else { "" }));
2801 if a_has_value {
2802 src_wrapper_impl.push(format!(" self.stack.push(EnumSynValue::{fnu}(val));"));
2803 }
2804 }
2805 }
2806 if !no_method {
2807 src_wrapper_impl.push(" }".to_string());
2808 }
2809 for a in last_it_alts {
2810 assert_eq!(flags, pinfo.flags[nt]);
2811 let owner_maybe = if flags & ruleflag::CHILD_REPEAT_LFORM == ruleflag::CHILD_REPEAT_LFORM {
2814 Some(var)
2815 } else if flags & ruleflag::CHILD_L_RECURSION != 0 {
2816 pinfo.parent[nt]
2817 } else {
2818 None
2819 };
2820 if let Some(owner) = owner_maybe {
2821 if self.nt_value[owner as usize] {
2822 let (variant, _, fnname) = &self.nt_name[owner as usize];
2823 let typ = self.get_nt_type(owner);
2824 let varname = if is_child_repeat_lform { "acc" } else { fnname };
2825 if VERBOSE { println!(" exitloop{fnname}({varname}) owner = {}", Symbol::NT(owner).to_str(self.get_symbol_table())); }
2826 src_listener_decl.push(" #[allow(unused_variables)]".to_string());
2827 src_listener_decl.push(format!(" fn exitloop_{fnname}(&mut self, {varname}: &mut {typ}) {{}}"));
2828 let (v, pf) = &self.parsing_table.alts[a as usize];
2829 let alt_str = if MATCH_COMMENTS_SHOW_DESCRIPTIVE_ALTS {
2830 self.full_alt_str(a, None, false)
2831 } else {
2832 pf.to_rule_str(*v, self.get_symbol_table(), self.parsing_table.flags[*v as usize])
2833 };
2834 src_exit.push(vec![format!(" {a} => self.exitloop_{fnpl}(),"), format!("// {alt_str}")]);
2835 exit_alt_done.insert(a);
2836 src_wrapper_impl.push(String::new());
2837 src_wrapper_impl.push(format!(" fn exitloop_{fnpl}(&mut self) {{"));
2838 src_wrapper_impl.push(format!(" let EnumSynValue::{variant}({varname}) = self.stack.last_mut().unwrap(){};",
2839 if syns.len() > 1 { " else { panic!() }" } else { "" }));
2840 src_wrapper_impl.push(format!(" self.listener.exitloop_{fnname}({varname});"));
2841 src_wrapper_impl.push(" }".to_string());
2842 }
2843 }
2844 }
2845 }
2846 }
2847
2848 fn source_wrapper_finalize(&mut self, span_init: HashSet<VarId>, sources: WrapperSources) -> (Vec<String>, Vec<String>, Vec<String>) {
2849 let WrapperSources { mut src, src_listener_decl, mut src_skel, mut src_types, src_init, src_exit, src_wrapper_impl } = sources;
2850
2851 src.add_space();
2853 src.push(format!("pub trait {}Listener {{", self.name));
2854 src.push(" /// Checks if the listener requests an abort. This happens if an error is too difficult to recover from".to_string());
2855 src.push(" /// and may corrupt the stack content. In that case, the parser immediately stops and returns `ParserError::AbortRequest`.".to_string());
2856 src.push(" fn check_abort_request(&self) -> Terminate { Terminate::None }".to_string());
2857 src.push(" fn get_log_mut(&mut self) -> &mut impl Logger;".to_string());
2858 let extra_span = if self.gen_span_params { ", span: PosSpan" } else { "" };
2859 let extra_ref_span = if self.gen_span_params { ", span: &PosSpan" } else { "" };
2860 if !self.terminal_hooks.is_empty() {
2861 src.push(" #[allow(unused_variables)]".to_string());
2862 src.push(format!(" fn hook(&mut self, token: TokenId, text: &str{extra_ref_span}) -> TokenId {{ token }}"));
2863 }
2864 src.push(" #[allow(unused_variables)]".to_string());
2865 src.push(format!(" fn intercept_token(&mut self, token: TokenId, text: &str{extra_ref_span}) -> TokenId {{ token }}"));
2866 if self.nt_value[self.start as usize] || self.gen_span_params {
2867 src.push(" #[allow(unused_variables)]".to_string());
2868 }
2869 if self.nt_value[self.start as usize] {
2870 src.push(format!(" fn exit(&mut self, {}: {}{extra_span}) {{}}", self.nt_name[self.start as usize].2, self.get_nt_type(self.start)));
2871 } else {
2872 src.push(format!(" fn exit(&mut self{extra_span}) {{}}"));
