├── LICENSE ├── README.md ├── benchmarks.cpp ├── hash_index.hpp ├── tests.cpp └── vs2015 ├── Tests.sln ├── Tests.vcxproj └── Tests.vcxproj.filters /LICENSE: -------------------------------------------------------------------------------- 1 | GNU GENERAL PUBLIC LICENSE 2 | Version 3, 29 June 2007 3 | 4 | Copyright (C) 2007 Free Software Foundation, Inc. 5 | Everyone is permitted to copy and distribute verbatim copies 6 | of this license document, but changing it is not allowed. 7 | 8 | Preamble 9 | 10 | The GNU General Public License is a free, copyleft license for 11 | software and other kinds of works. 12 | 13 | The licenses for most software and other practical works are designed 14 | to take away your freedom to share and change the works. By contrast, 15 | the GNU General Public License is intended to guarantee your freedom to 16 | share and change all versions of a program--to make sure it remains free 17 | software for all its users. We, the Free Software Foundation, use the 18 | GNU General Public License for most of our software; it applies also to 19 | any other work released this way by its authors. You can apply it to 20 | your programs, too. 21 | 22 | When we speak of free software, we are referring to freedom, not 23 | price. Our General Public Licenses are designed to make sure that you 24 | have the freedom to distribute copies of free software (and charge for 25 | them if you wish), that you receive source code or can get it if you 26 | want it, that you can change the software or use pieces of it in new 27 | free programs, and that you know you can do these things. 28 | 29 | To protect your rights, we need to prevent others from denying you 30 | these rights or asking you to surrender the rights. Therefore, you have 31 | certain responsibilities if you distribute copies of the software, or if 32 | you modify it: responsibilities to respect the freedom of others. 33 | 34 | For example, if you distribute copies of such a program, whether 35 | gratis or for a fee, you must pass on to the recipients the same 36 | freedoms that you received. You must make sure that they, too, receive 37 | or can get the source code. And you must show them these terms so they 38 | know their rights. 39 | 40 | Developers that use the GNU GPL protect your rights with two steps: 41 | (1) assert copyright on the software, and (2) offer you this License 42 | giving you legal permission to copy, distribute and/or modify it. 43 | 44 | For the developers' and authors' protection, the GPL clearly explains 45 | that there is no warranty for this free software. For both users' and 46 | authors' sake, the GPL requires that modified versions be marked as 47 | changed, so that their problems will not be attributed erroneously to 48 | authors of previous versions. 49 | 50 | Some devices are designed to deny users access to install or run 51 | modified versions of the software inside them, although the manufacturer 52 | can do so. This is fundamentally incompatible with the aim of 53 | protecting users' freedom to change the software. The systematic 54 | pattern of such abuse occurs in the area of products for individuals to 55 | use, which is precisely where it is most unacceptable. Therefore, we 56 | have designed this version of the GPL to prohibit the practice for those 57 | products. If such problems arise substantially in other domains, we 58 | stand ready to extend this provision to those domains in future versions 59 | of the GPL, as needed to protect the freedom of users. 60 | 61 | Finally, every program is threatened constantly by software patents. 62 | States should not allow patents to restrict development and use of 63 | software on general-purpose computers, but in those that do, we wish to 64 | avoid the special danger that patents applied to a free program could 65 | make it effectively proprietary. To prevent this, the GPL assures that 66 | patents cannot be used to render the program non-free. 67 | 68 | The precise terms and conditions for copying, distribution and 69 | modification follow. 70 | 71 | TERMS AND CONDITIONS 72 | 73 | 0. Definitions. 74 | 75 | "This License" refers to version 3 of the GNU General Public License. 76 | 77 | "Copyright" also means copyright-like laws that apply to other kinds of 78 | works, such as semiconductor masks. 79 | 80 | "The Program" refers to any copyrightable work licensed under this 81 | License. Each licensee is addressed as "you". "Licensees" and 82 | "recipients" may be individuals or organizations. 83 | 84 | To "modify" a work means to copy from or adapt all or part of the work 85 | in a fashion requiring copyright permission, other than the making of an 86 | exact copy. The resulting work is called a "modified version" of the 87 | earlier work or a work "based on" the earlier work. 88 | 89 | A "covered work" means either the unmodified Program or a work based 90 | on the Program. 