retex_engine/lib.rs
1//! The retex TeX engine: a clean-room implementation of LuaTeX's
2//! typesetting machine.
3//!
4//! The [`Engine`] holds all state of a TeX run. Its methods are spread
5//! over modules following the parts of *TeX: The Program*:
6//!
7//! * [`print`] and [`error`] — terminal and log output, error recovery;
8//! * [`hash`], [`eqtb`], [`save`] — control sequences, equivalents, grouping;
9//! * [`input`] — the input stack and the tokenizer (`get_next`);
10//! * [`expand`], [`cond`], [`scan`] — expansion, conditionals, scanning;
11//! * [`prefixed`] — assignments;
12//! * [`maincontrol`] — the chief executive.
13
14#![forbid(unsafe_code)]
15#![allow(clippy::too_many_lines)]
16
17pub mod align;
18pub mod backend;
19mod boxnote;
20pub mod build;
21pub mod clock;
22pub mod cmd;
23mod cmd_gen;
24pub mod codes;
25pub mod cond;
26pub mod diag;
27pub mod dirs;
28pub mod display;
29pub mod dvi;
30pub mod eqtb;
31pub mod error;
32pub mod expand;
33pub mod expansion;
34pub mod files;
35pub mod fonts;
36pub mod format;
37pub mod glue;
38pub mod hash;
39pub mod hmode;
40pub mod hyph;
41pub mod image;
42pub mod init;
43pub mod input;
44pub mod linebreak;
45pub mod maincontrol;
46pub mod math;
47pub mod mathdisplay;
48pub mod mathparams;
49pub mod meaning;
50pub mod misc;
51pub mod mlist;
52pub mod nodeproc;
53pub mod nodes;
54pub mod pack;
55pub mod page;
56pub mod pagenote;
57pub mod pdfext;
58pub(crate) mod pdfship;
59pub mod prefixed;
60pub mod primitives;
61pub mod primitives_gen;
62pub mod print;
63pub mod protrusion;
64pub mod save;
65pub mod scan;
66pub mod shipout;
67pub mod showbox;
68pub mod snapshot;
69pub mod synctex;
70pub mod the;
71pub mod timing;
72pub mod token;
73pub mod typeset;
74pub mod utf8x;
75
76use cmd::Cmd;
77use cond::CondState;
78use eqtb::Eqtb;
79use error::ErrorState;
80use files::FileState;
81use hash::{CsId, CsTable};
82use input::InputState;
83use print::Printer;
84use save::SaveStack;
85use token::Token;
86
87/// Callbacks from the engine into an embedded Lua interpreter.
88///
89/// The engine does not depend on Lua. `\directlua` and related primitives
90/// call this trait, which `retex-lua` implements. The host is shared
91/// (`Rc`), so nested calls (Lua → TeX → Lua) reach it too.
92pub trait LuaHost {
93 /// Runs a chunk of Lua code for `\directlua` (or `\latelua`, etc.).
94 /// `name` is the chunk name. Text printed with `tex.print` must be
95 /// pushed back into the engine's input by the host.
96 fn run(&self, engine: &mut Engine, name: &str, code: &[u8]) -> Result<(), String>;
97
98 /// Data to store in the format file (e.g. `lua.bytecode`).
99 fn dump(&self, _engine: &mut Engine) -> Vec<u8> {
100 Vec::new()
101 }
102
103 /// Restores data stored by [`dump`](Self::dump).
104 fn undump(&self, _engine: &mut Engine, _data: &[u8]) -> Result<(), String> {
105 Ok(())
106 }
107
108 /// Calls callback `name` without arguments, ignoring its results
109 /// (`pre_dump`, `stop_run`, …). Does nothing if none is registered.
110 fn call_simple(&self, _engine: &mut Engine, _name: &str) -> Result<(), String> {
111 Ok(())
112 }
113
114 /// Calls callback `name` with one string argument, ignoring its
115 /// results (`buildpage_filter`, `contribute_filter`).
