workflow/preprocess.rs
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077
// PREPROCESS.rs
// by Lut99
//
// Created:
// 02 Nov 2023, 14:52:26
// Last edited:
// 12 Jun 2024, 17:41:53
// Auto updated?
// Yes
//
// Description:
//! Defines a preprocessing step on a [WIR](Workflow) that simplifies it
//! to increase the support of the simpler checker workflow.
//
use std::collections::{HashMap, HashSet};
use std::error;
use std::fmt::{Display, Formatter, Result as FResult};
use std::panic::catch_unwind;
use std::sync::Arc;
use brane_ast::MergeStrategy;
use brane_ast::ast::{Edge, EdgeInstr, FunctionDef, SymTable, TaskDef, Workflow};
use brane_ast::func_id::FunctionId;
use brane_ast::spec::BuiltinFunctions;
use brane_exe::pc::{ProgramCounter, ResolvedProgramCounter};
use enum_debug::EnumDebug as _;
use log::{debug, trace};
use super::utils;
/***** TESTS *****/
#[cfg(test)]
mod tests {
use std::ffi::OsStr;
use std::path::PathBuf;
use brane_ast::traversals::print::ast;
use brane_ast::{CompileResult, ParserOptions, compile_program};
use brane_shr::utilities::{create_data_index_from, create_package_index_from, test_on_dsl_files_in};
use humanlog::{DebugMode, HumanLogger};
use specifications::data::DataIndex;
use specifications::package::PackageIndex;
use super::*;
/// Runs checks to verify the workflow inlining analysis
#[test]
fn test_checker_workflow_inline_analysis() {
// Setup logger if told
if std::env::var("TEST_LOGGER").map(|value| value == "1" || value == "true").unwrap_or(false) {
if let Err(err) = HumanLogger::terminal(DebugMode::Full).init() {
eprintln!("WARNING: Failed to setup test logger: {err} (no logging for this session)");
}
}
// Defines a few test files with expected inlinable functions
let tests: [(&str, &str, HashMap<usize, Option<HashSet<usize>>>); 5] = [
("case1", r#"println("Hello, world!");"#, HashMap::from([(1, None)])),
(
"case2",
r#"func hello_world() { return "Hello, world!"; } println(hello_world());"#,
HashMap::from([(1, None), (4, Some(HashSet::new()))]),
),
(
"case3",
r#"func foo() { return "Foo"; } func foobar() { return foo() + "Bar"; } println(foobar());"#,
HashMap::from([(1, None), (4, Some(HashSet::new())), (5, Some(HashSet::from([4])))]),
),
("case4", r#"import hello_world; println(hello_world());"#, HashMap::from([(1, None)])),
(
"case5",
r#"func hello_world(n) { if (n <= 0) { return "Hello, world!"; } else { return "Hello, " + hello_world(n - 1) + "\n"; } } println(hello_world(3));"#,
HashMap::from([(1, None), (4, None)]),
),
];
// Load example package- and data indices
let tests_path: PathBuf = PathBuf::from(super::super::tests::TESTS_DIR);
let pindex: PackageIndex = create_package_index_from(tests_path.join("packages"));
let dindex: DataIndex = create_data_index_from(tests_path.join("data"));
// Test them each
for (id, test, gold) in tests.into_iter() {
// Compile to BraneScript (we'll assume this works)
let wir: Workflow = match compile_program(test.as_bytes(), &pindex, &dindex, &ParserOptions::bscript()) {
CompileResult::Workflow(wir, _) => wir,
CompileResult::Err(errs) => {
for err in errs {
err.prettyprint(format!("<{id}>"), test);
}
panic!("Failed to compile BraneScript (see error above)");
},
CompileResult::Eof(err) => {
err.prettyprint(format!("<{id}>"), test);
panic!("Failed to compile BraneScript (see error above)");
},
_ => {
unreachable!();
},
};
// Emit the compiled workflow
println!("{}", (0..80).map(|_| '-').collect::<String>());
println!("Test '{id}'");
println!();
ast::do_traversal(&wir, std::io::stdout()).unwrap();
println!();
// Analyse function calls (we'll assume this works too)
let calls: HashMap<ProgramCounter, usize> = resolve_calls(&wir, &wir.table, &mut vec![], ProgramCounter::start(), None, None).unwrap().0;
println!(
"Resolved functions calls: {:?}",
calls.iter().map(|(pc, func_id)| (format!("{}", pc.resolved(&wir.table)), *func_id)).collect::<HashMap<String, usize>>()
);
// Analyse the inlinable funcs
let mut pred: HashMap<usize, Option<HashSet<usize>>> = HashMap::with_capacity(calls.len());
find_inlinable_funcs(&wir, &calls, &mut vec![], ProgramCounter::start(), None, &mut pred);
println!("Inlinable functions: {pred:?}");
println!();
// Neat, done, assert it was right
assert_eq!(pred, gold);
}
}
/// Runs the workflow inlining on the test files only
#[test]
fn test_checker_workflow_simplify() {
let tests_path: PathBuf = PathBuf::from(super::super::tests::TESTS_DIR);
// Setup logger if told