91 | 92 | To "propagate" a work means to do anything with it that, without 93 | permission, would make you directly or secondarily liable for 94 | infringement under applicable copyright law, except executing it on a 95 | computer or modifying a private copy. Propagation includes copying, 96 | distribution (with or without modification), making available to the 97 | public, and in some countries other activities as well. 98 | 99 | To "convey" a work means any kind of propagation that enables other 100 | parties to make or receive copies. Mere interaction with a user through 101 | a computer network, with no transfer of a copy, is not conveying. 102 | 103 | An interactive user interface displays "Appropriate Legal Notices" 104 | to the extent that it includes a convenient and prominently visible 105 | feature that (1) displays an appropriate copyright notice, and (2) 106 | tells the user that there is no warranty for the work (except to the 107 | extent that warranties are provided), that licensees may convey the 108 | work under this License, and how to view a copy of this License. If 109 | the interface presents a list of user commands or options, such as a 110 | menu, a prominent item in the list meets this criterion. 111 | 112 | 1. Source Code. 113 | 114 | The "source code" for a work means the preferred form of the work 115 | for making modifications to it. "Object code" means any non-source 116 | form of a work. 117 | 118 | A "Standard Interface" means an interface that either is an official 119 | standard defined by a recognized standards body, or, in the case of 120 | interfaces specified for a particular programming language, one that 121 | is widely used among developers working in that language. 122 | 123 | The "System Libraries" of an executable work include anything, other 124 | than the work as a whole, that (a) is included in the normal form of 125 | packaging a Major Component, but which is not part of that Major 126 | Component, and (b) serves only to enable use of the work with that 127 | Major Component, or to implement a Standard Interface for which an 128 | implementation is available to the public in source code form. A 129 | "Major Component", in this context, means a major essential component 130 | (kernel, window system, and so on) of the specific operating system 131 | (if any) on which the executable work runs, or a compiler used to 132 | produce the work, or an object code interpreter used to run it. 133 | 134 | The "Corresponding Source" for a work in object code form means all 135 | the source code needed to generate, install, and (for an executable 136 | work) run the object code and to modify the work, including scripts to 137 | control those activities. However, it does not include the work's 138 | System Libraries, or general-purpose tools or generally available free 139 | programs which are used unmodified in performing those activities but 140 | which are not part of the work. For example, Corresponding Source 141 | includes interface definition files associated with source files for 142 | the work, and the source code for shared libraries and dynamically 143 | linked subprograms that the work is specifically designed to require, 144 | such as by intimate data communication or control flow between those 145 | subprograms and other parts of the work. 146 | 147 | The Corresponding Source need not include anything that users 148 | can regenerate automatically from other parts of the Corresponding 149 | Source. 150 | 151 | The Corresponding Source for a work in source code form is that 152 | same work. 153 | 154 | 2. Basic Permissions. 155 | 156 | All rights granted under this License are granted for the term of 157 | copyright on the Program, and are irrevocable provided the stated 158 | conditions are met. This License explicitly affirms your unlimited 159 | permission to run the unmodified Program. The output from running a 160 | covered work is covered by this License only if the output, given its 161 | content, constitutes a covered work. This License acknowledges your 162 | rights of fair use or other equivalent, as provided by copyright law. 163 | 164 | You may make, run and propagate covered works that you do not 165 | convey, without conditions so long as your license otherwise remains 166 | in force. You may convey covered works to others for the sole purpose 167 | of having them make modifications exclusively for you, or provide you 168 | with facilities for running those works, provided that you comply with 169 | the terms of this License in conveying all material for which you do 170 | not control copyright. Those thus making or running the covered works 171 | for you must do so exclusively on your behalf, under your direction 172 | and control, on terms that prohibit them from making any copies of 173 | your copyrighted material outside their relationship with you. 174 | 175 | Conveying under any other circumstances is permitted solely under 176 | the conditions stated below. Sublicensing is not allowed; section 10 177 | makes it unnecessary. 