2873 }
2874 src.push(" #[allow(unused_variables)]".to_string());
2875 src.push(" fn abort(&mut self, terminate: Terminate) {}".to_string());
2876 src.extend(src_listener_decl);
2884 src.push("}".to_string());
2885
2886 src.add_space();
2888 src.push("pub struct Wrapper<T> {".to_string());
2889 src.push(" verbose: bool,".to_string());
2890 src.push(" listener: T,".to_string());
2891 src.push(" stack: Vec<EnumSynValue>,".to_string());
2892 src.push(" max_stack: usize,".to_string());
2893 src.push(" stack_t: Vec<String>,".to_string());
2894 if self.gen_span_params {
2895 src.push(" stack_span: Vec<PosSpan>,".to_string());
2896 }
2897 src.push("}".to_string());
2898 src.push(String::new());
2899 src.push(format!("impl<T: {}Listener> ListenerWrapper for Wrapper<T> {{", self.name));
2900 src.push(" fn switch(&mut self, call: Call, nt: VarId, alt_id: AltId, t_data: Option<Vec<String>>) {".to_string());
2901 src.push(" if self.verbose {".to_string());
2902 src.push(" println!(\"switch: call={call:?}, nt={nt}, alt={alt_id}, t_data={t_data:?}\");".to_string());
2903 src.push(" }".to_string());
2904 src.push(" if let Some(mut t_data) = t_data {".to_string());
2905 src.push(" self.stack_t.append(&mut t_data);".to_string());
2906 src.push(" }".to_string());
2907 src.push(" match call {".to_string());
2908 src.push(" Call::Enter => {".to_string());
2909 if self.gen_span_params {
2910 let mut seg_span = Segments::from_iter(span_init.into_iter().map(|v| Seg(v as u32, v as u32)));
2912 seg_span.normalize();
2913 let pattern = seg_span.into_iter().map(|Seg(a, b)| {
2914 if a == b {
2915 a.to_string()
2916 } else if b == a + 1 {
2917 format!("{a} | {b}")
2918 } else {
2919 format!("{a} ..= {b}")
2920 }
2921 }).join(" | ");
2922 if !pattern.is_empty() {
2923 src.push(format!(" if matches!(nt, {pattern}) {{"));
2924 src.push(" self.stack_span.push(PosSpan::empty());".to_string());
2925 src.push(" }".to_string());
2926 }
2927 }
2928 src.push(" match nt {".to_string());
2929 src.extend(columns_to_str(src_init, Some(vec![64, 0])));
2935 src.push(" _ => panic!(\"unexpected enter nonterminal id: {nt}\")".to_string());
2936 src.push(" }".to_string());
2937 src.push(" }".to_string());
2938 src.push(" Call::Loop => {}".to_string());
2939 src.push(" Call::Exit => {".to_string());
2940 src.push(" match alt_id {".to_string());
2941 src.extend(columns_to_str(src_exit, Some(vec![64, 0])));
2949 src.push(" _ => panic!(\"unexpected exit alternative id: {alt_id}\")".to_string());
2950 src.push(" }".to_string());
2951 src.push(" }".to_string());
2952 src.push(" Call::End(terminate) => {".to_string());
2953 src.push(" match terminate {".to_string());
2954 src.push(" Terminate::None => {".to_string());
2955 let mut args = vec![];
2956 let (_nu, _nl, npl) = &self.nt_name[self.start as usize];
2957 if self.nt_value[self.start as usize] {
2958 src.push(format!(" let val = self.stack.pop().unwrap().get_{npl}();"));
2959 args.push("val");
2960 }
2961 if self.gen_span_params {
2962 src.push(" let span = self.stack_span.pop().unwrap();".to_string());
2963 args.push("span");
2964 }
2965 src.push(format!(" self.listener.exit({});", args.join(", ")));
2966 src.push(" }".to_string());
2967 src.push(" Terminate::Abort | Terminate::Conclude => self.listener.abort(terminate),".to_string());
2968 src.push(" }".to_string());
2969 src.push(" }".to_string());
2970 src.push(" }".to_string());
2971 src.push(" self.max_stack = std::cmp::max(self.max_stack, self.stack.len());".to_string());
2972 src.push(" if self.verbose {".to_string());
2973 src.push(" println!(\"> stack_t: {}\", self.stack_t.join(\", \"));".to_string());