116 fn call_with_string(
117 &self,
118 _engine: &mut Engine,
119 _name: &str,
120 _arg: &str,
121 ) -> Result<(), String> {
122 Ok(())
123 }
124
125 /// Whether a Lua function is registered for callback `name`.
126 fn has_callback(&self, _name: &str) -> bool {
127 false
128 }
129
130 /// Calls the `define_font` callback. Returns the id of the font it
131 /// defined, or `None` if the font could not be loaded.
132 fn define_font(&self, _engine: &mut Engine, _name: &[u8], _size: i32) -> Option<u32> {
133 None
134 }
135
136 /// Calls a node-list filter callback (`pre_linebreak_filter`,
137 /// `hpack_filter`, `post_linebreak_filter`, …) on the list starting at
138 /// `head`. Returns the new head; [`ListFilter::Flush`] means the list
139 /// must be discarded.
140 fn list_filter(
141 &self,
142 _engine: &mut Engine,
143 _callback: &str,
144 head: nodes::NodeRef,
145 _group: &str,
146 ) -> ListFilter {
147 ListFilter::Keep(head)
148 }
149
150 /// Calls the `hpack_quality` or `vpack_quality` callback `cb` for box
151 /// `b` with `incident` (overfull, underfull, loose, tight) and its
152 /// detail; returns the node the callback gave (null if none).
153 #[allow(clippy::too_many_arguments)]
154 fn pack_quality(
155 &self,
156 _engine: &mut Engine,
157 _cb: &str,
158 _incident: &str,
159 _detail: i32,
160 _b: nodes::NodeRef,
161 _first: i32,
162 _last: i32,
163 ) -> nodes::NodeRef {
164 nodes::NodeRef::NULL
165 }
166
167 /// Runs the `hyphenate`, `ligaturing` or `kerning` callback `cb` on
168 /// the list after `head` (a temporary node), with its last node:
169 /// `true` if a function or `false` is registered, which replaces the
170 /// built-in pass.
171 fn node_pass(&self, _engine: &mut Engine, _cb: &str, _head: nodes::NodeRef) -> bool {
172 false
173 }
174
175 /// Calls the `linebreak_filter` callback on paragraph `head`
176 /// (`display`: a display follows): `Some(list)` if it returned a node
177 /// list, which replaces line breaking; `None` otherwise.
178 fn linebreak_filter(
179 &self,
180 _engine: &mut Engine,
181 _head: nodes::NodeRef,
182 _display: bool,
183 ) -> Option<nodes::NodeRef> {
184 None
185 }
186
187 /// Calls the `mlist_to_hlist` callback on math list `head` in style
188 /// `style` (0–7); `None` if the callback failed or none is set.
189 fn mlist_to_hlist(
190 &self,
191 _engine: &mut Engine,
192 _head: nodes::NodeRef,
193 _style: i32,
194 _penalties: bool,
195 ) -> Option<nodes::NodeRef> {
196 None
197 }
198
199 /// Runs shell command `cmd` under the shell escape rules (as Lua's
200 /// `io.popen` does) and returns its output; `None` if it may not run.
201 fn pipe_read(&self, _engine: &mut Engine, _cmd: &[u8]) -> Option<Vec<u8>> {
202 None
203 }
204
205 /// Calls the `input_level_string` callback with input-stack depth `n`;
206 /// `None` if it returned no string.
207 fn input_level_string(&self, _engine: &mut Engine, _n: i64) -> Option<Vec<u8>> {
208 None
209 }
210
211 /// Calls the `glyph_info` callback for glyph node `p`: `Some(text)` is
212 /// what to print, `None` means the callback returned no string.
213 fn glyph_info(&self, _engine: &mut Engine, _p: nodes::NodeRef) -> Option<Vec<u8>> {
214 None
215 }
216}
217
218/// Result of a node-list filter callback.
219#[derive(Clone, Copy, Debug, PartialEq, Eq)]
220pub enum ListFilter {
221 /// Continue with this list (possibly the original head).
