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 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312
/*! Macros to make using the `engine` module's interface more ergonomic. # Adapton programming model **Adapton roles**: Adapton proposes editor and achivist roles: - **Editor** role: Creates and mutates input, observes (demands) output of incremental computations. - **Archivist** role: Performs cached computations that consume incremental input and produce incremental output. The examples below illustrate these roles, in increasing complexity. **Programming primitives:** The following list of primitives covers the core features of the Adapton engine. Each primitive below is meaningful in each of the two, editor and archivist, roles: - **Ref cell allocation**: Mutable input (editor role), and cached data structures that change across runs (archivist role). - [`cell!`](https://docs.rs/adapton/0/adapton/macro.cell.html) -- Preferred version - [`let_cell!`](https://docs.rs/adapton/0/adapton/macro.let_cell.html) -- Useful in simple examples - [`engine::cell`](https://docs.rs/adapton/0/adapton/engine/fn.cell.html) -- Engine's raw interface - **Observation** and **demand**: Both editor and archivist role. - [`get!`](https://docs.rs/adapton/0/adapton/macro.get.html) -- Preferred version - [`engine::force`](https://docs.rs/adapton/0/adapton/engine/fn.force.html) -- Engine's raw interface - [`engine::force_map`](https://docs.rs/adapton/0/adapton/engine/fn.force_map.html) -- A variant for observations that compose before projections - **Thunk Allocation**: Both editor and archivist role. - Thunk allocation, **_without_ demand**: - [`thunk!`](https://docs.rs/adapton/0/adapton/macro.thunk.html) -- Preferred version - [`let_thunk!`](https://docs.rs/adapton/0/adapton/macro.let_thunk.html) -- Useful in simple examples - [`engine::thunk`](https://docs.rs/adapton/0/adapton/engine/fn.thunk.html) -- Engine's raw interface (can be cumbersome) - Thunk allocation, **_with_ demand**: - [`memo!`](https://docs.rs/adapton/0/adapton/macro.memo.html) -- Preferred version - [`let_memo!`](https://docs.rs/adapton/0/adapton/macro.let_memo.html) -- Useful in simple examples - [`eager!`](https://docs.rs/adapton/0/adapton/macro.eager.html) -- Deprecated; use `memo!` instead. ## Implicit counter for naming `cell`s `cell!(123)` uses a global counter to choose a unique name to hold `123`. Important note: This _may_ be appopriate for the Editor role, but is _never appropriate for the Archivist role_. ``` # #[macro_use] extern crate adapton; # fn main() { # use adapton::macros::*; # use adapton::engine::*; # manage::init_dcg(); let c : Art<usize> = cell!( 123 ); assert_eq!( get!(c), 123 ); assert_eq!( get!(c), force(&c) ); # } ``` ## Explicitly-named `cell`s Sometimes we name a cell using a Rust identifier. We specify this case using the notation `[ name ]`, which specifies that the cell's name is a string, constructed from the Rust identifer `name`: ``` # #[macro_use] extern crate adapton; # fn main() { # use adapton::macros::*; # use adapton::engine::*; # manage::init_dcg(); let c : Art<usize> = cell!([c] 123); assert_eq!(get!(c), 123); assert_eq!(get!(c), force(&c)); # } ``` ## Optionally-named `cell`s Most generally, we supply an expression `optional_name` of type `Option<Name>` to specify the name for the `Art`. This `Art` is created by either `cell` or `put`, in the case that `optional_name` is `Some(name)` or `None`, respectively: ``` # #[macro_use] extern crate adapton; # fn main() { # use adapton::macros::*; # use adapton::engine::*; # manage::init_dcg(); let n : Name = name_of_str(stringify!(c)); let c : Art<usize> = cell!([Some(n)]? 123); assert_eq!(get!(c), 123); assert_eq!(get!(c), force(&c)); let c = cell!([None]? 123); assert_eq!(get!(c), 123); assert_eq!(get!(c), force(&c)); # } ``` Demand-driven change propagation ================================= The example below demonstrates _demand-driven change propagation_, which is unique to Adapton's approach to incremental computation. The example constructs two mutable inputs, `nom` and `den`, an intermediate subcomputation `div` that divides the numerator in `nom` by the denominator in `den`, and a thunk `check` that first checks whether the denominator is zero (returning zero if so) and if non-zero, returns the value of the division. ``` # #[macro_use] extern crate adapton; # fn main() { use adapton::macros::*; use adapton::engine::*; manage::init_dcg(); // Two mutable inputs, for numerator and denominator of division let num = cell!(42); let den = cell!