2974 src.push(" println!(\"> stack: {}\", self.stack.iter().map(|it| format!(\"{it:?}\")).collect::<Vec<_>>().join(\", \"));".to_string());
2975 src.push(" }".to_string());
2976 src.push(" }".to_string());
2977 src.push(String::new());
2978 src.push(" fn check_abort_request(&self) -> Terminate {".to_string());
2979 src.push(" self.listener.check_abort_request()".to_string());
2980 src.push(" }".to_string());
2981 src.push(String::new());
2982 src.push(" fn abort(&mut self) {".to_string());
2983 src.push(" self.stack.clear();".to_string());
2984 if self.gen_span_params {
2985 src.push(" self.stack_span.clear();".to_string());
2986 }
2987 src.push(" self.stack_t.clear();".to_string());
2988 src.push(" }".to_string());
2989 src.push(String::new());
2990 src.push(" fn get_log_mut(&mut self) -> &mut impl Logger {".to_string());
2991 src.push(" self.listener.get_log_mut()".to_string());
2992 src.push(" }".to_string());
2993 if self.gen_span_params {
2994 src.push(String::new());
2995 src.push(" fn push_span(&mut self, span: PosSpan) {".to_string());
2996 src.push(" self.stack_span.push(span);".to_string());
2997 src.push(" }".to_string());
2998 }
2999 src.push(String::new());
3000 src.push(" fn is_stack_empty(&self) -> bool {".to_string());
3001 src.push(" self.stack.is_empty()".to_string());
3002 src.push(" }".to_string());
3003 src.push(String::new());
3004 src.push(" fn is_stack_t_empty(&self) -> bool {".to_string());
3005 src.push(" self.stack_t.is_empty()".to_string());
3006 src.push(" }".to_string());
3007 if self.gen_span_params {
3008 src.add_space();
3009 src.push(" fn is_stack_span_empty(&self) -> bool {".to_string());
3010 src.push(" self.stack_span.is_empty()".to_string());
3011 src.push(" }".to_string());
3012 }
3013 let unused_span = if self.gen_span_params { "" } else { "_" };
3014 let extra_span_arg = if self.gen_span_params { ", span" } else { "" };
3015 if !self.terminal_hooks.is_empty() {
3016 src.add_space();
3017 src.push(format!(" fn hook(&mut self, token: TokenId, text: &str, {unused_span}span: &PosSpan) -> TokenId {{"));
3018 src.push(format!(" self.listener.hook(token, text{extra_span_arg})"));
3019 src.push(" }".to_string());
3020 }
3021 src.add_space();
3022 src.push(format!(" fn intercept_token(&mut self, token: TokenId, text: &str, {unused_span}span: &PosSpan) -> TokenId {{"));
3023 src.push(format!(" self.listener.intercept_token(token, text{extra_span_arg})"));
3024 src.push(" }".to_string());
3025 src.push("}".to_string());
3026
3027 src.add_space();
3028 src.push(format!("impl<T: {}Listener> Wrapper<T> {{", self.name));
3029 src.push(" pub fn new(listener: T, verbose: bool) -> Self {".to_string());
3030 src.push(format!(
3031 " Wrapper {{ verbose, listener, stack: Vec::new(), max_stack: 0, stack_t: Vec::new(){} }}",
3032 if self.gen_span_params { ", stack_span: Vec::new()" } else { "" }
3033 ));
3034 src.push(" }".to_string());
3035 src.push(String::new());
3036 src.push(" pub fn get_listener(&self) -> &T {".to_string());
3037 src.push(" &self.listener".to_string());
3038 src.push(" }".to_string());
3039 src.push(String::new());
3040 src.push(" pub fn get_listener_mut(&mut self) -> &mut T {".to_string());
3041 src.push(" &mut self.listener".to_string());
3042 src.push(" }".to_string());
3043 src.push(String::new());
3044 src.push(" pub fn give_listener(self) -> T {".to_string());
3045 src.push(" self.listener".to_string());
3046 src.push(" }".to_string());
3047 src.push(String::new());
3048 src.push(" pub fn set_verbose(&mut self, verbose: bool) {".to_string());
3049 src.push(" self.verbose = verbose;".to_string());
3050 src.push(" }".to_string());
3051src.extend(src_wrapper_impl);
3071 src.push("}".to_string());
3072
