Skip to main content

retex_engine/
init.rs

1//! Engine construction, IniTeX initialization and the job driver.
2
3use std::collections::HashMap;
4use std::rc::Rc;
5
6use crate::Engine;
7use crate::cmd::Cmd;
8use crate::cond::CondState;
9use crate::eqtb::{Eqtb, Meaning};
10use crate::error::{ErrorState, Interaction, JumpOut};
11use crate::files::FileState;
12use crate::fonts::{FontInfo, FontTable};
13use crate::hash::{CsId, CsTable, NO_CS};
14use crate::input::{InputState, LineSource};
15use crate::maincontrol::Mode;
16use crate::nodes::Nodes;
17use crate::primitives_gen::{self as pg, Group};
18use crate::print::{Printer, Selector};
19use crate::save::SaveStack;
20use crate::scan::ValLevel;
21use crate::token::Token;
22
23/// Well-known control sequences.
24#[derive(Clone, Copy, Debug, Default)]
25pub struct KnownCs {
26    /// `\par`.
27    pub par: CsId,
28    /// `\inaccessible` (`frozen_protection`).
29    pub frozen_protection: CsId,
30    /// Frozen `\cr`.
31    pub frozen_cr: CsId,
32    /// Frozen `\endgroup`.
33    pub frozen_end_group: CsId,
34    /// Frozen `\right`.
35    pub frozen_right: CsId,
36    /// Frozen `\fi`.
37    pub frozen_fi: CsId,
38    /// Frozen `\endtemplate` (`end_template`).
39    pub frozen_end_template: CsId,
40    /// Frozen `\endtemplate` (`endv`).
41    pub frozen_endv: CsId,
42    /// Frozen `\relax`.
43    pub frozen_relax: CsId,
44    /// `\endwrite`.
45    pub end_write: CsId,
46    /// `\notexpanded:` (`frozen_dont_expand`).
47    pub frozen_dont_expand: CsId,
48    /// Frozen `\special`.
49    pub frozen_special: CsId,
50    /// Frozen `\nullfont`.
51    pub frozen_null_font: CsId,
52    /// The marker ending a local control loop.
53    pub end_local: CsId,
54}
55
56/// Options for a run.
57#[derive(Clone, Debug)]
58pub struct Options {
59    /// IniTeX mode.
60    pub ini: bool,
61    /// Initial interaction mode.
62    pub interaction: Interaction,
63    /// `-jobname`.
64    pub job_name: Option<String>,
65    /// `-output-directory`.
66    pub output_dir: Option<std::path::PathBuf>,
67    /// The first line: the remaining command-line arguments joined.
68    pub first_line: String,
69    /// `-file-line-error`.
70    pub file_line_error: bool,
71    /// Seconds since the epoch for `\time`, `\day`, … (from
72    /// `SOURCE_DATE_EPOCH` or the clock).
73    pub epoch: Option<i64>,
74    /// `-utc`: use UTC instead of local time for the job's date.
75    pub utc: bool,
76    /// Shell escape: 0 disabled, 1 enabled, 2 restricted to the commands
77    /// in `shell_escape_commands` (LuaTeX's `status.shell_escape`).
78    pub shell_escape: u8,
79    /// `max_print_line`, `error_line`, `half_error_line`.
80    pub line_lengths: (usize, usize, usize),
81    /// `-halt-on-error`.
82    pub halt_on_error: bool,
83    /// `-output-comment`: the DVI preamble comment (else LuaTeX's dated
84    /// one).
85    pub output_comment: Option<String>,
86    /// Initial values of integer parameters, set after the format is
87    /// loaded (`-output-format` sets `\outputmode`, `-draftmode`
88    /// `\draftmode`, `-synctex` `\synctex`).