222 Keep(nodes::NodeRef),
223 /// The callback returned `false`: discard the list.
224 Flush,
225}
226
227impl Engine {
228 /// Whether the Lua callback `name` is set.
229 pub fn has_callback(&self, name: &str) -> bool {
230 self.lua.as_ref().is_some_and(|h| h.has_callback(name))
231 }
232
233 /// Calls the notification callback `name` (`buildpage_filter`,
234 /// `contribute_filter`) with `info`, if one is set.
235 pub fn notify_callback(&mut self, name: &str, info: &str) {
236 if let Some(h) = self.lua.clone()
237 && h.has_callback(name)
238 && let Err(e) = h.call_with_string(self, name, info)
239 {
240 self.lua_error(&e);
241 }
242 }
243
244 /// Calls `buildpage_filter` with `info` (where the page builder was
245 /// called from, observed), then builds the page.
246 pub fn build_page_from(&mut self, info: &str) {
247 self.notify_callback("buildpage_filter", info);
248 self.build_page();
249 }
250
251 /// Runs a node-list filter callback if one is registered (the hook
252 /// typesetting code calls at LuaTeX's callback points).
253 pub fn run_list_filter(
254 &mut self,
255 callback: &str,
256 head: nodes::NodeRef,
257 group: &str,
258 ) -> ListFilter {
259 match self.lua.clone() {
260 Some(h) if h.has_callback(callback) => h.list_filter(self, callback, head, group),
261 _ => ListFilter::Keep(head),
262 }
263 }
264}
265
266impl Engine {
267 /// Converts a finished formula's math list into an hlist, through the
268 /// `mlist_to_hlist` callback when one is set (LuaTeX calls it for the
269 /// outermost list of each formula and equation number; observed: not
270 /// for an empty one).
271 pub fn formula_to_hlist(
272 &mut self,
273 head: nodes::NodeRef,
274 style: i32,
275 penalties: bool,
276 ) -> nodes::NodeRef {
277 if head.is_null() {
278 return nodes::NodeRef::NULL;
279 }
280 if let Some(h) = self.lua.clone()
281 && h.has_callback("mlist_to_hlist")
282 {
283 if let Some(r) = h.mlist_to_hlist(self, head, style, penalties) {
284 if !r.is_null() {
285 self.nodes.set_prev(r, nodes::NodeRef::NULL);
286 }
287 return r;
288 }
289 return nodes::NodeRef::NULL;
290 }
291 self.mlist_to_hlist(head, style, penalties)
292 }
293
294 /// Prints the input level before a line of macro tracing through the
295 /// `input_level_string` callback (LaTeX's `\tracingstacklevels`);
296 /// returns false if no callback is set. A newline in the string ends
297 /// the line.
298 pub fn print_input_level(&mut self) -> bool {
299 let Some(h) = self.lua.clone() else {
300 return false;
301 };
302 if !h.has_callback("input_level_string") {
303 return false;
304 }
305 let n = self.input.stack.len().saturating_sub(1) as i64;
306 if let Some(s) = h.input_level_string(self, n) {
307 for (i, part) in s.split(|&b| b == b'\n').enumerate() {
308 if i > 0 {
309 self.printer.print_ln();
310 }
311 self.printer.print_bytes(part);
312 }
313 }
314 true
315 }
316
317 /// Runs the `glyph_info` callback for glyph `p` when one is set:
318 /// `None` without a callback, `Some(None)` when it returned no string.
319 pub fn run_glyph_info(&mut self, p: nodes::NodeRef) -> Option<Option<Vec<u8>>> {
320 match self.lua.clone() {
321 Some(h) if h.has_callback("glyph_info") => Some(h.glyph_info(self, p)),
322 _ => None,
323 }
324 }
325}
326
327/// A one-shot callback into the host, run with the engine borrowed.
328pub type EngineHook = Box<dyn FnOnce(&mut Engine)>;
329
330/// Shows diagnostics before the terminal is read (see
331/// [`Engine::prompt_hook`]).