3073 src_types.extend(vec![
3075 String::new(),
3076 format!("// {:-<80}", ""),
3077 ]);
3078 if let Some(line) = src_skel.last() {
3079 if line.is_empty() {
3080 src_skel.pop();
3081 }
3082 }
3083 src_skel.extend(vec![
3084 "}".to_string(),
3085 String::new(),
3086 format!("// {:-<80}", ""),
3087 ]);
3088 self.log.add_info(format!("Template for the user types:\n\n{}\n", src_types.join("\n")));
3089 self.log.add_info(format!("Template for the listener implementation:\n\n{}\n", src_skel.join("\n")));
3090
3091 (src, src_types, src_skel)
3092 }
3093}
3094
3095impl LogReader for ParserGen {
3096 type Item = BufLog;
3097
3098 fn get_log(&self) -> &Self::Item {
3099 &self.log
3100 }
3101
3102 fn give_log(self) -> Self::Item {
3103 self.log
3104 }
3105}
3106
3107impl HasBuildErrorSource for ParserGen {
3108 const SOURCE: BuildErrorSource = BuildErrorSource::ParserGen;
3109}
3110
3111impl<T> BuildFrom<ProdRuleSet<T>> for ParserGen where ProdRuleSet<LL1>: BuildFrom<ProdRuleSet<T>> {
3112 fn build_from(mut rules: ProdRuleSet<T>) -> Self {
3118 let name = rules.name.take().unwrap_or(DEFAULT_LISTENER_NAME.to_string());
3119 ParserGen::build_from_rules(rules, name)
3120 }
3121}
3122
3123impl ParserGen {
3127
3128 pub fn get_nt_tree(&self) -> VecTree<VarId> {
3129 let mut tree = VecTree::new();
3130 let root = tree.add_root(0);
3131 let mut idx = HashMap::new();
3132 for group in self.nt_parent.iter().filter(|vf| !vf.is_empty()) {
3133 idx.clear();
3134 let tree_ids = tree.add_iter(None, group.iter().cloned()).to_vec();
3144 idx.extend(group.iter().zip(tree_ids));
3145 for &child in group.iter() {
3146 tree.attach_child(
3147 self.parsing_table.parent[child as usize]
3148 .map(|p| idx[&p])
3149 .unwrap_or(root),
3150 idx[&child]);
3151 }
3152 }
3153 tree
3154 }
3155
3156 pub fn get_indented_nt(&self) -> Vec<(VarId, String)>{
3157 let tree = self.get_nt_tree();
3158 let mut indented = vec![];
3159 let mut indent = vec![];
3160 for node in tree.iter_pre_depth_simple().skip(1) {
3161 let depth = node.depth as usize;
3162 if indent.len() < depth {
3163 indent.push((1..depth).map(|i| if i & 1 == 0 { " " } else { ". " }).join(""));
3164 }
3165 indented.push((*node, format!("{}{}", &indent[depth - 1], Symbol::NT(*node).to_str(self.get_symbol_table()))));
3166 }
3167 indented
3168 }
3169
3170 pub fn nt_info_str(&self) -> Vec<String> {
3171 let indented = self.get_indented_nt();
3172 let mut cols = vec![
3173 vec![" NT".to_string(), " name".to_string(), " val".to_string(), " flags".to_string(), String::new()]];
3174 for (v, line) in indented {
3175 let nt = v as usize;
3176 cols.push(vec![
3178 format!("| {v:3}"),
3179 format!("| {line}"),
3180 if self.nt_value[nt] { "| y".to_string() } else { "|".to_string() },
3181 format!("| {}", ruleflag::to_string(self.parsing_table.flags[nt]).join(", ")),
3183 "|".to_string(),
3184 ]);
3185 }
3186 let mut txt = columns_to_str(cols, Some(vec![3, 5, 0, 0, 0]));
3187 if let Some(max) = txt.get(1).map(|s| s.charlen()) {
3188 let sep = format!("+{:-<1$}+", "", max - 2);
3189 txt.insert(1, sep.clone());
3190 txt.push(sep);
3191 }
3192 txt
3193 }
3194
3195 pub fn log_nt_info(&mut self) {
3196 let mut txt = self.nt_info_str();
3197 txt.push(String::new());
3198 self.log.add_info("nonterminal information:");
3199 self.log.extend_messages(txt.into_iter().map(LogMsg::Info));
3200 }
3201
3202 pub fn alt_info_str(&self) -> Vec<String> {
3203 let indented = self.get_indented_nt();
3204 let mut cols = vec![
3205 vec![" NT".to_string(), " alt".to_string(), " opcodes".to_string(), " spans".to_string(), " item_ops".to_string(), String::new()]];
3206 for (v, line) in indented {
3207 let nt = v as usize;
3208 for &alt_id in &self.var_alts[nt] {