(2); // In Rust, cloning is explicit: let den2 = den.clone(); // clone _global reference_ to cell. let den3 = den.clone(); // clone _global reference_ to cell, again. // Two subcomputations: The division, and a check thunk with a conditional expression let div = thunk![ get!(num) / get!(den) ]; let check = thunk![ if get!(den2) == 0 { None } else { Some(get!(div)) } ]; // Observe output of `check` while we change the input `den` // Step 1: (Explained in detail, below) assert_eq!(get!(check), Some(21)); // Step 2: (Explained in detail, below) set(&den3, 0); assert_eq!(get!(check), None); // Step 3: (Explained in detail, below) set(&den3, 2); assert_eq!(get!(check), Some(21)); // division is reused # } ``` The programmer's changes and observations in the last lines induce the following change propagation behavior: 1. When the `check` is demanded the first time, it executes the condition, and `den` holds `2`, which is non-zero. Hence, the `else` branch executes `get!(div)`, which demands the output of the division, `21`. 2. After this first observation of `check`, the programmer changes `den` to `0`, and re-demands the output of `check`. In response, change propagation first re-executes the condition (not the division), and the condition branches to the `then` branch, resulting in `None`; in particular, it does _not_ re-demand the `div` node, though this node still exists in the DCG. 3. Next, the programmer changes `den` back to its original value, `2`, and re-demands the output of `check`. In response, change propagation re-executes the condition, which re-demands the output of `div`. Change propagation attempts to "clean" the `div` node before re-executing it. To do so, it compares its _last observations_ of `num` and `den` to their current values, of `42` and `2`, respectively. In so doing, it finds that these earlier observations match the current values. Consequently, it _reuses_ the output of the division (`21`) _without_ having to re-execute the division. For a graphical illustration of this behavior, see [these slides](https://github.com/cuplv/adapton-talk/blob/master/adapton-example--div-by-zero/). In the academic literature on Adapton, we refer to this three-step pattern as _switching_: The demand of `div` switches from being present (in step 1) to absent (in step 2) to present (in step 3). Past work on self-adjusting computation does not support this switching pattern directly: Because of its change propagation semantics, it would "forget" the division in step 2, and rerun it _from-scratch_ in step 3. Furthermore, some other change propagation algorithms base their re-execution schedule on "node height" (of the graph's topological ordering). These algorithms may also have undesirable behavior. In particular, they may re-execute the division in step 2, though it is not presently in demand. For an example, see [this gist](https://gist.github.com/khooyp/98abc0e64dc296deaa48). Memoization ============ Memoization provides a mechanism for caching the results of subcomputations; it is a crtical feature of Adapton's approach to incremental computation. In Adapton, each _memoization point_ has three ingredients: - A function expression (of type `Fn`) - Zero or more arguments. Each argument type must have an implementation for the traits `Eq + Clone + Hash + Debug`. The traits `Eq` and `Clone` are both critical to Adapton's caching and change propagation engine. The trait `Hash` is required when Adapton's naming strategy is _structural_ (e.g., where function names are based on the hashes of their arguments). The trait `Debug` is useful for debugging, and reflection. - An optional _name_, which identifies the function call for reuse later. - When this optional name is `None`, the memoization point may be treated in one of two ways: either as just an ordinary, uncached function call, or as a cached function call that is identified _structurally_, by its function pointer and arguments. Adapton permits structural subcomputations via the engine's [structural](https://docs.rs/adapton/0/adapton/engine/fn.structural.html) function. - When this is `Some(name)`, the memoization point uses `name` to identify the work performed by the function call, and its result. Critically, in future incremental runs, it is possible for `name` to associate with different functions and/or argument values. Optional name version ---------------------- The following form is preferred: `memo!