178 | 179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 180 | 181 | No covered work shall be deemed part of an effective technological 182 | measure under any applicable law fulfilling obligations under article 183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or 184 | similar laws prohibiting or restricting circumvention of such 185 | measures. 186 | 187 | When you convey a covered work, you waive any legal power to forbid 188 | circumvention of technological measures to the extent such circumvention 189 | is effected by exercising rights under this License with respect to 190 | the covered work, and you disclaim any intention to limit operation or 191 | modification of the work as a means of enforcing, against the work's 192 | users, your or third parties' legal rights to forbid circumvention of 193 | technological measures. 194 | 195 | 4. Conveying Verbatim Copies. 196 | 197 | You may convey verbatim copies of the Program's source code as you 198 | receive it, in any medium, provided that you conspicuously and 199 | appropriately publish on each copy an appropriate copyright notice; 200 | keep intact all notices stating that this License and any 201 | non-permissive terms added in accord with section 7 apply to the code; 202 | keep intact all notices of the absence of any warranty; and give all 203 | recipients a copy of this License along with the Program. 204 | 205 | You may charge any price or no price for each copy that you convey, 206 | and you may offer support or warranty protection for a fee. 207 | 208 | 5. Conveying Modified Source Versions. 209 | 210 | You may convey a work based on the Program, or the modifications to 211 | produce it from the Program, in the form of source code under the 212 | terms of section 4, provided that you also meet all of these conditions: 213 | 214 | a) The work must carry prominent notices stating that you modified 215 | it, and giving a relevant date. 216 | 217 | b) The work must carry prominent notices stating that it is 218 | released under this License and any conditions added under section 219 | 7. This requirement modifies the requirement in section 4 to 220 | "keep intact all notices". 221 | 222 | c) You must license the entire work, as a whole, under this 223 | License to anyone who comes into possession of a copy. This 224 | License will therefore apply, along with any applicable section 7 225 | additional terms, to the whole of the work, and all its parts, 226 | regardless of how they are packaged. This License gives no 227 | permission to license the work in any other way, but it does not 228 | invalidate such permission if you have separately received it. 229 | 230 | d) If the work has interactive user interfaces, each must display 231 | Appropriate Legal Notices; however, if the Program has interactive 232 | interfaces that do not display Appropriate Legal Notices, your 233 | work need not make them do so. 234 | 235 | A compilation of a covered work with other separate and independent 236 | works, which are not by their nature extensions of the covered work, 237 | and which are not combined with it such as to form a larger program, 238 | in or on a volume of a storage or distribution medium, is called an 239 | "aggregate" if the compilation and its resulting copyright are not 240 | used to limit the access or legal rights of the compilation's users 241 | beyond what the individual works permit. Inclusion of a covered work 242 | in an aggregate does not cause this License to apply to the other 243 | parts of the aggregate. 244 | 245 | 6. Conveying Non-Source Forms. 246 | 247 | You may convey a covered work in object code form under the terms 248 | of sections 4 and 5, provided that you also convey the 249 | machine-readable Corresponding Source under the terms of this License, 250 | in one of these ways: 251 | 252 | a) Convey the object code in, or embodied in, a physical product 253 | (including a physical distribution medium), accompanied by the 254 | Corresponding Source fixed on a durable physical medium 255 | customarily used for software interchange. 256 | 257 | b) Convey the object code in, or embodied in, a physical product 258 | (including a physical distribution medium), accompanied by a 259 | written offer, valid for at least three years and valid for as 260 | long as you offer spare parts or customer support for that product 261 | model, to give anyone who possesses the object code either (1) a 262 | copy of the Corresponding Source for all the software in the 263 | product that is covered by this License, on a durable physical 264 | medium customarily used for software interchange, for a price no 265 | more than your reasonable cost of physically performing this 266 | conveying of source, or (2) access to copy the 267 | Corresponding Source from a network server at no charge. 