3209 let a_id = alt_id as usize;
3210 let alt = &self.parsing_table.alts[a_id].1;
3211 let opcodes = self.opcodes[a_id].iter().map(|o| o.to_str_quote(self.get_symbol_table())).join(" ");
3212 let item_ops = self.item_ops.get(a_id)
3213 .map(|ops| ops.iter().map(|s| s.to_str(self.get_symbol_table())).join(" "))
3214 .unwrap_or_else(|| "-".to_string());
3215 cols.push(vec![
3216 format!("| {v:3}"),
3217 format!("| {alt_id:4}: {line} -> {}", alt.to_str(self.get_symbol_table())),
3218 format!("| {opcodes}"),
3219 format!("| {}{}",
3220 &self.span_nbrs[a_id],
3221 if let Some(ispan) = self.span_nbrs_sep_list.get(&alt_id) { format!(", {ispan}") } else { String::new() }),
3222 format!("| {item_ops}"),
3223 "|".to_string(),
3224 ]);
3225 }
3226 }
3227 let mut txt = columns_to_str(cols, Some(vec![3, 5, 0, 0, 0, 0]));
3228 if let Some(max) = txt.get(1).map(|s| s.charlen()) {
3229 let sep = format!("+{:-<1$}+", "", max - 2);
3230 txt.insert(1, sep.clone());
3231 txt.push(sep);
3232 }
3233 txt
3234 }
3235
3236 pub fn log_alt_info(&mut self) {
3237 let mut txt = self.alt_info_str();
3238 txt.push("legend: ►nt = enter nonterminal nt, ◄0 = exit alt, ●nt = loop nonterminal, Xyz! = variable terminal, \"…\" = fixed terminal, ▲ = hook".to_string());
3239 txt.push(String::new());
3240 self.log.add_note("rule alternatives:");
3241 self.log.extend_messages(txt.into_iter().map(LogMsg::Info));
3242 }
3243
3244 pub fn print_items(&self, indent: usize, show_symbols: bool, show_span: bool) {
3245 let tbl = self.get_symbol_table();
3246 let fields = (0..self.parsing_table.alts.len())
3247 .map(|a| {
3248 let a_id = a as AltId;
3249 let (v, alt) = &self.parsing_table.alts[a];
3250 let ops = &self.opcodes[a];
3251 let it = &self.item_ops[a_id as usize];
3252 let mut cols = vec![];
3253 if show_symbols {
3254 let symbols = format!("symbols![{}]", it.iter().map(|s| s.to_macro_item()).join(", "));
3255 let value = if show_span {
3256 assert!(self.gen_span_params, "ParserGen is not configured for spans");
3257 format!("({}, {symbols})", self.span_nbrs[a_id as usize])
3258 } else {
3259 symbols
3260 };
3261 cols.push(format!("{a_id} => {value},"));
3262 }
3263 cols.extend([
3264 format!("// {a_id:2}: {} -> {}", Symbol::NT(*v).to_str(tbl), alt.iter().map(|s| s.to_str_quote(tbl)).join(" ")),
3265 format!("| {}", ops.iter().map(|s| s.to_str_quote(tbl)).join(" ")),
3266 format!(
3267 "| {}{}",
3268 &self.span_nbrs[a_id as usize],
3269 if let Some(ispan) = self.span_nbrs_sep_list.get(&a_id) { format!(", {ispan}") } else { String::new() }),
3270 format!("| {}", it.iter().map(|s| s.to_str(tbl)).join(" ")),
3271 ]);
3272 cols
3273 }).to_vec();
3274 let widths = if show_symbols { vec![40, 0, 0, 0, 0] } else { vec![16, 0, 0, 0, 0] };
3275 for l in columns_to_str(fields, Some(widths)) {
3276 println!("{:indent$}{l}", "", indent = indent)
3277 }
3278 }
3279
3280 pub fn print_flags(&self, indent: usize) {
3281 let tbl: Option<&SymbolTable> = self.get_symbol_table();
3282 let prefix = format!("{:width$}//", "", width = indent);
3283 let nt_flags = self.get_parsing_table().flags.iter().index().filter_map(|(nt, &f)|
3284 if f != 0 { Some(format!("{prefix} - {}: {} ({})", Symbol::NT(nt).to_str(tbl), ruleflag::to_string(f).join(" | "), f)) } else { None }
3285 ).join("\n");
3286 let parents = self.get_parsing_table().parent.iter().index().filter_map(|(c, &par)|
3287 par.map(|p| format!("{prefix} - {} -> {}", Symbol::NT(c).to_str(tbl), Symbol::NT(p).to_str(tbl)))
3288 ).join("\n");
3289 if !nt_flags.is_empty() {
3290 println!("{prefix} NT flags:\n{nt_flags}");
3291 }
3292 if !parents.is_empty() {
3293 println!("{prefix} parents:\n{parents}");
3294 }
3295 }
3296}