( [ optional_name ]? fnexp ; lab1 : arg1, ..., labk : argk )` It accepts an optional name, of type `Option<Name>`, and an arbitrary function expression `fnexp` (closure or function pointer). Like the other forms, it requires that the programmer label each argument. Example ------- ``` # #[macro_use] extern crate adapton; # fn main() { # use adapton::macros::*; # use adapton::engine::*; # manage::init_dcg(); let opnm : Option<Name> = Some(name_unit()); let (t,z) : (Art<usize>, usize) = memo!([opnm]? |x:usize,y:usize|{ if x > y { x } else { y }}; x:10, y:20 ); assert_eq!(z, 20); assert_eq!(force(&t), 20); # } ``` Thunks ------------------------------- The following form is preferred: `thunk!( [ optional_name ]? fnexp ; lab1 : arg1, ..., labk : argk )` It accepts an optional name, of type `Option<Name>`, and an arbitrary function expression `fnexp` (closure or function pointer). Like the other forms, it requires that the programmer label each argument. Example ------- ``` # #[macro_use] extern crate adapton; # fn main() { # use adapton::macros::*; # use adapton::engine::*; # manage::init_dcg(); let opnm : Option<Name> = Some(name_unit()); let t : Art<usize> = thunk!([opnm]? |x:usize,y:usize|{ if x > y { x } else { y }}; x:10, y:20 ); assert_eq!(force(&t), 20); # } ``` Mapping observations makes them more precise ============================================== Suppose that we want to project only one field of type `A` from a pair within an `Art<(A,B)>`. If the field of type `B` changes, our observation of the `A` field will not be affected. Below, we show that using `force_map` prunes the dirtying phase of change propagation. Doing so means that computations that would otherwise be dirty and cleaned via re-execution are never diritied in the first place. We show a simple example of projecting a pair. To observe this fact, this test traces the engine, counts the number of dirtying steps, and ensures that this count is zero, as expected. ``` # #[macro_use] extern crate adapton; # fn main() { use adapton::macros::*; use adapton::engine::*; use adapton::reflect; manage::init_dcg(); // Trace the behavior of change propagation; ensure dirtying works as expected reflect::dcg_reflect_begin(); let pair = cell!((1234, 5678)); let pair1 = pair.clone(); let t = thunk![{ // Project the first component of pair: let fst = force_map(&pair, |_,x| x.0); fst + 100 }]; // The output is `1234 + 100` = `1334` assert_eq!(force(&t), 1334); // Update the second component of the pair; the first is still 1234 set(&pair1, (1234, 8765)); // The output is still `1234 + 100` = `1334` assert_eq!(force(&t), 1334); // Assert that nothing was dirtied (due to using `force_map`) let traces = reflect::dcg_reflect_end(); let counts = reflect::trace::trace_count(&traces, None); assert_eq!(counts.dirty.0, 0); assert_eq!(counts.dirty.1, 0); # } ``` Nominal memoization: Toy Examples =================================== Adapton offers nominal memoization, which uses first-class _names_ (each of type `Name`) to identify cached computations and data. Behind the scenes, these names control how and when the engine _overwrites_ cached data and computations. As such, they permit patterns of programmatic _cache eviction_. For a simple illustration, we memoize several function calls to `sum` with different names and arguments. In real applications, the memoized function typically performs more work than summing two machine words. :) ``` # #[macro_use] extern crate adapton; # fn main() { use adapton::macros::*; use adapton::engine::*; use adapton::reflect; // create an empty DCG (demanded computation graph) manage::init_dcg(); // a simple function (memoized below for illustration purposes; // probably actually not worth it!) fn sum(x:usize, y:usize) -> usize { x + y } // Optional: Traces what the engine does below (for diagnostics, testing, illustration) reflect::dcg_reflect_begin(); let nm_a_0 : Name = name_of_str("a"); // name "a" let nm_a_1 : Name = name_of_str("a"); // name "a" (another copy) let nm_b_0 : Name = name_of_str("b"); // name "b" let nm_b_1 : Name = name_of_str("b"); // name "b" (another copy) // create a memo entry, named "a", that remembers that `sum(42,43) = 85` let res1 : usize = memo!(nm_a_0 =>> sum, x:42, y:43); // same name "a", same arguments (42, 43) => reuses the memo entry above for `res1` let res2 : usize = memo!(nm_a_1 =>> sum, x:42, y:43); // different name "b", same arguments (42, 43) => will *not* match `res1`; creates a new entry let res3 : usize = memo!(nm_b_0 =>> sum, x:42, y:43); // same name "b", different arguments; will *overwrite* entry named "b" with new args & result let res4 : usize = memo!(nm_b_1 =>> sum, x:55, y:66); // Optional: Assert what happened above, in terms of analytical counts let traces = reflect::dcg_reflect_end(); let counts = reflect::trace::trace_count(&traces, None); // Editor allocated two thunks ("a" and "b") assert_eq!