if std::env::var("TEST_LOGGER").map(|value| value == "1" || value == "true").unwrap_or(false) {
if let Err(err) = HumanLogger::terminal(DebugMode::Full).init() {
eprintln!("WARNING: Failed to setup test logger: {err} (no logging for this session)");
}
}
// Scope the function
let test_file: Option<String> = std::env::var("TEST_FILE").ok();
// Run the compiler for every applicable DSL file
test_on_dsl_files_in("BraneScript", &tests_path, |path: PathBuf, code: String| {
// Skip if not the file we're looking for
if let Some(test_file) = &test_file {
if path.file_name().is_none() || path.file_name().unwrap().to_string_lossy() != test_file.as_str() {
return;
}
}
// Start by the name to always know which file this is
println!("{}", (0..80).map(|_| '-').collect::<String>());
println!("File '{}' gave us:", path.display());
// Skip some files, sadly
if let Some(name) = path.file_name() {
if name == OsStr::new("class.bs") {
println!("Skipping test, since instance calling is not supported in checker workflows...");
println!("{}\n\n", (0..80).map(|_| '-').collect::<String>());
return;
}
}
// Load the package index
let pindex: PackageIndex = create_package_index_from(tests_path.join("packages"));
let dindex: DataIndex = create_data_index_from(tests_path.join("data"));
// Compile the raw source to WIR
let wir: Workflow = match compile_program(code.as_bytes(), &pindex, &dindex, &ParserOptions::bscript()) {
CompileResult::Workflow(wir, warns) => {
// Print warnings if any
for w in warns {
w.prettyprint(path.to_string_lossy(), &code);
}
wir
},
CompileResult::Eof(err) => {
// Print the error
err.prettyprint(path.to_string_lossy(), &code);
panic!("Failed to compile to WIR (see output above)");
},
CompileResult::Err(errs) => {
// Print the errors
for e in errs {
e.prettyprint(path.to_string_lossy(), &code);
}
panic!("Failed to compile to WIR (see output above)");
},
_ => {
unreachable!();
},
};
// Alright preprocess it
let wir: Workflow = match simplify(wir) {
Ok((wir, _)) => wir,
Err(err) => {
panic!("Failed to preprocess WIR: {err}");
},
};
// Now print the file for prettyness
ast::do_traversal(&wir, std::io::stdout()).unwrap();
println!("{}\n\n", (0..80).map(|_| '-').collect::<String>());
});
}
}
/***** ERRORS *****/
/// Defines errors that may occur when preprocessing a [`Workflow`].
#[derive(Debug)]
pub enum Error {
/// Unknown task given.
UnknownTask { id: usize },
/// Unknown function given.
UnknownFunc { id: FunctionId },
/// A [`Call`](ast::Edge::Call)-edge was encountered while we didn't know of a function ID on the stack.
CallingWithoutId { pc: ResolvedProgramCounter },
}
impl Display for Error {
fn fmt(&self, f: &mut Formatter<'_>) -> FResult {
use Error::*;
match self {
UnknownTask { id } => write!(f, "Encountered unknown task ID {id} in Node"),
UnknownFunc { id } => write!(f, "Encountered unknown function ID {id} in Call"),
CallingWithoutId { pc } => write!(f, "Attempted to call function at {pc} without statically known task ID on the stack"),
}
}
}
impl error::Error for Error {}
/***** ANALYSIS FUNCTIONS *****/
/// Checks whether the given stream of instructions would end with a function ID on top of the stack.
///
/// # Arguments
/// - `instrs`: The list of instructions to analyse.
/// - `idx`: The index of the particular instruction (i.e., the previous one) to examine. When calling this functio non-recursively, use the **last** instruction.
///
/// # Returns
/// A double [`Option`] detailling what's possible:
/// - [`Some(Some(...))`] means that there was a function ID on top.
/// - [`Some(None)`] means that we _know_ there is _no_ function ID on top.
/// - [`None`] means that nothing was pushed, i.e., whatever was on top is still on top.
fn pushes_func_id(instrs: &[EdgeInstr], idx: usize) -> Option<Option<usize>> {
// Pop the next instruction
let instr: &EdgeInstr = if idx < instrs.len() {
&instrs[idx]
} else {
// If we reached the last instruction, then we know no value was pushed :celebrate:
return None;
};
// Examine what it does
// NOTE: The BraneScript compiler only supports function calls over identifiers and projections. So we can ignore gnarly array stuff etc!
// NOTE: Actually... we know violently little statically of class calls in general, because they are fully pushed to dynamic land. We _could_ learn it by tracking
// a variable's contents over multiple edges, but that fucks; let's give up and only support direct calls for now.
match instr {
// What we're looking for!