268 | 269 | c) Convey individual copies of the object code with a copy of the 270 | written offer to provide the Corresponding Source. This 271 | alternative is allowed only occasionally and noncommercially, and 272 | only if you received the object code with such an offer, in accord 273 | with subsection 6b. 274 | 275 | d) Convey the object code by offering access from a designated 276 | place (gratis or for a charge), and offer equivalent access to the 277 | Corresponding Source in the same way through the same place at no 278 | further charge. You need not require recipients to copy the 279 | Corresponding Source along with the object code. If the place to 280 | copy the object code is a network server, the Corresponding Source 281 | may be on a different server (operated by you or a third party) 282 | that supports equivalent copying facilities, provided you maintain 283 | clear directions next to the object code saying where to find the 284 | Corresponding Source. Regardless of what server hosts the 285 | Corresponding Source, you remain obligated to ensure that it is 286 | available for as long as needed to satisfy these requirements. 287 | 288 | e) Convey the object code using peer-to-peer transmission, provided 289 | you inform other peers where the object code and Corresponding 290 | Source of the work are being offered to the general public at no 291 | charge under subsection 6d. 292 | 293 | A separable portion of the object code, whose source code is excluded 294 | from the Corresponding Source as a System Library, need not be 295 | included in conveying the object code work. 296 | 297 | A "User Product" is either (1) a "consumer product", which means any 298 | tangible personal property which is normally used for personal, family, 299 | or household purposes, or (2) anything designed or sold for incorporation 300 | into a dwelling. In determining whether a product is a consumer product, 301 | doubtful cases shall be resolved in favor of coverage. For a particular 302 | product received by a particular user, "normally used" refers to a 303 | typical or common use of that class of product, regardless of the status 304 | of the particular user or of the way in which the particular user 305 | actually uses, or expects or is expected to use, the product. A product 306 | is a consumer product regardless of whether the product has substantial 307 | commercial, industrial or non-consumer uses, unless such uses represent 308 | the only significant mode of use of the product. 309 | 310 | "Installation Information" for a User Product means any methods, 311 | procedures, authorization keys, or other information required to install 312 | and execute modified versions of a covered work in that User Product from 313 | a modified version of its Corresponding Source. The information must 314 | suffice to ensure that the continued functioning of the modified object 315 | code is in no case prevented or interfered with solely because 316 | modification has been made. 317 | 318 | If you convey an object code work under this section in, or with, or 319 | specifically for use in, a User Product, and the conveying occurs as 320 | part of a transaction in which the right of possession and use of the 321 | User Product is transferred to the recipient in perpetuity or for a 322 | fixed term (regardless of how the transaction is characterized), the 323 | Corresponding Source conveyed under this section must be accompanied 324 | by the Installation Information. But this requirement does not apply 325 | if neither you nor any third party retains the ability to install 326 | modified object code on the User Product (for example, the work has 327 | been installed in ROM). 328 | 329 | The requirement to provide Installation Information does not include a 330 | requirement to continue to provide support service, warranty, or updates 331 | for a work that has been modified or installed by the recipient, or for 332 | the User Product in which it has been modified or installed. Access to a 333 | network may be denied when the modification itself materially and 334 | adversely affects the operation of the network or violates the rules and 335 | protocols for communication across the network. 336 | 337 | Corresponding Source conveyed, and Installation Information provided, 338 | in accord with this section must be in a format that is publicly 339 | documented (and with an implementation available to the public in 340 | source code form), and must require no special password or key for 341 | unpacking, reading or copying. 342 | 343 | 7. Additional Terms. 344 | 345 | "Additional permissions" are terms that supplement the terms of this 346 | License by making exceptions from one or more of its conditions. 347 | Additional permissions that are applicable to the entire Program shall 348 | be treated as though they were included in this License, to the extent 349 | that they are valid under applicable law. If additional permissions 350 | apply only to part of the Program, that part may be used separately 351 | under those permissions, but the entire Program remains governed by 352 | this License without regard to the additional permissions. 353 | 354 | When you convey a copy of a covered work, you may at your option 355 | remove any additional permissions from that copy, or from any part of 356 | it. (Additional permissions may be written to require their own 357 | removal in certain cases when you modify the work.) You may place 358 | additional permissions on material, added by you to a covered work, 359 | for which you have or can give appropriate copyright permission. 