(counts.alloc_fresh.0, 2); // Editor allocated one thunk without changing it ("a", with same args) assert_eq!(counts.alloc_nochange.0, 1); // Editor allocated one thunk by changing it ("b", different args) assert_eq!(counts.alloc_change.0, 1); // Archivist allocated nothing assert_eq!(counts.alloc_fresh.1, 0); # drop((res1,res2,res3,res4)); # } ``` Some notes about the code above: - **Callsite argument names**: The macro `memo!` relies on programmer-supplied variable names in its macro expansion of these call sites, shown as `x` and `y` in the uses above. These can be chosen arbitrarily: So long as these symbols are distinct from one another, they can be _any_ symbols, and need not actually match the formal argument names. - **Type arguments**: If the function call expects type arguments, `memo!` accomodates these calls with alternative syntax. - **Spurious arguments**: If the function call expects some later arguments that do not implement `Eq`, but are _functionally determined_ by earlier ones that do (including the supplied `Name`), `memo!` accomodates these calls with alternative syntax. We call these arguments "spurious", since the Adapton engine does _not check_ their identity when performing change propagation. Common examples include function values (e.g., anonymous closures). Nominal Firewalls =================== This example demonstrates how nominal allocation mixes dirtying and cleaning behind the scenes: when the input changes, dirtying proceeds incrementally through the edges of the DCG, _during cleaning_. In some situations (Run 2, below), nominal allocation prevents dirtying from cascading, leading to finer-grained dependency tracking, and more incremental reuse. One might call this design pattern _"nominal firewalls"_ (thanks to @nikomatsakis for suggesting the term "firewall" in this context). First, consider this DCG: ``` // cell +---- Legend ------------------+ // a | [ 2 ] ref cell holding 2 | // [ 2 ] | (g) thunk named 'g' | // ^ | ----> force/observe edge | // | force | --->> allocation edge | // | 2 +------------------------------+ // | // | cell cell // | alloc 4 b force 4 alloc 4 c // (g)------------->>[ 4 ]<--------------(h)-------------->>[ 4 ] // ^ ^ // | force | force h, // | returns b | returns c // | | // (f)------------------------------------+ // ^ // | force f, // | returns cell c // | // (root of demand) ``` In this graph, the ref cell `b` acts as the "firewall". Below, we show a particular input change for cell `a` where a subcomputation `h` is never dirtied nor cleaned by change propagation (input change 2 to -2). We show another change to the same input where this subcomputation `h` *is* _eventually_ dirtied and cleaned by Adapton, though not immediately (input change -2 to 3). Here's the Rust code for generating this DCG, and these changes to its input cell, named `"a"`: ``` # #[macro_use] extern crate adapton; # fn main() { use adapton::macros::*; use adapton::engine::*; fn demand_graph(a: Art<i32>) -> Art<i32> { let_memo!{ c =(f)= { let a = a.clone(); let_memo!{ b =(g)={ let x = get!(a); cell!([b] x * x) }; c =(h)={ let x = get!(b); cell!([c] if x < 100 { x } else { 100 }) }; c }}; c } } manage::init_dcg(); // 1. Initialize input cell "a" to hold 2, and do the computation illustrated above: let c = demand_graph(let_cell!{a = 2; a}); // 2. Change input cell "a" to hold -2, and do the computation illustrated above: let c = demand_graph(let_cell!{a = -2; a}); // 3. Change input cell "a" to hold 3, and do the computation illustrated above: let c = demand_graph(let_cell!{a = 3; a}); # drop(c) # } ``` The `let_memo!` macro above expands as follows: ``` # #[macro_use] extern crate adapton; # fn main() { # use adapton::macros::*; # use adapton::engine::*; fn demand_graph__mid_macro_expansion(a: Art<i32>) -> Art<i32> { let f = let_thunk!{f = { let a = a.clone(); let g = let_thunk!{g = {let x = get!(a); cell!([b] x * x)}; g }; let b = force(&g); let h = let_thunk!{h = {let x = get!(b); cell!([c] if x < 100 { x } else { 100 })}; h }; let c = force(&h); c }; f }; let c = force(&f); c }; # } ``` In this example DCG, thunk `f` allocates and forces two sub-computations, thunks `g` and `h`. The first observes the input `a` and produces an intermediate result (ref cell `b`); the second observes this intermediate result and produces a final result (ref cell `c`), which both thunks `h` and `f` return as their final result. **Run 1.** In the first computation, the input cell `a` holds 2, and the final resulting cell `c` holds `4`. **Run 2.