EdgeInstr::Function { def } => Some(Some(*def)),
// Things instructions only pop, potentially (accidentally) removing our function
// Jep just tell the thign we don't know, we don't need it for direct function calls
EdgeInstr::Pop {} | EdgeInstr::PopMarker {} | EdgeInstr::DynamicPop {} | EdgeInstr::VarSet { .. } => Some(None),
// Alright some weird local branching; fuck it, also give up because we don't know which of the branches will do it
EdgeInstr::Branch { .. } | EdgeInstr::BranchNot { .. } => Some(None),
// These instructions never pop- or push anything
EdgeInstr::VarDec { .. } | EdgeInstr::VarUndec { .. } => Some(None),
// These instructions push invalid things _for sure_
EdgeInstr::Cast { .. }
| EdgeInstr::Not {}
| EdgeInstr::Neg {}
| EdgeInstr::And {}
| EdgeInstr::Or {}
| EdgeInstr::Add {}
| EdgeInstr::Sub {}
| EdgeInstr::Mul {}
| EdgeInstr::Div {}
| EdgeInstr::Mod {}
| EdgeInstr::Eq {}
| EdgeInstr::Ne {}
| EdgeInstr::Lt {}
| EdgeInstr::Le {}
| EdgeInstr::Gt {}
| EdgeInstr::Ge {}
| EdgeInstr::Array { .. }
| EdgeInstr::ArrayIndex { .. }
| EdgeInstr::Instance { .. }
| EdgeInstr::Proj { .. }
| EdgeInstr::VarGet { .. }
| EdgeInstr::Boolean { .. }
| EdgeInstr::Integer { .. }
| EdgeInstr::Real { .. }
| EdgeInstr::String { .. } => Some(None),
}
}
/// Analyses the edges in an [`Workflow`] to resolve function calls to the ID of the functions they call.
///
/// # Arguments
/// - `wir`: The [`Workflow`] to analyse.
/// - `table`: A running [`VirtualSymTable`] that determines the current types in scope.
/// - `trace`: A stack of call pointers that keeps track of the trace of function calls. Allows us to avoid recursion.
/// - `stack_id`: The function ID currently known to be on the stack. Is [`None`] if we don't know this.
/// - `pc`: The program-counter-index of the edge to analyse. These are pairs of `(function, edge_idx)`, where main is referred to by [`usize::MAX`](usize).
/// - `breakpoint`: An optional program-counter-index that, if given, will not analyse that edge onwards (excluding it too).
///
/// # Returns
/// A tuple with a [`HashMap`] that maps call indices (as program-counter-indices) to function IDs and an optional top call ID currently on the stack.
///
/// Note that, if a call ID occurs in the map but has [`None`] as function ID, it means it does not map to a body (e.g., a builtin).
///
/// # Errors
/// This function may error if we failed to statically discover the function IDs.
fn resolve_calls(
wir: &Workflow,
table: &SymTable,
trace: &mut Vec<ProgramCounter>,
pc: ProgramCounter,
stack_id: Option<usize>,
breakpoint: Option<ProgramCounter>,
) -> Result<(HashMap<ProgramCounter, usize>, Option<usize>), Error> {
// Quit if we're at the breakpoint
if let Some(breakpoint) = breakpoint {
if pc == breakpoint {
return Ok((HashMap::new(), None));
}
}
// Get the edge in the workflow
let edge: &Edge = match utils::get_edge(wir, pc) {
Some(edge) => edge,
None => return Ok((HashMap::new(), None)),
};
// Match to recursively process it
trace!("Attempting to resolve calls in {} ({:?})", pc.resolved(table), edge.variant());
match edge {
Edge::Node { task, next, .. } => {
// Attempt to discover the return type of the Node.
let def: &TaskDef = match table.tasks.get(*task) {
Some(def) => def,
None => return Err(Error::UnknownTask { id: *task }),
};
// Alright, recurse with the next instruction
resolve_calls(wir, table, trace, pc.jump(*next), if def.func().ret.is_void() { stack_id } else { None }, breakpoint)
},
Edge::Linear { instrs, next } => {
// Analyse the instructions to find out if we can deduce a new `stack_id`
let stack_id: Option<usize> = if !instrs.is_empty() { pushes_func_id(instrs, instrs.len() - 1).unwrap_or(stack_id) } else { stack_id };
// Analyse the next one
resolve_calls(wir, table, trace, pc.jump(*next), stack_id, breakpoint)
},
Edge::Stop {} => Ok((HashMap::new(), None)),
Edge::Branch { true_next, false_next, merge } => {
// First, analyse the branches
let (mut calls, mut stack_id): (HashMap<_, _>, Option<usize>) =
resolve_calls(wir, table, trace, pc.jump(*true_next), stack_id, merge.map(|merge| pc.jump(merge)))?;
if let Some(false_next) = false_next {
let (false_calls, false_stack) = resolve_calls(wir, table, trace, pc.jump(*false_next), stack_id, merge.map(|merge| pc.jump(merge)))?;
calls.extend(false_calls);
if stack_id != false_stack {
stack_id = None;
}
}
// Analyse the remaining part next
if let Some(merge) = merge {
let (merge_calls, merge_stack) = resolve_calls(wir, table, trace, pc.jump(*merge), stack_id, breakpoint)?;
calls.extend(merge_calls);
stack_id = merge_stack;
}
// Alright, return the found results
Ok((calls, stack_id))
},
Edge::Parallel { branches, merge } => {
// Simply analyse all branches first. No need to worry about their return values and such, since that's not until the `Join`.
let mut calls: HashMap<_, _> = HashMap::new();
for branch in branches {
calls.extend(resolve_calls(wir, table, trace, pc.jump(*branch), stack_id, breakpoint)?.0);
}
// OK, then analyse the rest assuming the stack is unchanged (we can do that because the parallel's branches get clones)
let (new_calls, stack_id): (HashMap<_, _>, Option<usize>) = resolve_calls(wir, table, trace, pc.jump(*merge), stack_id, breakpoint)?;
calls.extend(new_calls);
Ok((calls, stack_id))
},
Edge::Join { merge, next } => {
// Simply do the next, only _not_ resetting the stack ID if no value is returned.