332pub type PromptHook = Box<dyn FnMut(&[diag::Diagnostic], &str) -> bool>;
333
334/// The global state of a TeX run.
335pub struct Engine {
336 /// Control-sequence names.
337 pub hash: CsTable,
338 /// The table of equivalents.
339 pub eqtb: Eqtb,
340 /// The save stack and grouping state.
341 pub save: SaveStack,
342 /// The input stack.
343 pub input: InputState,
344 /// Terminal, log and `\write` output.
345 pub printer: Printer,
346 /// Error-handling state.
347 pub err: ErrorState,
348 /// Conditionals.
349 pub cond: CondState,
350 /// Files: `\openin`/`\read`, `\openout`, input file stack.
351 pub files: FileState,
352 /// The file finder.
353 pub kpse: Option<retex_kpath::Shared>,
354 /// The Lua interpreter, when available.
355 pub lua: Option<std::rc::Rc<dyn LuaHost>>,
356
357 /// The current command code (`cur_cmd`).
358 pub cur_cmd: Cmd,
359 /// The current modifier (`cur_chr`).
360 pub cur_chr: u32,
361 /// The current control sequence (`cur_cs`); [`hash::NO_CS`] for
362 /// characters.
363 pub cur_cs: CsId,
364 /// The current token (`cur_tok`).
365 pub cur_tok: Token,
366
367 /// The value most recently scanned (`cur_val`).
368 pub cur_val: i32,
369 /// The glue most recently scanned, when `cur_val_level` is glue.
370 pub cur_glue: glue::Glue,
371 /// The token list most recently scanned, when `cur_val_level` is tokens.
372 pub cur_toks: Option<token::TokenList>,
373 /// The level of `cur_val`.
374 pub cur_val_level: scan::ValLevel,
375
376 /// Well-known control sequences.
377 pub cs: init::KnownCs,
378 /// Whether the run is in IniTeX mode.
379 pub ini_version: bool,
380 /// Shell escape mode (see [`init::Options::shell_escape`]).
381 pub shell_escape: u8,
382 /// The format identification printed in the banner.
383 pub format_ident: String,
384 /// The job name, once known.
385 pub job_name: Option<Vec<u8>>,
386 /// Token saved by `\afterassignment` (`after_token`).
387 pub after_token: Token,
388 /// Number of nested expansions (LuaTeX's `expand_depth_count`).
389 pub expand_depth: u32,
390 /// The current mode (copy of the innermost nest level's mode).
391 pub mode: maincontrol::Mode,
392 /// The mode last shown by `show_cur_cmd_chr`.
393 pub shown_mode: maincontrol::Mode,
394 /// The format name (`\formatname`).
395 pub format_name: String,
396 /// Node memory.
397 pub nodes: nodes::Nodes,
398 /// The semantic nest; the last entry is the current list.
399 pub nest: Vec<maincontrol::ListState>,
400 /// Loaded fonts.
401 pub fonts: fonts::FontTable,
402 /// The `long_state` of the macro being called.
403 pub long_state: Cmd,
404 /// The control sequence being defined (for diagnostics).
405 pub def_ref_cs: CsId,
406 /// The partial token list printed by `runaway`.
407 pub runaway_list: Vec<Token>,
408 /// The control sequence last made by `\csname` (`\lastnamedcs`).
409 pub last_cs_name: CsId,
410 /// Frozen copies made by `\primitive`.
411 pub frozen_prims: std::collections::HashMap<Vec<u8>, CsId>,
412 /// Mark texts.
413 pub marks: typeset::Marks,
414 /// Set when a `\read` line has been consumed.
415 pub input_read_finished: bool,
416 /// Whether the long `\errmessage` help has been shown.
417 pub long_help_seen: bool,
418 /// `\deadcycles`.
419 pub dead_cycles: i32,
420 /// `\insertpenalties`.
421 pub insert_penalties: i32,
422 /// `\badness`.