** When the input cell `a` changes, e.g., from 2 to -2, thunks `f` and `g` are dirtied. Thunk `g` is dirty because it observes the changed input. Thunk `f` is dirty because it demanded (observed) the output of thunk `g` in the extent of its own computation. _Importantly, thunk `h` is *not* immediately dirtied when cell `a` changes._ In a sense, cell `a` is an indirect ("transitive") input to thunk `h`. This fact may suggest that when cell `a` is changed from 2 to -2, we should dirty thunk `h` immediately. However, thunk `h` is related to this input only by reading a *different* ref cell (ref cell b) that depends, indirectly, on cell `a`, via the behavior of thunk `g`, on which thunk `h` does *not* directly depend: thunk `h` does not force thunk `g`. Rather, when thunk `f` is re-demanded, Adapton will necessarily perform a cleaning process (aka, "change propagation"), re-executing `g`, its immediate dependent, which is dirty. Since thunk `g` merely squares its input, and 2 and -2 both square to 4, the output of thunk `g` will not change in this case. Consequently, the observers of cell `b`, which holds this output, will not be dirtied or re-executed. In this case, thunk `h` is this observer. In situations like these, Adapton's dirtying + cleaning algorithms do not dirty nor clean thunk `h`. In sum, under this change, after `f` is re-demanded, the cleaning process will first re-execute `g`, the immediate observer of cell `a`. Thunk `g` will again allocate cell `b` to hold 4, the same value as before. It also yields this same cell pointer (to cell `b`). Consequently, thunk `f` is not re-executed, and is cleaned. Meanwhile, the outgoing (dependency) edges thunk of `h` are never dirtied. **Run 3.** For some other change, e.g., from 2 to 3, thunk `h` would _eventually_ be dirtied and cleaned. */ // Adapton uses memoization under the covers, which needs an efficient // mechanism to search for function pointers and compare them for // equality. // // Meanwhile, Rust does not provide Eq and Hash implementations for // trait Fn. So, to identify Rust functions as values that we can // hash and compare, we need to bundle additional static information // along with the function pointer as use this data as a proxy for the // function itself. The idea is that this information uniquely // identifies the function pointer (i.e., two distinct functions will // always have two distinct identities). // use std::cell::RefCell; use std::fmt::{Formatter,Result,Debug}; #[doc(hidden)] pub use std::rc::Rc; thread_local!(static NAME_COUNTER: RefCell<usize> = RefCell::new(0)); #[doc(hidden)] /// Program points: used by the Adapton engine to distinguish different memoized functions. #[derive(PartialEq,Eq,Clone,Hash)] pub struct ProgPt { // Symbolic identity, in Rust semantics: pub symbol:&'static str, // via stringify!(...) // module:Rc<String>, // via module!() // Location in local filesystem: //pub file:&'static str, // via file!() //pub line:u32, // via line!() //pub column:u32, // via column!() } impl Debug for ProgPt { fn fmt(&self, f: &mut Formatter) -> Result { self.symbol.fmt(f) } } #[doc(hidden)] /// Convenience function: A global counter for creating unique names, /// e.g., in unit tests. Avoid using this outside of unit tests. pub fn bump_name_counter() -> usize { NAME_COUNTER.with(|ctr|{let c = *ctr.borrow(); *ctr.borrow_mut() = c + 1; c}) } #[doc(hidden)] /// Generate a "program point", used as a unique ID for memoized functions. #[macro_export] macro_rules! prog_pt { ($symbol:expr) => {{ ProgPt{ symbol:$symbol, //file:file!(), //line:line!(), //column:column!(), } }} } /** Convenience wrapper for `engine::force` Example usage: ``` # #[macro_use] extern crate adapton; # fn main() { # use adapton::macros::*; # use adapton::engine::*; # manage::init_dcg(); let c = cell!(123); assert_eq!(get!(c), 123); assert_eq!(get!(c), force(&c)); # } ``` */ #[macro_export] macro_rules! get { ($art:expr) => {{ force(&($art)) }} } /** Convenience wrappers for `engine::cell`. ## Optional-name version In this verion, supply an expression `optional_name` of type `Option<Name>` to specify the name for the cell, created by either `cell` or `put`, in the case that `optional_name` is `Some(name)` or `None`, respectively: ``` # #[macro_use] extern crate adapton; # fn main() { # use adapton::macros::*; # use adapton::engine::*; # manage::init_dcg(); let n = name_of_str(stringify!(c)); let c = cell!([Some(n)]? 123); assert_eq!(get!(c), 123); assert_eq!(get!(c), force(&c)); let c = cell!([None]? 123); assert_eq!(get!(c), 123); assert_eq!(get!