resolve_calls(wir, table, trace, pc.jump(*next), if *merge == MergeStrategy::None { stack_id } else { None }, breakpoint)
},
Edge::Loop { cond, body, next } => {
// Traverse the three individually, using the stack ID of the codebody that precedes it
let (mut calls, mut cond_id): (HashMap<_, _>, Option<usize>) =
resolve_calls(wir, table, trace, pc.jump(*cond), stack_id, Some(pc.jump(*body - 1)))?;
let (body_calls, _): (HashMap<_, _>, Option<usize>) = resolve_calls(wir, table, trace, pc.jump(*body), cond_id, Some(pc.jump(*cond)))?;
calls.extend(body_calls);
if let Some(next) = next {
let (next_calls, next_id): (HashMap<_, _>, Option<usize>) = resolve_calls(wir, table, trace, pc.jump(*next), cond_id, breakpoint)?;
calls.extend(next_calls);
cond_id = next_id;
}
// Done!
Ok((calls, cond_id))
},
Edge::Call { input: _, result: _, next } => {
// Alright time to jump functions based on the current top-of-the-stack
let stack_id: usize = match stack_id {
Some(id) => id,
None => {
return Err(Error::CallingWithoutId { pc: pc.resolved(table) });
},
};
// We can early quit upon recursion
if trace.contains(&pc) {
let mut calls: HashMap<ProgramCounter, usize> = HashMap::from([(pc, stack_id)]);
let (next_calls, next_id): (HashMap<_, _>, Option<usize>) = resolve_calls(wir, table, trace, pc.jump(*next), None, breakpoint)?;
calls.extend(next_calls);
return Ok((calls, next_id));
}
// Add the mapping to the table
let mut calls: HashMap<ProgramCounter, usize> = HashMap::from([(pc, stack_id)]);
// Resolve the call of the function (builtins simply return nothing, so are implicitly handled)
trace.push(pc);
let (call_calls, call_id): (HashMap<_, _>, Option<usize>) = resolve_calls(wir, table, trace, ProgramCounter::call(stack_id), None, None)?;
trace.pop();
calls.extend(call_calls);
// Then continue with the next one
let (next_calls, next_id): (HashMap<_, _>, Option<usize>) = resolve_calls(wir, table, trace, pc.jump(*next), call_id, breakpoint)?;
calls.extend(next_calls);
Ok((calls, next_id))
},
Edge::Return { result: _ } => {
// If we're in the main function, this acts as an [`Elem::Stop`] with value
if pc.is_main() {
return Ok((HashMap::new(), None));
}
// To see whether we pass a function ID, consult the function definition
let def: &FunctionDef = match catch_unwind(|| table.func(pc.func_id)) {
Ok(def) => def,
Err(_) => return Err(Error::UnknownFunc { id: pc.func_id }),
};
// Only return the current one if the function returns void
if def.ret.is_void() { Ok((HashMap::new(), stack_id)) } else { Ok((HashMap::new(), None)) }
},
}
}
/// Attempts to find all non-recursive functions in the given WIR.
///
/// The only moment when we don't consider a function inlinable is if the function call is:
/// - Recursive
/// - A builtin
/// - Undecidable
///
/// # Arguments
/// - `wir`: The input [WIR](Workflow) to analyse.
/// - `calls`: The map of call indices to which function is actually called.
/// - `trace`: A trace of function IDs that we've "called".
/// - `pc`: Points to the current [`Edge`] to analyse.
/// - `breakpoint`: If given, then analysis should stop when this PC is hit.
/// - `inlinable`: The result we're recursively building. This set simply collects all function IDs and maps them to inlinable or not. If they are, then their ID is mapped to a list of functions on which the call depends (or else [`None`]).
///
/// # Returns
/// A list of all function calls found (that are inlinable). This builds a dependency tree of which calls the given depends on.
fn find_inlinable_funcs(
wir: &Workflow,
calls: &HashMap<ProgramCounter, usize>,
trace: &mut Vec<usize>,
pc: ProgramCounter,
breakpoint: Option<ProgramCounter>,
inlinable: &mut HashMap<usize, Option<HashSet<usize>>>,
) -> HashSet<usize> {
// Stop on the breakpoint
if let Some(breakpoint) = breakpoint {
if pc == breakpoint {
return HashSet::new();
}
}
// Attempt to get the edge
let edge: &Edge = match utils::get_edge(wir, pc) {
Some(edge) => edge,
None => return HashSet::new(),
};
// Match on its kind
trace!("Finding inlinable functions in {} ({:?})", pc.resolved(&wir.table), edge.variant());
match edge {
Edge::Node { next, .. } | Edge::Linear { next, .. } => {
// Doesn't call any functions, so just proceed with the next one
find_inlinable_funcs(wir, calls, trace, pc.jump(*next), breakpoint, inlinable)
},
Edge::Stop {} => HashSet::new(),
Edge::Branch { true_next, false_next, merge } => {
// Analyse the left branch...
let mut dependencies: HashSet<usize> =
find_inlinable_funcs(wir, calls, trace, pc.jump(*true_next), merge.map(|merge| pc.jump(merge)), inlinable);
// ...the right branch...