423 pub last_badness: i32,
424 /// Token lists show a macro parameter character once, not doubled
425 /// (`\luaescapestring`).
426 pub single_mac_param: bool,
427 /// The direction the next box made by `hpack` or `vpackage` gets
428 /// (from `dir`/`bdir` of a box specification), or -1. It is set before
429 /// packing because over- and underfull box reports show it.
430 pub pack_dir: i8,
431 /// `\randomseed`: the seed `\setrandomseed` last gave (its absolute
432 /// value), or the one chosen at startup.
433 pub random_seed: i32,
434 /// The state of the random number generator (METAFONT's, see
435 /// [`retex_arith::random`]) behind `\uniformdeviate`, `\normaldeviate`
436 /// and the `tex` library's randomizers.
437 pub randoms: retex_arith::random::Randoms,
438 /// `\lastxpos`.
439 pub last_x_pos: i32,
440 /// `\lastypos`.
441 pub last_y_pos: i32,
442 /// The magnification frozen by `prepare_mag`.
443 pub mag_set: i32,
444 /// The font dimension located by `find_font_dimen`.
445 pub font_dimen_loc: Option<(u32, u32)>,
446 /// `\showboxdepth` in effect while displaying boxes.
447 pub depth_threshold: i32,
448 /// `\showboxbreadth` in effect while displaying boxes.
449 pub breadth_max: i32,
450 /// The paragraph-ending control sequence (`\partokenname`).
451 pub par_token_cs: CsId,
452 /// Whether `\dump` ended the run.
453 pub dump_requested: bool,
454 /// Integer parameters set from the command line when the job starts
455 /// ([`init::Options::int_params`]).
456 pub start_ints: Vec<(u32, i32)>,
457 /// `-output-comment`.
458 pub output_comment: Option<String>,
459 /// The job's start time (`\time`, the banner, the PDF creation date).
460 pub start_time: clock::StartTime,
461 /// The page builder.
462 pub page: typeset::PageState,
463 /// Lua functions registered for `\luafunction` (by number).
464 pub lua_functions: std::collections::HashMap<i32, u32>,
465 /// Called to store the format when `\dump` is executed.
466 pub format_hook: Option<EngineHook>,
467 /// Called at the end of the run to finish output files.
468 pub finish_hook: Option<EngineHook>,
469 /// Non-immediate `\write`s waiting for shipout.
470 pub deferred_writes: Vec<(i32, token::TokenList)>,
471 /// Text directions opened in the current box: (group level,
472 /// direction, the text direction before), innermost last (see
473 /// [`dirs`]).
474 pub text_dirs: Vec<(u16, i8, i8)>,
475 /// Box resources to write after the page or form being written, one
476 /// list per open page or form.
477 pub pending_forms: Vec<Vec<u32>>,
478 /// Objects referred to (`\pdfextension refobj`) in the box resources
479 /// being written, one list per open form.
480 pub form_objects: Vec<Vec<pdfext::PdfObject>>,
481 /// The `text_dirs` of the enclosing boxes.
482 pub outer_text_dirs: Vec<Vec<(u16, i8, i8)>>,
483 /// The math and text directions outside the open formulas, with the
484 /// save level of each formula's group.
485 pub outer_math_dirs: Vec<(u16, i32, i32)>,
486 /// Contents of files written by `\write`, for the pass driver.
487 pub written_sink: files::WrittenSink,
488 /// State of the typesetting routines.
489 pub ts: typeset::TsState,
490 /// Tokens printed from Lua waiting to be pushed as a token list.
491 pub pending_lua_tokens: Option<Vec<Token>>,
492 /// Default output mode: errors and warnings recorded for the driver
493 /// (`None` in compatibility mode).
494 pub diagnostics: Option<Vec<diag::Diagnostic>>,
495 /// Default output mode, interactive: shows the diagnostics recorded
496 /// since the last prompt, and the prompt, before the terminal is read.