360 | 361 | Notwithstanding any other provision of this License, for material you 362 | add to a covered work, you may (if authorized by the copyright holders of 363 | that material) supplement the terms of this License with terms: 364 | 365 | a) Disclaiming warranty or limiting liability differently from the 366 | terms of sections 15 and 16 of this License; or 367 | 368 | b) Requiring preservation of specified reasonable legal notices or 369 | author attributions in that material or in the Appropriate Legal 370 | Notices displayed by works containing it; or 371 | 372 | c) Prohibiting misrepresentation of the origin of that material, or 373 | requiring that modified versions of such material be marked in 374 | reasonable ways as different from the original version; or 375 | 376 | d) Limiting the use for publicity purposes of names of licensors or 377 | authors of the material; or 378 | 379 | e) Declining to grant rights under trademark law for use of some 380 | trade names, trademarks, or service marks; or 381 | 382 | f) Requiring indemnification of licensors and authors of that 383 | material by anyone who conveys the material (or modified versions of 384 | it) with contractual assumptions of liability to the recipient, for 385 | any liability that these contractual assumptions directly impose on 386 | those licensors and authors. 387 | 388 | All other non-permissive additional terms are considered "further 389 | restrictions" within the meaning of section 10. If the Program as you 390 | received it, or any part of it, contains a notice stating that it is 391 | governed by this License along with a term that is a further 392 | restriction, you may remove that term. If a license document contains 393 | a further restriction but permits relicensing or conveying under this 394 | License, you may add to a covered work material governed by the terms 395 | of that license document, provided that the further restriction does 396 | not survive such relicensing or conveying. 397 | 398 | If you add terms to a covered work in accord with this section, you 399 | must place, in the relevant source files, a statement of the 400 | additional terms that apply to those files, or a notice indicating 401 | where to find the applicable terms. 402 | 403 | Additional terms, permissive or non-permissive, may be stated in the 404 | form of a separately written license, or stated as exceptions; 405 | the above requirements apply either way. 406 | 407 | 8. Termination. 408 | 409 | You may not propagate or modify a covered work except as expressly 410 | provided under this License. Any attempt otherwise to propagate or 411 | modify it is void, and will automatically terminate your rights under 412 | this License (including any patent licenses granted under the third 413 | paragraph of section 11). 414 | 415 | However, if you cease all violation of this License, then your 416 | license from a particular copyright holder is reinstated (a) 417 | provisionally, unless and until the copyright holder explicitly and 418 | finally terminates your license, and (b) permanently, if the copyright 419 | holder fails to notify you of the violation by some reasonable means 420 | prior to 60 days after the cessation. 421 | 422 | Moreover, your license from a particular copyright holder is 423 | reinstated permanently if the copyright holder notifies you of the 424 | violation by some reasonable means, this is the first time you have 425 | received notice of violation of this License (for any work) from that 426 | copyright holder, and you cure the violation prior to 30 days after 427 | your receipt of the notice. 428 | 429 | Termination of your rights under this section does not terminate the 430 | licenses of parties who have received copies or rights from you under 431 | this License. If your rights have been terminated and not permanently 432 | reinstated, you do not qualify to receive new licenses for the same 433 | material under section 10. 434 | 435 | 9. Acceptance Not Required for Having Copies. 436 | 437 | You are not required to accept this License in order to receive or 438 | run a copy of the Program. Ancillary propagation of a covered work 439 | occurring solely as a consequence of using peer-to-peer transmission 440 | to receive a copy likewise does not require acceptance. However, 441 | nothing other than this License grants you permission to propagate or 442 | modify any covered work. These actions infringe copyright if you do 443 | not accept this License. Therefore, by modifying or propagating a 444 | covered work, you indicate your acceptance of this License to do so. 445 | 446 | 10. Automatic Licensing of Downstream Recipients. 447 | 448 | Each time you convey a covered work, the recipient automatically 449 | receives a license from the original licensors, to run, modify and 450 | propagate that work, subject to this License. You are not responsible 451 | for enforcing compliance by third parties with this License. 