if let Some(false_next) = false_next {
dependencies.extend(find_inlinable_funcs(wir, calls, trace, pc.jump(*false_next), merge.map(|merge| pc.jump(merge)), inlinable));
}
// ...and the merge!
if let Some(merge) = merge {
dependencies.extend(find_inlinable_funcs(wir, calls, trace, pc.jump(*merge), breakpoint, inlinable));
}
dependencies
},
Edge::Parallel { branches, merge } => {
// Collect all the branches
let mut dependencies: HashSet<usize> = HashSet::new();
for branch in branches {
dependencies.extend(find_inlinable_funcs(wir, calls, trace, pc.jump(*branch), Some(pc.jump(*merge)), inlinable));
}
// Run merge and done is Cees
dependencies.extend(find_inlinable_funcs(wir, calls, trace, pc.jump(*merge), breakpoint, inlinable));
dependencies
},
Edge::Join { next, .. } => find_inlinable_funcs(wir, calls, trace, pc.jump(*next), breakpoint, inlinable),
Edge::Loop { cond, body, next } => {
// Traverse the condition...
let mut dependencies: HashSet<usize> = find_inlinable_funcs(wir, calls, trace, pc.jump(*cond), Some(pc.jump(*body - 1)), inlinable);
// ...the body...
dependencies.extend(find_inlinable_funcs(wir, calls, trace, pc.jump(*body), Some(pc.jump(*cond)), inlinable));
// ...and finally, the next step, if any
if let Some(next) = next {
dependencies.extend(find_inlinable_funcs(wir, calls, trace, pc.jump(*next), breakpoint, inlinable));
}
dependencies
},
Edge::Call { next, .. } => {
// OK, the exciting point!
// Resolve the function ID we're calling
let func_id: usize = match calls.get(&pc) {
Some(id) => *id,
None => {
panic!("Encountered unresolved call after running call analysis");
},
};
let def: &FunctionDef = match wir.table.funcs.get(func_id) {
Some(def) => def,
None => panic!("Failed to get definition of function {func_id} after call analysis"),
};
// Analyse next, since all codepaths do this always
let mut dependencies: HashSet<usize> = find_inlinable_funcs(wir, calls, trace, pc.jump(*next), None, inlinable);
dependencies.insert(func_id);
// Functions are not inlinable if builtins; if so, return
if BuiltinFunctions::is_builtin(&def.name) {
trace!("Function {} ('{}') is not inlinable because it is a builtin", func_id, def.name);
inlinable.insert(func_id, None);
return dependencies;
}
// Examine if this call would introduce a recursive problem
if trace.contains(&func_id) {
// It's been in our callstack before - that means recursion!
// Change our minds about its inlinability
trace!("Function {} ('{}') is not inlinable because it is recursive", func_id, def.name);
inlinable.insert(func_id, None);
return dependencies;
}
if inlinable.contains_key(&func_id) {
// We've already seen this one! However, _don't_ change our mind about its inlinability because it means a repeated function call
// NOTE: No need to go into the call body, as we've done this the first time we saw it
trace!("Function {} ('{}') is skipped because we have seen it before", func_id, def.name);
return dependencies;
}
trace!("Function {} ('{}') is assumed as inlinable until we see it recursive", func_id, def.name);
// For now assume that the function exist with no deps; we inject these later
inlinable.insert(func_id, Some(HashSet::new()));
// If we get this far, recurse into the body
trace.push(func_id);
let func_deps: HashSet<usize> = find_inlinable_funcs(wir, calls, trace, ProgramCounter::call(func_id), None, inlinable);
trace.pop();
// Now we can inject the entries
if let Some(deps) = inlinable.get_mut(&func_id).unwrap() {
deps.extend(func_deps);
}
// Return the dependencies in _this_ body.
dependencies
},
Edge::Return { result: _ } => HashSet::new(),
}
}
/// Orders a given map of inlinable functions such that, when ordered inline, every function will have its calls inlined if possible.
///
/// More specifically, the order makes sure that functions on which other functions depend (i.e., they make calls to it) are inlined first so that they can be inlined properly in the functions calling them.
///
/// # Arguments
/// - `ordered`: The vector of ordered function IDs that is being populated. The inline order is left-to-right (i.e., the leftmost function should never have a dependency, the second-to-left can only depend on the leftmost, etc).
/// - `inlinable`: The map of inlinable functions to their dependencies.
fn order_inlinable<'i>(ordered: &mut Vec<usize>, inlinable: &HashMap<usize, Option<HashSet<usize>>>, mut next: impl Iterator<Item = &'i usize>) {
// Get a function to inline
let func_id: usize = match next.next() {
Some(id) => *id,
None => return,
};
let deps: &HashSet<usize> = match inlinable.get(&func_id).unwrap() {
Some(deps) => deps,
None => {
// No need to inline this one, so just continue
trace!("order_inlinable(): Not considering function {func_id} because it is not inlinable (deps is None)");
order_inlinable(ordered, inlinable, next);
return;
},
};
// Examine the dependencies
if deps.is_empty() {
// Base-case; add to the list first before any other
trace!("order_inlinable(): Function {func_id} is inlinable but has no dependencies");
ordered.push(func_id);
order_inlinable(ordered, inlinable, next);
trace!("order_inlinable(): New result: {ordered:?}");
} else {
// Recursive case: add all the dependencies first
trace!("order_inlinable(): Function {func_id} is inlinable and has dependencies");
order_inlinable(ordered, inlinable, deps.iter());
ordered.push(func_id);
trace!("order_inlinable(): New result: {ordered:?}");
order_inlinable(ordered, inlinable, next);
}
}
/// Given a vector, removes all duplicates from it.