497 /// Returns `false` if it showed nothing (the error only repeats one
498 /// already reported): TeX then goes on without asking.
499 pub prompt_hook: Option<PromptHook>,
500 /// Default output mode: called when a page is shipped or a file opened,
501 /// to show progress and the errors recorded so far.
502 pub progress_hook: Option<fn(&mut Engine)>,
503 /// The number of diagnostics the prompt hook has shown.
504 pub diagnostics_shown: usize,
505 /// Whether the terminal was read (an error or a `\read` prompt).
506 pub prompted: bool,
507 /// Image resources (`\saveimageresource`), in the order saved (their
508 /// indices are their object numbers).
509 pub images: Vec<image::ImageResource>,
510 /// `\lastsavedimageresourceindex`.
511 pub last_image_index: i32,
512 /// `\lastsavedimageresourcepages`.
513 pub last_image_pages: i32,
514 /// Box resources (`\saveboxresource`), in the order saved (their
515 /// indices are their object numbers).
516 pub box_resources: Vec<image::BoxResource>,
517 /// `\lastsavedboxresourceindex`.
518 pub last_box_index: i32,
519 /// Scratch buffer for macro arguments (reused by `macro_call`).
520 pub arg_scratch: Vec<Token>,
521 /// Scratch buffer for the argument list of a macro call.
522 pub param_scratch: Vec<token::TokenList>,
523 /// The file of each `\write` stream opened (see [`snapshot`]).
524 pub write_paths: Vec<Option<std::path::PathBuf>>,
525 /// The macro at whose expansion [`snapshot_hook`](Self::snapshot_hook)
526 /// runs (LaTeX's `\document`).
527 pub snapshot_cs: Option<CsId>,
528 /// Called once, when `snapshot_cs` is expanded (see [`snapshot`]).
529 pub snapshot_hook: Option<snapshot::Hook>,
530 /// The file and line the user asked to edit at an error (`E`); the
531 /// program runs the editor at the end.
532 pub edit_request: Option<(Vec<u8>, i32)>,
533 /// The output file written at the end: name, pages and bytes.
534 pub output_written: Option<(String, u32, u64)>,
535 /// The line number reported while reading Lua output.
536 pub input_line_for_lua: i32,
537 /// Internal tokens made by `\pdfvariable`, by variable name.
538 pub backend_cs: std::collections::HashMap<&'static str, CsId>,
539 /// Nesting depth of local control loops (`tex.runtoks`).
540 pub local_level: u32,
541 /// The box `token.scan_list` scanned, once finished.
542 pub lua_scanned_box: Option<nodes::NodeRef>,
543 /// Whether `token.scan_list` runs local control for its box.
544 pub lua_scan_waiting: bool,
545 /// The output backend (PDF writer), if any.
546 pub backend: Option<Box<dyn shipout::Backend>>,
547 /// The DVI writer (`\outputmode=0`).
548 pub dvi: dvi::DviWriter,
549 /// Hyphenation languages, by `\language` number.
550 pub languages: Vec<Option<retex_hyph::Language>>,
551 /// Document-level PDF data from `\pdfextension`.
552 pub pdf: pdfext::PdfState,
553 /// Current values of the colour stacks during shipout.
554 pub colorstack_values: std::collections::HashMap<i32, Vec<Vec<u8>>>,
555 /// Number of pages shipped out.
556 pub total_pages: u32,
557 /// Muglue values of `\Umath…` spacing parameters.
558 pub math_glues: Vec<glue::Glue>,
559 /// Whether the parameters set by the first formula are initialized
560 /// (see [`mathparams`]).
561 pub math_params_initialized: bool,
562 /// SyncTeX output.
563 pub synctex: synctex::Synctex,
564}
565
566impl std::fmt::Debug for Engine {
567 fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
568 f.debug_struct("Engine")
569 .field("cur_cmd", &self.cur_cmd)
570 .field("cur_chr", &self.cur_chr)
571 .finish_non_exhaustive()
572 }
573}