452 | 453 | An "entity transaction" is a transaction transferring control of an 454 | organization, or substantially all assets of one, or subdividing an 455 | organization, or merging organizations. If propagation of a covered 456 | work results from an entity transaction, each party to that 457 | transaction who receives a copy of the work also receives whatever 458 | licenses to the work the party's predecessor in interest had or could 459 | give under the previous paragraph, plus a right to possession of the 460 | Corresponding Source of the work from the predecessor in interest, if 461 | the predecessor has it or can get it with reasonable efforts. 462 | 463 | You may not impose any further restrictions on the exercise of the 464 | rights granted or affirmed under this License. For example, you may 465 | not impose a license fee, royalty, or other charge for exercise of 466 | rights granted under this License, and you may not initiate litigation 467 | (including a cross-claim or counterclaim in a lawsuit) alleging that 468 | any patent claim is infringed by making, using, selling, offering for 469 | sale, or importing the Program or any portion of it. 470 | 471 | 11. Patents. 472 | 473 | A "contributor" is a copyright holder who authorizes use under this 474 | License of the Program or a work on which the Program is based. The 475 | work thus licensed is called the contributor's "contributor version". 476 | 477 | A contributor's "essential patent claims" are all patent claims 478 | owned or controlled by the contributor, whether already acquired or 479 | hereafter acquired, that would be infringed by some manner, permitted 480 | by this License, of making, using, or selling its contributor version, 481 | but do not include claims that would be infringed only as a 482 | consequence of further modification of the contributor version. For 483 | purposes of this definition, "control" includes the right to grant 484 | patent sublicenses in a manner consistent with the requirements of 485 | this License. 486 | 487 | Each contributor grants you a non-exclusive, worldwide, royalty-free 488 | patent license under the contributor's essential patent claims, to 489 | make, use, sell, offer for sale, import and otherwise run, modify and 490 | propagate the contents of its contributor version. 491 | 492 | In the following three paragraphs, a "patent license" is any express 493 | agreement or commitment, however denominated, not to enforce a patent 494 | (such as an express permission to practice a patent or covenant not to 495 | sue for patent infringement). To "grant" such a patent license to a 496 | party means to make such an agreement or commitment not to enforce a 497 | patent against the party. 498 | 499 | If you convey a covered work, knowingly relying on a patent license, 500 | and the Corresponding Source of the work is not available for anyone 501 | to copy, free of charge and under the terms of this License, through a 502 | publicly available network server or other readily accessible means, 503 | then you must either (1) cause the Corresponding Source to be so 504 | available, or (2) arrange to deprive yourself of the benefit of the 505 | patent license for this particular work, or (3) arrange, in a manner 506 | consistent with the requirements of this License, to extend the patent 507 | license to downstream recipients. "Knowingly relying" means you have 508 | actual knowledge that, but for the patent license, your conveying the 509 | covered work in a country, or your recipient's use of the covered work 510 | in a country, would infringe one or more identifiable patents in that 511 | country that you have reason to believe are valid. 512 | 513 | If, pursuant to or in connection with a single transaction or 514 | arrangement, you convey, or propagate by procuring conveyance of, a 515 | covered work, and grant a patent license to some of the parties 516 | receiving the covered work authorizing them to use, propagate, modify 517 | or convey a specific copy of the covered work, then the patent license 518 | you grant is automatically extended to all recipients of the covered 519 | work and works based on it. 520 | 521 | A patent license is "discriminatory" if it does not include within 522 | the scope of its coverage, prohibits the exercise of, or is 523 | conditioned on the non-exercise of one or more of the rights that are 524 | specifically granted under this License. You may not convey a covered 525 | work if you are a party to an arrangement with a third party that is 526 | in the business of distributing software, under which you make payment 527 | to the third party based on the extent of your activity of conveying 528 | the work, and under which the third party grants, to any of the 529 | parties who would receive the covered work from you, a discriminatory 530 | patent license (a) in connection with copies of the covered work 531 | conveyed by you (or copies made from those copies), or (b) primarily 532 | for and in connection with specific products or compilations that 533 | contain the covered work, unless you entered into that arrangement, 534 | or that patent license was granted, prior to 28 March 2007. 