///
/// Retains the **first** occurrences.
///
/// # Arguments
/// - `data`: The vector to deduplicate.
fn keep_unique_first(data: &mut Vec<usize>) {
// A buffer of seen elements
let mut seen: HashSet<usize> = HashSet::new();
data.retain(|elem| {
if seen.contains(elem) {
false
} else {
seen.insert(*elem);
true
}
});
}
/// Traverses the given function body and replaces all [`Edge::Return`] with an [`Edge::Linear`] pointing to the given edge index.
///
/// Also bumps definition pointers with the given values. This is necessary because we need to pull function scopes one layer up.
///
/// # Arguments
/// - `edges`: The edges to traverse.
/// - `calls`: The map of program counters to calls that we update with any nested call's' new position.
/// - `func_id`: The ID of this function.
/// - `start_idx`: The index to add all next indices.
/// - `ret_idx`: The index to point the returning linears to.
/// - `pc`: Points to the current [`Edge`] to replace potentially.
/// - `breakpoint`: If given, then analysis should stop when this PC is hit.
fn prep_func_body(
edges: &mut [Edge],
calls: &mut HashMap<ProgramCounter, usize>,
func_id: usize,
start_idx: usize,
ret_idx: usize,
pc: usize,
breakpoint: Option<usize>,
) {
// Stop on the breakpoint
if let Some(breakpoint) = breakpoint {
if pc == breakpoint {
return;
}
}
// Attempt to get the edge
let edge: &mut Edge = match edges.get_mut(pc) {
Some(edge) => edge,
None => return,
};
// Match on its kind
match edge {
Edge::Node { next, .. } | Edge::Linear { next, .. } => {
// Update the nexts
let old_next: usize = *next;
*next += start_idx;
// Continue traversing
prep_func_body(edges, calls, func_id, start_idx, ret_idx, old_next, breakpoint);
},
Edge::Stop {} => (),
Edge::Branch { true_next, false_next, merge } => {
let (old_true_next, old_false_next, old_merge): (usize, Option<usize>, Option<usize>) = (*true_next, *false_next, *merge);
// Update the nexts
*true_next += start_idx;
if let Some(false_next) = false_next {
*false_next += start_idx;
}
if let Some(merge) = merge {
*merge += start_idx;
}
// Analyse the left branch...
prep_func_body(edges, calls, func_id, start_idx, ret_idx, old_true_next, old_merge);
// ...the right branch...
if let Some(old_false_next) = old_false_next {
prep_func_body(edges, calls, func_id, start_idx, ret_idx, old_false_next, old_merge);
}
// ...and the merge!
if let Some(old_merge) = old_merge {
prep_func_body(edges, calls, func_id, start_idx, ret_idx, old_merge, breakpoint);
}
},
Edge::Parallel { branches, merge } => {
let (old_branches, old_merge): (Vec<usize>, usize) = (branches.clone(), *merge);
// Update the nexts
for branch in branches {
*branch += start_idx;
}
*merge += start_idx;
// Collect all the branches
for old_branch in old_branches {
prep_func_body(edges, calls, func_id, start_idx, ret_idx, old_branch, Some(old_merge));
}
// Run merge and done is Cees
prep_func_body(edges, calls, func_id, start_idx, ret_idx, old_merge, breakpoint);
},
Edge::Join { next, .. } => {
let old_next: usize = *next;
*next += start_idx;
prep_func_body(edges, calls, func_id, start_idx, ret_idx, old_next, breakpoint);
},
Edge::Loop { cond, body: lbody, next } => {
let (old_cond, old_lbody, old_next): (usize, usize, Option<usize>) = (*cond, *lbody, *next);
// Update the nexts
*cond += start_idx;
*lbody += start_idx;
if let Some(next) = next {
*next += start_idx;
}
// Traverse the condition...
prep_func_body(edges, calls, func_id, start_idx, ret_idx, old_cond, Some(old_lbody - 1));
// ...the body...
prep_func_body(edges, calls, func_id, start_idx, ret_idx, old_lbody, Some(old_cond));
// ...and finally, the next step, if any
if let Some(old_next) = old_next {
prep_func_body(edges, calls, func_id, start_idx, ret_idx, old_next, breakpoint);
}
},
Edge::Call { next, .. } => {
let old_next: usize = *next;
// Update the next
*next += start_idx;
// Update the call list with this dude's new position
calls.insert(
ProgramCounter::new(FunctionId::Main, start_idx + pc),
*calls.get(&ProgramCounter::new(func_id, pc)).unwrap_or_else(|| panic!("Encountered unresolved call after call ID analysis")),
);
// Prepare the remainder
prep_func_body(edges, calls, func_id, start_idx, ret_idx, old_next, breakpoint);
},
Edge::Return { result: _ } => {
// Yank it
trace!("Yanking return edge at '{pc}' with a linear edge to '{ret_idx}'");
*edge = Edge::Linear { instrs: vec![], next: ret_idx };
},
}
}
/// Inlines the given set of functions in the given WIR function body.