(c), force(&c)); # } ``` ## Explicit-names version: In this verion, use `[ name ]` to specify the cell's name is `name`: ``` # #[macro_use] extern crate adapton; # fn main() { # use adapton::macros::*; # use adapton::engine::*; # manage::init_dcg(); let c = cell!([c] 123); assert_eq!(get!(c), 123); assert_eq!(get!(c), force(&c)); # } ``` ## Global counter version: Uses a global counter to choose a unique name. Important note: This _may_ be appopriate for the Editor role, but is _never appropriate for the Archivist role_. ``` # #[macro_use] extern crate adapton; # fn main() { # use adapton::macros::*; # use adapton::engine::*; # manage::init_dcg(); let c = cell!( 123 ); assert_eq!( get!(c), 123 ); assert_eq!( get!(c), force(&c) ); # } ``` */ #[macro_export] macro_rules! cell { ( $value:expr ) => {{ cell(name_of_usize(bump_name_counter()), $value) }} ; ( [ $nm:expr ] ? $value:expr ) => {{ match $nm { Some(n) => cell(n, $value), None => put($value) } }} ; ( [ $nm:ident ] $value:expr ) => {{ cell(name_of_str(stringify!($nm)), $value) }} } /** Thunks ======================= The following form is preferred: `thunk!( [ optional_name ]? fnexp ; lab1 : arg1, ..., labk : argk )` It accepts an optional name, of type `Option<Name>`, and an arbitrary function expression `fnexp` (closure or function pointer). Like the other forms, it requires that the programmer label each argument. Example ------- ``` # #[macro_use] extern crate adapton; # fn main() { # use adapton::macros::*; # use adapton::engine::*; # manage::init_dcg(); let opnm : Option<Name> = Some(name_unit()); let t : Art<usize> = thunk!([opnm]? |x:usize,y:usize|{ if x > y { x } else { y }}; x:10, y:20 ); assert_eq!(force(&t), 20); # } ``` */ #[macro_export] macro_rules! thunk { ( [ $nmop:expr ] ? $fun:expr ; $( $lab:ident :$arg:expr ),* ) => {{ thunk( match $nmop { None => { ArtIdChoice::Eager }, Some(n) => { ArtIdChoice::Nominal(n) }}, prog_pt!(stringify!($fun)), Rc::new(Box::new( |($($lab),*),()| $fun ( $($lab),* ) )), ( $( $arg ),* ), () ) }} ; [ $suspended_body:expr ] => {{ thunk (ArtIdChoice::Nominal(name_of_usize(bump_name_counter())), prog_pt!(stringify!("anonymous")), Rc::new(Box::new( move |(),()|{ $suspended_body })), (), () ) }} ; [ $nm:ident =>>> $suspended_body:expr ] => {{ thunk!(name_of_str(stringify!($nm)), $suspended_body) }} ; [ $nm:expr =>> $suspended_body:expr ] => {{ thunk (ArtIdChoice::Nominal($nm), prog_pt!(stringify!("anonymous")), Rc::new(Box::new( move |(),()|{ $suspended_body })), (), () ) }} ; ( $nm:expr =>> $f:ident :: < $( $ty:ty ),* > , $( $lab:ident : $arg:expr ),* ) => {{ thunk (ArtIdChoice::Nominal($nm), prog_pt!(stringify!($f)), Rc::new(Box::new( |args, _|{ let ($( $lab ),*, _) = args ; $f :: < $( $ty ),* >( $( $lab ),* ) })), ( $( $arg ),*, ()), () ) }} ; ( $nm:expr =>> $f:path , $( $lab:ident : $arg:expr ),* ) => {{ thunk (ArtIdChoice::Nominal($nm), prog_pt!(stringify!($f)), Rc::new(Box::new( |args, _|{ let ($( $lab ),*, _) = args ; $f ( $( $lab ),* ) })), ( $( $arg ),*, () ), () ) }} ; ( $f:ident :: < $( $ty:ty ),* > , $( $lab:ident : $arg:expr ),* ) => {{ thunk (ArtIdChoice::Structural, prog_pt!(stringify!($f)), Rc::new(Box::new( |args, _|{ let ($( $lab ),*, _) = args ; $f :: < $( $ty ),* >( $( $lab ),* ) })), ( $( $arg ),*, () ), () ) }} ; ( $f:path , $( $lab:ident : $arg:expr ),* ) => {{ thunk (ArtIdChoice::Structural, prog_pt!(stringify!($f)), Rc::new(Box::new( |args, _|{ let ($( $lab ),*, _) = args ; $f ( $( $lab ),* ) })), ( $( $arg ),*, () ), () ) }} ; ( $nm:expr =>> $f:ident =>> < $( $ty:ty ),* > , $( $lab1:ident : $arg1:expr ),* ;; $( $lab2:ident : $arg2:expr ),* ) => {{ let t = thunk (ArtIdChoice::Nominal($nm), prog_pt!(stringify!($f)), Rc::new(Box::new( |args1, args2|{ let ($( $lab1 ),*, _) = args1 ; let ($( $lab2 ),*, _) = args2 ; $f :: < $( $ty ),* > ( $( $lab1 ),* , $( $lab2 ),* ) })), ( $( $arg1 ),*, () ), ( $( $arg2 ),*, () ), ); t }} ; } #[macro_export] macro_rules! fork { { $nmop:expr } => {{ match $nmop { None => (None,None), Some(n) => { let (l,r) = name_fork(n); (Some(l),Some(r)) } } }} ; [ $nm:expr ] => {{ name_fork($nm) }}; } /** Memoization ============ Memoization provides a mechanism for caching the results of subcomputations; it is a crtical feature of Adapton's approach to incremental computation. In Adapton, each _memoization point_ has three ingredients: - A function expression (of type `Fn`) - Zero or more arguments. Each argument type must have an implementation for the traits `Eq + Clone + Hash + Debug`. The traits `Eq` and `Clone` are both critical to Adapton's caching and change propagation engine. The trait `Hash` is required when Adapton's naming strategy