89    pub int_params: Vec<(u32, i32)>,
90}
91
92impl Default for Options {
93    fn default() -> Self {
94        Options {
95            ini: true,
96            interaction: Interaction::ErrorStop,
97            job_name: None,
98            output_dir: None,
99            first_line: String::new(),
100            file_line_error: false,
101            epoch: None,
102            utc: false,
103            shell_escape: 2,
104            line_lengths: (79, 79, 50),
105            halt_on_error: false,
106            output_comment: None,
107            int_params: Vec::new(),
108        }
109    }
110}
111
112/// Outcome of a run.
113#[derive(Clone, Copy, Debug, PartialEq, Eq)]
114pub enum Outcome {
115    /// `\end` was reached.
116    Ended,
117    /// `\dump` was reached in IniTeX mode.
118    Dumped,
119    /// The run was aborted by a fatal error.
120    Aborted,
121}
122
123impl Engine {
124    /// A fresh IniTeX engine writing to `term`.
125    pub fn new(options: &Options, term: Box<dyn std::io::Write>) -> Engine {
126        let mut printer = Printer::new(term);
127        printer.max_print_line = options.line_lengths.0;
128        printer.error_line = options.line_lengths.1;
129        printer.half_error_line = options.line_lengths.2;
130        let mut e = Engine {
131            hash: CsTable::new(),
132            eqtb: Eqtb::initex(),
133            save: SaveStack::default(),
134            input: InputState::default(),
135            printer,
136            err: ErrorState::default(),
137            cond: CondState::default(),
138            files: FileState::default(),
139            kpse: None,
140            lua: None,
141            cur_cmd: Cmd::Relax,
142            cur_chr: 0,
143            cur_cs: NO_CS,
144            cur_tok: Token(0),
145            cur_val: 0,
146            cur_glue: crate::glue::Glue::ZERO,
147            cur_toks: None,
148            cur_val_level: ValLevel::Int,
149            cs: KnownCs::default(),
150            ini_version: options.ini,
151            shell_escape: options.shell_escape,
152            format_ident: String::new(),
153            format_name: String::new(),
154            job_name: options.job_name.clone().map(String::into_bytes),
155            after_token: Token(0),
156            expand_depth: 0,
157            mode: Mode::Vertical,
158            shown_mode: Mode::None,
159            nodes: Nodes::new(),
160            nest: Vec::new(),
161            fonts: FontTable::default(),
162            long_state: Cmd::Call,
163            def_ref_cs: NO_CS,
164            runaway_list: Vec::new(),
165            last_cs_name: NO_CS,
166            frozen_prims: HashMap::new(),
167            marks: crate::typeset::Marks::default(),
168            input_read_finished: false,
169            long_help_seen: false,
170            dead_cycles: 0,
171            insert_penalties: 0,
172            last_badness: 0,
173            single_mac_param: false,
174            pack_dir: -1,
175            random_seed: 0,
176            randoms: retex_arith::random::Randoms::new(0),
177            last_x_pos: 0,
178            last_y_pos: 0,
179            mag_set: 0,
180            font_dimen_loc: None,
181            depth_threshold: 0,
182            breadth_max: 0,
183            par_token_cs: NO_CS,
184            dump_requested: false,
185            start_ints: options.int_params.clone(),
186            output_comment: options.output_comment.clone(),
187            start_time: crate::clock::StartTime::default(),
188            page: crate::typeset::PageState::default(),
189            lua_functions: HashMap::new(),
190            format_hook: None,
191            finish_hook: None,
192            deferred_writes: Vec::new(),
193            text_dirs: Vec::new(),
194            outer_text_dirs: Vec::new(),
195            outer_math_dirs: Vec::new(),
196            pending_forms: Vec::new(),
197            form_objects: Vec::new(),