///
/// Note that this is a rather confusing operation space-wise. To prevent program counter pointers from becoming invalid, we simply replace the call with an empty [`Edge::Linear`] that connects to the body appended at the end of the stream. Then, the body connects back to the call's old `next`.
///
/// # Arguments
/// - `body`: A [WIR](Workflow) function body to inline functions _in_.
/// - `calls`: The map of call indices to which function is actually called.
/// - `funcs`: A map of call IDs to function bodies ready to be substituted in the `body`.
/// - `inlinable`: A collection of functions that determines if functions are inlinable. If the set of `deps` is [`Some`], it's inlinable; else it's not.
/// - `func_id`: The ID of the function we're inlining.
/// - `pc`: Points to the current [`Edge`] to analyse.
/// - `breakpoint`: If given, then analysis should stop when this PC is hit.
// It's a compiler function, too many arguments are kinda its thing :P No it's not worth it to come up with structs for this.
#[allow(clippy::too_many_arguments)]
fn inline_funcs_in_body(
body: &mut Vec<Edge>,
calls: &mut HashMap<ProgramCounter, usize>,
funcs: &HashMap<usize, Vec<Edge>>,
inlinable: &HashMap<usize, Option<HashSet<usize>>>,
func_id: FunctionId,
pc: usize,
breakpoint: Option<usize>,
) {
// Stop on the breakpoint
if let Some(breakpoint) = breakpoint {
if pc == breakpoint {
return;
}
}
// Attempt to get the edge
let body_len: usize = body.len();
let edge: &mut Edge = match body.get_mut(pc) {
Some(edge) => edge,
None => return,
};
// Match on its kind
match edge {
Edge::Node { next, .. } | Edge::Linear { next, .. } => {
let next: usize = *next;
inline_funcs_in_body(body, calls, funcs, inlinable, func_id, next, breakpoint)
},
Edge::Stop {} => (),
Edge::Branch { true_next, false_next, merge } => {
let (true_next, false_next, merge): (usize, Option<usize>, Option<usize>) = (*true_next, *false_next, *merge);
// Analyse the left branch...
inline_funcs_in_body(body, calls, funcs, inlinable, func_id, true_next, merge);
// ...the right branch...
if let Some(false_next) = false_next {
inline_funcs_in_body(body, calls, funcs, inlinable, func_id, false_next, merge)
}
// ...and the merge!
if let Some(merge) = merge {
inline_funcs_in_body(body, calls, funcs, inlinable, func_id, merge, breakpoint)
}
},
Edge::Parallel { branches, merge } => {
let (branches, merge): (Vec<usize>, usize) = (branches.clone(), *merge);
// Collect all the branches
for branch in branches {
inline_funcs_in_body(body, calls, funcs, inlinable, func_id, branch, Some(merge));
}
// Run merge and done is Cees
inline_funcs_in_body(body, calls, funcs, inlinable, func_id, merge, breakpoint);
},
Edge::Join { next, .. } => {
let next: usize = *next;
inline_funcs_in_body(body, calls, funcs, inlinable, func_id, next, breakpoint)
},
Edge::Loop { cond, body: lbody, next } => {
let (cond, lbody, next): (usize, usize, Option<usize>) = (*cond, *lbody, *next);
// Traverse the condition...
inline_funcs_in_body(body, calls, funcs, inlinable, func_id, cond, Some(lbody - 1));
// ...the body...
inline_funcs_in_body(body, calls, funcs, inlinable, func_id, lbody, Some(cond));
// ...and finally, the next step, if any
if let Some(next) = next {
inline_funcs_in_body(body, calls, funcs, inlinable, func_id, next, breakpoint);
}
},
Edge::Call { next, .. } => {
let next: usize = *next;
// Resolve the function ID we're calling
let call_id: usize = match calls.get(&ProgramCounter::new(func_id, pc)) {
Some(id) => *id,
None => {
panic!("Encountered unresolved call after running inline analysis");
},
};
// Assert this is an inlinable function (and not external)
if inlinable.get(&call_id).map(|deps| deps.is_none()).unwrap_or(true) {
// Simply skip after doing the next
trace!("Not inlining function call to function {call_id} at {pc}");
inline_funcs_in_body(body, calls, funcs, inlinable, func_id, next, breakpoint);
return;
}
trace!("Inlining function call to function {call_id} at {pc}");
// Otherwise, yank the call with a linear that refers to the inlined body instead (we'll put it after all the other edges to avoid them moving)
// Note: we insert a pop to consume the function reference pushed on the stack to execute the call
*edge = Edge::Linear { instrs: vec![EdgeInstr::Pop {}], next: body_len };
// Prepare the call body by replacing returns with normal links and by bumping all definitions
let mut call_body: Vec<Edge> = funcs
.get(&call_id)
.unwrap_or_else(|| {
panic!("Encountered function ID '{call_id}' without function body after inline analysis (might be an uninlined dependency)")
})
.clone();
prep_func_body(&mut call_body, calls, call_id, body_len, next, 0, None);
// Append it to the main body and the inlining is complete
body.extend(call_body);
// End with the next edges
inline_funcs_in_body(body, calls, funcs, inlinable, func_id, next, breakpoint);
},
Edge::Return { result: _ } => (),
}
}
/***** SIMPLIFICATION FUNCTIONS *****/
/// Attempts to inline functions in the WIR as much as possible.