is _structural_ (e.g., where function names are based on the hashes of their arguments). The trait `Debug` is useful for debugging, and reflection. - An optional _name_, which identifies the function call for reuse later. - When this optional name is `None`, the memoization point may be treated in one of two ways: either as just an ordinary, uncached function call, or as a cached function call that is identified _structurally_, by its function pointer and arguments. Adapton permits structural subcomputations via the engine's [structural](https://docs.rs/adapton/0/adapton/engine/fn.structural.html) function. - When this is `Some(name)`, the memoization point uses `name` to identify the work performed by the function call, and its result. Critically, in future incremental runs, it is possible for `name` to associate with different functions and/or argument values. Optional name version ---------------------- The following form is preferred: `memo!( [ optional_name ]? fnexp ; lab1 : arg1, ..., labk : argk )` It accepts an optional name, of type `Option<Name>`, and an arbitrary function expression `fnexp` (closure or function pointer). Like the other forms, it requires that the programmer label each argument. Example ------- ``` # #[macro_use] extern crate adapton; # fn main() { # use adapton::macros::*; # use adapton::engine::*; # manage::init_dcg(); let opnm : Option<Name> = Some(name_unit()); let (t,z) : (Art<usize>, usize) = memo!([opnm]? |x:usize,y:usize|{ if x > y { x } else { y }}; x:10, y:20 ); assert_eq!(z, 20); assert_eq!(force(&t), 20); # } ``` */ #[macro_export] macro_rules! memo { ( [ $nmop:expr ] ? $fun:expr ; $( $lab:ident :$arg:expr ),* ) => {{ match $nmop { None => { let x = put( $fun ( $( $arg ),* ) ); let y = force(&x); (x, y) } Some(n) => { let t = thunk!( [ Some(n) ]? $fun ; $( $lab : $arg ),* ); let x = force(&t); (t, x) } } }} ; ( $nm:expr =>> $f:ident :: < $( $ty:ty ),* > , $( $lab:ident : $arg:expr ),* ) => {{ let t = thunk (ArtIdChoice::Nominal($nm), prog_pt!(stringify!($f)), Rc::new(Box::new( |args, _|{ let ($( $lab ),*) = args ; $f :: < $( $ty ),* >( $( $lab ),* ) })), ( $( $arg ),*, ), () ); force(&t) }} ; ( $nm:expr =>> $f:path , $( $lab:ident : $arg:expr ),* ) => {{ let t = thunk (ArtIdChoice::Nominal($nm), prog_pt!(stringify!($f)), Rc::new(Box::new( |args, _|{ let ($( $lab ),*) = args ; $f ( $( $lab ),* ) })), ( $( $arg ),* ), () ); force(&t) }} ; ( $f:ident :: < $( $ty:ty ),* > , $( $lab:ident : $arg:expr ),* ) => {{ let t = thunk (ArtIdChoice::Structural, prog_pt!(stringify!($f)), Rc::new(Box::new( |args, _|{ let ($( $lab ),*) = args ; $f :: < $( $ty ),* >( $( $lab ),* ) })), ( $( $arg ),* ), () ); force(&t) }} ; ( $f:path , $( $lab:ident : $arg:expr ),* ) => {{ let t = thunk (ArtIdChoice::Structural, prog_pt!(stringify!($f)), Rc::new(Box::new( |args, _|{ let ($( $lab ),*, _) = args ; $f ( $( $lab ),* ) })), ( $( $arg ),*, () ), () ); force(&t) }} ; ( $nm:expr =>> $f:path , $( $lab1:ident : $arg1:expr ),* ;; $( $lab2:ident : $arg2:expr ),* ) => {{ let t = thunk (ArtIdChoice::Nominal($nm), prog_pt!(stringify!($f)), Rc::new(Box::new( |args1, args2|{ let ($( $lab1 ),*, _) = args1 ; let ($( $lab2 ),*, _) = args2 ; $f ( $( $lab1 ),* , $( $lab2 ),* ) })), ( $( $arg1 ),*, () ), ( $( $arg2 ),*, () ), ); force(&t) }} ; ( $nm:expr =>> $f:ident =>> < $( $ty:ty ),* > , $( $lab1:ident : $arg1:expr ),* ;; $( $lab2:ident : $arg2:expr ),* ) => {{ let t = thunk (ArtIdChoice::Nominal($nm), prog_pt!(stringify!($f)), Rc::new(Box::new( |args1, args2|{ let ($( $lab1 ),*, _) = args1 ; let ($( $lab2 ),*, _) = args2 ; $f :: < $( $ty ),* > ( $( $lab1 ),* , $( $lab2 ),* ) })), ( $( $arg1 ),*, () ), ( $( $arg2 ),*, () ), ); force(&t) }} ; } /// Similar to `memo!`, except return both the thunk and its observed (`force`d) value. #[macro_export] macro_rules! eager { ( $nm:expr =>> $f:ident :: < $( $ty:ty ),* > , $( $lab:ident : $arg:expr ),* ) => {{ let t = thunk (ArtIdChoice::Nominal($nm), prog_pt!(stringify!($f)), Rc::new(Box::new( |args, _|{ let ($( $lab ),*) = args ; $f :: < $( $ty ),* >( $( $lab ),* ) })), ( $( $arg ),*, ), () ); let res = force(&t) ; (t, res) }} ; ( $nm:expr =>> $f:path , $( $lab:ident : $arg:expr ),* ) => {{ let t = thunk (ArtIdChoice::Nominal($nm), prog_pt!(stringify!($f)), Rc::new(Box::new( |args, _|{ let ($( $lab ),*) = args ; $f ( $( $lab ),* ) })), ( $( $arg ),* ), () ); let res = force(&t) ; (t, res) }} ; ( $f:ident :: < $( $ty:ty ),* > , $( $lab:ident : $arg:expr ),* ) => {{ let t = thunk (ArtIdChoice::Structural, prog_pt!(stringify!