198            written_sink: crate::files::WrittenSink::default(),
199            math_glues: Vec::new(),
200            math_params_initialized: false,
201            synctex: Default::default(),
202            ts: crate::typeset::TsState::default(),
203            total_pages: 0,
204            colorstack_values: HashMap::new(),
205            pdf: crate::pdfext::PdfState::default(),
206            languages: Vec::new(),
207            backend: None,
208            dvi: Default::default(),
209            local_level: 0,
210            lua_scanned_box: None,
211            lua_scan_waiting: false,
212            backend_cs: HashMap::new(),
213            pending_lua_tokens: None,
214            diagnostics: None,
215            prompt_hook: None,
216            progress_hook: None,
217            diagnostics_shown: 0,
218            prompted: false,
219            output_written: None,
220            arg_scratch: Vec::new(),
221            param_scratch: Vec::new(),
222            write_paths: Vec::new(),
223            snapshot_cs: None,
224            snapshot_hook: None,
225            edit_request: None,
226            images: Vec::new(),
227            last_image_index: 0,
228            last_image_pages: 0,
229            box_resources: Vec::new(),
230            last_box_index: 0,
231            input_line_for_lua: 0,
232        };
233        e.err.interaction = options.interaction;
234        e.err.file_line_error = options.file_line_error;
235        e.err.halt_on_error = options.halt_on_error;
236        e.files.output_dir = options.output_dir.clone();
237        e.files.command_line = options.first_line.clone().into_bytes();
238        e.init_params(options.epoch, options.utc);
239        e.init_known_cs();
240        e.fonts.fonts.push(FontInfo::null());
241        e.fonts.id_cs.push(e.cs.frozen_null_font);
242        e.push_nest(Mode::Vertical);
243        e.page.init(&mut e.nodes);
244        // The contribution list is the list of the outer vertical mode.
245        let ch = e.nest[0].head;
246        e.nodes.free(e.page.contrib_head);
247        e.page.contrib_head = ch;
248        e.ts.font_in_short_display = -1;
249        if options.ini {
250            e.install_group(Group::Tex, b"");
251            e.install_group(Group::Core, b"");
252        }
253        e.err.history = crate::error::History::Spotless;
254        e
255    }
256
257    /// Sets IniTeX parameter defaults from LuaTeX's recorded values.
258    fn init_params(&mut self, epoch: Option<i64>, utc: bool) {
259        let data = include_str!("../data/luatex-initex-params.txt");
260        for line in data.lines() {
261            let mut it = line.splitn(3, ' ');
262            let (Some(cmd), Some(name), Some(value)) = (it.next(), it.next(), it.next()) else {
263                continue;
264            };
265            let Some(p) = crate::primitives::lookup(name.as_bytes()) else {
266                continue;
267            };
268            match cmd {
269                "assign_int" => {
270                    if let Ok(v) = value.parse::<i32>() {
271                        self.eqtb.int_par[p.chr as usize].val = v;
272                    }
273                }
274                "assign_dimen" => {
275                    if let Ok(v) = value.parse::<i32>() {
276                        self.eqtb.dimen_par[p.chr as usize].val = v;
277                    }
278                }
279                _ => {}
280            }
281        }
282        // Date and time (local, or UTC with `-utc`).
283        self.start_time = crate::clock::StartTime::new(epoch, utc);
284        self.set_date_and_time();
285        self.random_seed = crate::clock::startup_random_seed();
286        self.randoms.init(self.random_seed);
287    }
288
289    /// Sets `\time`, `\day`, `\month` and `\year` to the job's start
290    /// (`fix_date_and_time`), also after a format brought its own values.