///
/// The only moment when we don't is if the function call is:
/// - Recursive
/// - A builtin
/// - Undecidable
///
/// # Arguments
/// - `wir`: The input [WIR](Workflow) to simply.
/// - `calls`: The map of call indices to which function is actually called.
///
/// # Returns
/// The same `wir` as given, but then optimized.
///
/// # Errors
/// This function may error if the input workflow is incoherent.
pub fn inline_functions(mut wir: Workflow, calls: &mut HashMap<ProgramCounter, usize>) -> Workflow {
// Analyse which functions in the WIR are non-recursive
let mut inlinable: HashMap<usize, Option<HashSet<usize>>> = HashMap::with_capacity(calls.len());
find_inlinable_funcs(&wir, calls, &mut vec![], ProgramCounter::start(), None, &mut inlinable);
debug!(
"Inlinable functions: {}",
inlinable
.iter()
.filter_map(|(id, deps)| if let Some(deps) = deps {
Some(format!(
"'{}' (depends on {})",
wir.table.funcs.get(*id).map(|def| def.name.as_str()).unwrap_or("???"),
deps.iter()
.map(|id| format!("'{}'", wir.table.funcs.get(*id).map(|def| def.name.as_str()).unwrap_or("???")))
.collect::<Vec<String>>()
.join(", "),
))
} else {
None
})
.collect::<Vec<String>>()
.join(", ")
);
// Order them so that we satisfy function dependencies
let mut inline_order: Vec<usize> = Vec::with_capacity(inlinable.len());
order_inlinable(&mut inline_order, &inlinable, inlinable.keys());
keep_unique_first(&mut inline_order);
debug!(
"Inline order: {}",
inline_order
.iter()
.map(|id| format!("'{}'", wir.table.funcs.get(*id).map(|def| def.name.as_str()).unwrap_or("???"),))
.collect::<Vec<String>>()
.join(", ")
);
{
// Tear open the Workflow to satisfy the borrow checker
let Workflow { id: _, graph: wir_graph, metadata: _, funcs: wir_funcs, table: wir_table, user: _ } = &mut wir;
// Extract the graph behind the Arc
let mut graph: Arc<Vec<Edge>> = Arc::new(vec![]);
std::mem::swap(&mut graph, wir_graph);
let mut graph: Vec<Edge> = Arc::into_inner(graph).unwrap();
// Extract the functions behind the Arc
let mut funcs: Arc<HashMap<usize, Vec<Edge>>> = Arc::new(HashMap::new());
std::mem::swap(&mut funcs, wir_funcs);
let mut funcs: HashMap<usize, Vec<Edge>> = Arc::into_inner(funcs).unwrap();
// Extract the WIR table
let mut table: Arc<SymTable> = Arc::new(SymTable::new());
std::mem::swap(&mut table, wir_table);
let table: SymTable = Arc::into_inner(table).unwrap();
// Inline non-main function bodies first
let mut new_funcs: HashMap<usize, Vec<Edge>> = HashMap::new();
for id in inline_order {
// Acquire the body
let mut new_body: Vec<Edge> = funcs.get(&id).unwrap().clone();
// Inline the functions in this body
debug!("Inlining functions in function {id}");
inline_funcs_in_body(&mut new_body, calls, &new_funcs, &inlinable, FunctionId::Func(id), 0, None);
new_funcs.insert(id, new_body);
}
funcs = new_funcs;
// Now inline the main with all function bodies inlined correctly
debug!("Inlining functions in main");
inline_funcs_in_body(&mut graph, calls, &funcs, &inlinable, FunctionId::Main, 0, None);
// Write the functions and graphs back
let mut table: Arc<SymTable> = Arc::new(table);
std::mem::swap(wir_table, &mut table);
let mut funcs: Arc<HashMap<usize, Vec<Edge>>> = Arc::new(funcs);
std::mem::swap(wir_funcs, &mut funcs);
let mut graph: Arc<Vec<Edge>> = Arc::new(graph);
std::mem::swap(wir_graph, &mut graph);
}
// OK, we did all we could
wir
}
/***** LIBRARY *****/
/// Simplifies the given WIR-workflow as much as possible to increase the compatability with checker workflows.
///
/// Most importantly, it:
/// - Attempts to inline functions as long as they're non-recursive (since functions are not supported)
///
/// # Arguments
/// - `wir`: The input [WIR](Workflow) to simply.
///
/// # Returns
/// A tuple of the same `wir` as given, but then optimized, and a mapping of (remaining) [`Edge::Call`]s to whatever function they actually map.
///
/// # Errors
/// This function may error if the input workflow is incoherent.
pub fn simplify(mut wir: Workflow) -> Result<(Workflow, HashMap<ProgramCounter, usize>), Error> {
// Analyse call dependencies first
let (mut calls, _): (HashMap<ProgramCounter, usize>, _) = resolve_calls(&wir, &wir.table, &mut vec![], ProgramCounter::start(), None, None)?;
debug!("Resolved calls as: {:?}", calls.iter().map(|(pc, id)| (format!("{}", pc.resolved(&wir.table)), *id)).collect::<HashMap<String, usize>>());
// Simplify functions as much as possible
wir = inline_functions(wir, &mut calls);
// Done!
Ok((wir, calls))
}