($f)), Rc::new(Box::new( |args, _|{ let ($( $lab ),*) = args ; $f :: < $( $ty ),* >( $( $lab ),* ) })), ( $( $arg ),* ), () ); let res = force(&t) ; (t, res) }} ; ( $f:path , $( $lab:ident : $arg:expr ),* ) => {{ let t = thunk (ArtIdChoice::Structural, prog_pt!(stringify!($f)), Rc::new(Box::new( |args, _|{ let ($( $lab ),*, _) = args ; $f ( $( $lab ),* ) })), ( $( $arg ),*, () ), () ); let res = force(&t) ; (t, res) }} ; ( $nm:expr =>> $f:ident =>> < $( $ty:ty ),* > , $( $lab1:ident : $arg1:expr ),* ;; $( $lab2:ident : $arg2:expr ),* ) => {{ let t = thunk (ArtIdChoice::Nominal($nm), prog_pt!(stringify!($f)), Rc::new(Box::new( |args1, args2|{ let ($( $lab1 ),*, _) = args1 ; let ($( $lab2 ),*, _) = args2 ; $f :: < $( $ty ),* > ( $( $lab1 ),* , $( $lab2 ),* ) })), ( $( $arg1 ),*, () ), ( $( $arg2 ),*, () ), ); let res = force(&t) ; (t, res) }} ; } /// Convenience wrapper: Call a function and place the result into an `engine::cell`. #[macro_export] macro_rules! cell_call { ( $nm:expr =>> $f:ident :: < $( $ty:ty ),* > , $( $lab:ident : $arg:expr ),* ) => {{ let res = { $f :: < $( $ty ),* >( $( $arg ),*, ) } ; let cell = cell($nm, res) ; cell }} ; ( $nm:expr =>> $f:ident , $( $lab:ident : $arg:expr ),* ) => {{ let res = { $f ( $( $arg ),*, ) } ; let cell = cell($nm, res) ; cell }} } /** Let-bind a nominal ref cell via `cell`, using the let-bound variable identifier as its name. Permits sequences of bindings. Example usage: [Adapton Example: Nominal firewalls](https://docs.rs/adapton/0/adapton/macros/index.html#nominal-firewalls). */ #[macro_export] macro_rules! let_cell { { $var:ident = $rhs:expr; $body:expr } => {{ { let name = name_of_str(stringify!($var)); let value = $rhs; let $var = cell(name, value); $body } }}; { $var1:ident = $rhs1:expr ; $( $var2:ident = $rhs2:expr ),+ ; $body:expr} => {{ let_cell!($var1 = $rhs1; let_cell!( $( $var2 = $rhs2 ),+ ; $body )) }}; } /** Let-bind a nominal thunk via `thunk!`, without forcing it. Permits sequences of bindings. Example usage: [Adapton Example: Nominal firewalls](https://docs.rs/adapton/0/adapton/macros/index.html#nominal-firewalls). */ #[macro_export] macro_rules! let_thunk { { $var:ident = $rhs:expr; $body:expr } => {{ let name = name_of_str(stringify!($var)); let $var = thunk![name =>> $rhs]; $body }}; { $var1:ident = $rhs1:expr ; $( $var2:ident = $rhs2:expr ),+ ; $body:expr} => {{ let_thunk!($var1 = $rhs1; let_thunk!( $( $var2 = $rhs2 ),+ ; $body )) }}; } #[test] fn test_let_cell_let_thunk_macros() { use adapton::macros::*; use adapton::engine::*; fn demand_graph(a: Art<i32>) -> Art<i32> { let c : Art<i32> = get!(let_thunk!{f = { let a = a.clone(); let b : Art<i32> = get!(let_thunk!{g = {let x = get!(a); let_cell!{b = x * x; b}}; g}); let c : Art<i32> = get!(let_thunk!{h = {let x = get!(b); let_cell!{c = if x < 100 { x } else { 100 }; c}}; h}); c}; f}); return c }; manage::init_dcg(); // 1. Initialize input cell "a" to hold 2, and do the computation illustrated above: let _ = demand_graph(cell(name_of_str("a"), 2)); // 2. Change input cell "a" to hold -2, and do the computation illustrated above: let _ = demand_graph(cell(name_of_str("a"), -2)); // 3. Change input cell "a" to hold 3, and do the computation illustrated above: let _ = demand_graph(cell(name_of_str("a"), 3)); } /** Let-bind a nominal thunk, force it, and let-bind its result. Permits sequences of bindings. Example usage: [Adapton Example: Nominal firewalls](https://docs.rs/adapton/0/adapton/macros/index.html#nominal-firewalls). */ #[macro_export] macro_rules! let_memo { { $var:ident = ( $thkvar1:ident ) = $rhs:expr; $body:expr } => {{ let name = name_of_str(stringify!($thkvar1)); let $thkvar1 = thunk![name =>> $rhs]; let $var = get!($thkvar1); $body }}; { $var1:ident = ( $thkvar1:ident ) = $rhs1:expr ; $( $var2:ident = ( $thkvar2:ident ) = $rhs2:expr ),+ ; $body:expr} => {{ let_memo!($var1 = ( $thkvar1 ) = $rhs1; let_memo!( $( $var2 = ( $thkvar2 ) = $rhs2 ),+ ; $body )) }}; } #[test] fn test_memo_macros() { use adapton::macros::*; use adapton::engine::*; fn demand_graph(a: Art<i32>) -> Art<i32> { let_memo!{c =(f)= { let a = a.clone(); let_memo!{b =(g)= {let x = get!(a); let_cell!{b = x * x; b}}; c =(h)= {let x = get!(b); let_cell!{c = if x < 100 { x } else { 100 }; c}}; c}}; c} } manage::init_dcg(); // 1. Initialize input cell "a" to hold 2, and do the computation illustrated above: let _ = demand_graph(cell(name_of_str("a"), 2)); // 2. Change input cell "a" to hold -2, and do the computation illustrated above: let _ = demand_graph(cell(name_of_str("a"), -2)); // 3. Change input cell "a" to hold 3, and do the computation illustrated above: let _ = demand_graph(cell(name_of_str("a"), 3)); }