291    fn set_date_and_time(&mut self) {
292        let c = self.start_time.civil();
293        use pg::assign_int as ip;
294        self.eqtb.int_par[ip::TIME as usize].val = (c.hour * 60 + c.minute) as i32;
295        self.eqtb.int_par[ip::DAY as usize].val = c.day as i32;
296        self.eqtb.int_par[ip::MONTH as usize].val = c.month as i32;
297        self.eqtb.int_par[ip::YEAR as usize].val = c.year as i32;
298    }
299
300    fn init_known_cs(&mut self) {
301        let h = &mut self.hash;
302        let par = h.lookup_or_insert(b"par");
303        let frozen = |h: &mut CsTable, n: &str| h.new_frozen(n.as_bytes());
304        let cs = KnownCs {
305            par,
306            frozen_protection: frozen(h, "inaccessible"),
307            frozen_cr: frozen(h, "cr"),
308            frozen_end_group: frozen(h, "endgroup"),
309            frozen_right: frozen(h, "right"),
310            frozen_fi: frozen(h, "fi"),
311            frozen_end_template: frozen(h, "endtemplate"),
312            frozen_endv: frozen(h, "endtemplate"),
313            frozen_relax: frozen(h, "relax"),
314            end_write: frozen(h, "endwrite"),
315            frozen_dont_expand: frozen(h, "notexpanded:"),
316            frozen_special: frozen(h, "special"),
317            frozen_null_font: frozen(h, "nullfont"),
318            end_local: frozen(h, "endlocalcontrol"),
319        };
320        self.cs = cs;
321        let set = |e: &mut Engine, id: CsId, m: Meaning| {
322            e.eqtb.ensure_cs(id);
323            e.eqtb.cs[id.0 as usize].val = m;
324            e.eqtb.cs[id.0 as usize].level = crate::eqtb::LEVEL_ONE;
325        };
326        let prim = |name: &str| {
327            let p = crate::primitives::lookup(name.as_bytes()).expect("primitive");
328            Meaning::prim(p.cmd, p.chr)
329        };
330        set(self, cs.frozen_cr, prim("cr"));
331        set(self, cs.frozen_end_group, prim("endgroup"));
332        // Observed: the `\right` that LuaTeX inserts for a missing one
333        // acts as `\Umiddle`, so the formula stays open (`$\left($$`
334        // repeats the error until there are 100).
335        set(self, cs.frozen_right, prim("Umiddle"));
336        set(self, cs.frozen_fi, prim("fi"));
337        set(self, cs.frozen_relax, prim("relax"));
338        set(self, cs.frozen_special, prim("special"));
339        set(self, cs.frozen_null_font, Meaning::prim(Cmd::SetFont, 0));
340        set(
341            self,
342            cs.frozen_end_template,
343            Meaning::prim(Cmd::EndTemplate, 0),
344        );
345        set(self, cs.frozen_endv, Meaning::prim(Cmd::Endv, 0));
346        set(
347            self,
348            cs.frozen_dont_expand,
349            Meaning::prim(Cmd::DontExpand, 0),
350        );
351        set(self, cs.end_write, Meaning::prim(Cmd::OuterCall, 0));
352        set(self, cs.frozen_protection, Meaning::UNDEFINED);
353        set(
354            self,
355            cs.end_local,
356            Meaning::prim(Cmd::Extension, crate::misc::ext::END_LOCAL_CONTROL),
357        );
358        self.par_token_cs = par;
359    }
360
361    /// Defines all primitives of `group` with the given name prefix.
362    pub fn install_group(&mut self, group: Group, prefix: &[u8]) {
363        let prims: Vec<_> = crate::primitives::group(group).collect();
364        for p in prims {
365            self.define_primitive(prefix, p.name.as_bytes());
366        }
367    }
368
369    /// Defines primitive `name` as `prefix name`, unless that control
370    /// sequence already has a meaning (`tex.enableprimitives`). Returns
371    /// whether it was defined.
372    pub fn define_primitive(&mut self, prefix: &[u8], name: &[u8]) -> bool {
373        let Some(p) = crate::primitives::lookup(name) else {
374            return false;
375        };
376        if p.cmd == Cmd::UndefinedCs {
377            return false;
378        }
379        let mut full = prefix.to_vec();
380        full.extend_from_slice(name);
381        let cs = self.hash.lookup_or_insert(&full);
382        if self.eqtb.meaning(cs).cmd != Cmd::UndefinedCs {
383            return false;
384        }
385        self.eqtb.ensure_cs(cs);
386        self.eqtb.cs[cs.0 as usize].val = Meaning::prim(p.cmd, p.chr);
387        self.eqtb.cs[cs.0 as usize].level = crate::eqtb::LEVEL_ONE;
388        true
389    }
390
391    /// Runs a complete job: reads the first line, executes until `\end`
392    /// or `\dump`, and cleans up.
393    pub fn run(&mut self) -> Outcome {
394        let result = std::panic::catch_unwind(std::panic::AssertUnwindSafe(|| {
395            self.run_inner();
396        }));
397        match result {
398            Ok(()) => {
399                if self.dump_requested {
400                    Outcome::Dumped
401                } else {
402                    Outcome::Ended
403                }
404            }
405            Err(payload) => {
406                if payload.downcast_ref::<JumpOut>().is_some() {
407                    Outcome::Aborted
408                } else {
409                    std::panic::resume_unwind(payload);
410                }
411            }
412        }
413    }
414
415    fn run_inner(&mut self) {
416        for (i, v) in std::mem::take(&mut self.start_ints) {
417            self.eqtb.set(
418                crate::eqtb::Loc::IntPar(i),
419                crate::eqtb::Val::Int(i64::from(v)),
420                crate::eqtb::LEVEL_ONE,
421            );
422            if i == crate::primitives_gen::assign_int::SYNCTEX {
423                self.synctex.option = Some(v);
424            }
425        }
426        self.synctex_changed();
427        let banner = format!(
428            "This is retex, Version {} (LuaTeX 1.22.0 compatible){}",
429            env!("CARGO_PKG_VERSION"),
430            if self.ini_version { "  (INITEX)" } else { "" }
431        );
432        self.printer.term_raw(banner.as_bytes());
433        self.printer.term_raw(b"\n");
434        let shell = self.shell_escape_banner();
435        if !shell.is_empty() {
436            self.printer.term_raw(shell.as_bytes());
437            self.printer.term_raw(b"\n");
438        }
439        self.printer.term_offset = 0;
440        self.printer.update_terminal();
441        // The first line goes into a terminal level.
442        let first = self.files.command_line.clone();
443        self.begin_file_reading(LineSource::Terminal, Vec::new(), false);
444        if let Some(crate::input::Level::Text(t)) = self.input.stack.last_mut() {
445            t.buf = crate::utf8x::chars(&first);
446            t.loc = 0;
447            t.state = crate::input::LexState::NewLine;
448        }
449        let elc = self.eqtb.int(pg::assign_int::ENDLINECHAR);
450        if (0..=0x10_FFFF).contains(&elc)
451            && let Some(crate::input::Level::Text(t)) = self.input.stack.last_mut()
452        {
453            t.buf.push(elc as u32);
454        }
455        self.printer.new_line_char = self.eqtb.int(pg::assign_int::NEWLINECHAR);
456        self.printer.selector = if self.err.interaction == Interaction::Batch {
457            Selector::NoPrint
458        } else {
459            Selector::TermOnly
460        };
461        let starts_with_escape = first
462            .first()
463            .is_some_and(|&b| self.eqtb.catcode(u32::from(b)) == 0);
464        if !first.is_empty() && !starts_with_escape {
465            self.start_input();
466        }
467        self.history_spotless();
468        self.main_control();
469        self.final_cleanup();
470        if self.dump_requested {
471            self.store_fmt_file();
472        }
473        self.close_files_and_terminate();
474    }
475
476    fn history_spotless(&mut self) {
477        if self.err.history == crate::error::History::FatalErrorStop {
478            self.err.history = crate::error::History::Spotless;
479        }
480    }
481
482    /// Writes the format file (`store_fmt_file`, §1302).
483    pub fn store_fmt_file(&mut self) {
484        if self.save.cur_level > crate::eqtb::LEVEL_ONE {
485            self.print_err("You can't dump inside a group");
486            self.help(&["`{...\\dump}' is a no-no."]);
487            self.fatal_error("*** (cannot \\dump inside a group)");
488        }
489        // The pre_dump callback runs first; LaTeX uses it to precompile
490        // data (e.g. Unicode tables) into lua.bytecode.
491        if let Some(h) = self.lua.clone()
492            && let Err(msg) = h.call_simple(self, "pre_dump")
493        {
494            self.lua_error(&msg);
495        }
496        let job =
497            String::from_utf8_lossy(self.job_name.as_deref().unwrap_or(b"texput")).into_owned();
498        let c = self.start_time.civil();
499        self.format_ident = format!(" (format={job} {}.{}.{})", c.year, c.month, c.day);
500        self.format_name = job.clone();
501        let mut path = std::path::PathBuf::from(format!("{job}.fmt"));
502        if let Some(dir) = &self.files.output_dir {
503            path = dir.join(path);
504        }
505        self.printer.selector = if self.err.interaction == Interaction::Batch {
506            Selector::LogOnly
507        } else {
508            Selector::TermAndLog
509        };
510        self.printer.print_nl("Beginning to dump on file ");
511        self.printer.print_bytes(path.to_string_lossy().as_bytes());
512        self.printer.print_nl("");
513        let ident = self.format_ident.clone();
514        self.printer.print_str(&ident);
515        let engine_section = self.dump_engine();
516        let lua_section = match self.lua.clone() {
517            Some(h) => h.dump(self),
518            None => Vec::new(),
519        };
520        if crate::timing::enabled() {
521            eprintln!(
522                "[format] {:>10} bytes  Lua (lua.bytecode)",
523                lua_section.len()
524            );
525        }
526        let data = crate::format::assemble(&[("engine", &engine_section), ("lua", &lua_section)]);
527        self.printer.print_nl("");
528        self.printer.print_int(self.hash.len() as i64);
529        self.printer.print_str(" multiletter control sequences");
530        self.printer.print_nl("");
531        self.printer.print_int(self.fonts.fonts.len() as i64 - 1);
532        self.printer.print_str(" preloaded fonts");
533        if let Err(e) = std::fs::write(&path, &data) {
534            self.print_err("I can't write on file `");
535            self.printer.print_bytes(path.to_string_lossy().as_bytes());
536            self.printer.print_str("'.");
537            self.help_owned(vec![e.to_string()]);
538            self.error();
539        }
540    }
541
542    /// Loads a format file into a freshly created (non-IniTeX) engine.
543    pub fn load_format(&mut self, data: &[u8]) -> Result<(), crate::format::FormatError> {
544        let sections = crate::format::disassemble(data)?;
545        crate::timing::mark("format: sections checked");
546        for (name, bytes) in &sections {
547            if *name == "engine" {
548                self.undump_engine(bytes)?;
549            }
550        }
551        // The date of the job, not of the format.
552        self.set_date_and_time();
553        // Formats written by earlier versions have a frozen `\right` that
554        // acts as `\right`.
555        let p = crate::primitives::lookup(b"Umiddle").expect("primitive");
556        let id = self.cs.frozen_right;
557        self.eqtb.ensure_cs(id);
558        self.eqtb.cs[id.0 as usize].val = Meaning::prim(p.cmd, p.chr);
559        crate::timing::mark("format: engine state loaded");
560        if let Some(h) = self.lua.clone() {
561            for (name, bytes) in &sections {
562                if *name == "lua" {
563                    h.undump(self, bytes)
564                        .map_err(|_| crate::format::FormatError::Corrupt("lua section"))?;
565                }
566            }
567        }
568        crate::timing::mark("format: Lua state loaded");
569        self.format_name = self
570            .format_ident
571            .trim_start_matches(" (format=")
572            .split(' ')
573            .next()
574            .unwrap_or("")
575            .to_owned();
576        Ok(())
577    }
578
579    /// Defines a macro from a string (for tests and bootstrapping).
580    pub fn define_macro_str(&mut self, name: &str, body: &str) {
581        let cs = self.hash.lookup_or_insert(name.as_bytes());
582        let mut toks = vec![Token::END_MATCH];
583        toks.extend(crate::the::str_toks(body.as_bytes()));
584        self.eqtb.ensure_cs(cs);
585        self.eqtb.cs[cs.0 as usize].val = Meaning {
586            cmd: Cmd::Call,
587            chr: 0,
588            mac: Some(Rc::new(crate::token::Macro {
589                protected: false,
590                tokens: Rc::from(toks),
591            })),
592        };
593    }
594}