535 | 536 | Nothing in this License shall be construed as excluding or limiting 537 | any implied license or other defenses to infringement that may 538 | otherwise be available to you under applicable patent law. 539 | 540 | 12. No Surrender of Others' Freedom. 541 | 542 | If conditions are imposed on you (whether by court order, agreement or 543 | otherwise) that contradict the conditions of this License, they do not 544 | excuse you from the conditions of this License. If you cannot convey a 545 | covered work so as to satisfy simultaneously your obligations under this 546 | License and any other pertinent obligations, then as a consequence you may 547 | not convey it at all. For example, if you agree to terms that obligate you 548 | to collect a royalty for further conveying from those to whom you convey 549 | the Program, the only way you could satisfy both those terms and this 550 | License would be to refrain entirely from conveying the Program. 551 | 552 | 13. Use with the GNU Affero General Public License. 553 | 554 | Notwithstanding any other provision of this License, you have 555 | permission to link or combine any covered work with a work licensed 556 | under version 3 of the GNU Affero General Public License into a single 557 | combined work, and to convey the resulting work. The terms of this 558 | License will continue to apply to the part which is the covered work, 559 | but the special requirements of the GNU Affero General Public License, 560 | section 13, concerning interaction through a network will apply to the 561 | combination as such. 562 | 563 | 14. Revised Versions of this License. 564 | 565 | The Free Software Foundation may publish revised and/or new versions of 566 | the GNU General Public License from time to time. Such new versions will 567 | be similar in spirit to the present version, but may differ in detail to 568 | address new problems or concerns. 569 | 570 | Each version is given a distinguishing version number. If the 571 | Program specifies that a certain numbered version of the GNU General 572 | Public License "or any later version" applies to it, you have the 573 | option of following the terms and conditions either of that numbered 574 | version or of any later version published by the Free Software 575 | Foundation. If the Program does not specify a version number of the 576 | GNU General Public License, you may choose any version ever published 577 | by the Free Software Foundation. 578 | 579 | If the Program specifies that a proxy can decide which future 580 | versions of the GNU General Public License can be used, that proxy's 581 | public statement of acceptance of a version permanently authorizes you 582 | to choose that version for the Program. 583 | 584 | Later license versions may give you additional or different 585 | permissions. However, no additional obligations are imposed on any 586 | author or copyright holder as a result of your choosing to follow a 587 | later version. 588 | 589 | 15. Disclaimer of Warranty. 590 | 591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT 593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY 594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, 595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM 597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF 598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 599 | 600 | 16. Limitation of Liability. 601 | 602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING 603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS 604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY 605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE 606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF 607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), 609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF 610 | SUCH DAMAGES. 611 | 612 | 17. Interpretation of Sections 15 and 16. 613 | 614 | If the disclaimer of warranty and limitation of liability provided 615 | above cannot be given local legal effect according to their terms, 616 | reviewing courts shall apply local law that most closely approximates 617 | an absolute waiver of all civil liability in connection with the 618 | Program, unless a warranty or assumption of liability accompanies a 619 | copy of the Program in return for a fee. 620 | 621 | END OF TERMS AND CONDITIONS 622 | 623 | How to Apply These Terms to Your New Programs 624 | 625 | If you develop a new program, and you want it to be of the greatest 626 | possible use to the public, the best way to achieve this is to make it 627 | free software which everyone can redistribute and change under these terms. 628 | 629 | To do so, attach the following notices to the program. It is safest 630 | to attach them to the start of each source file to most effectively 631 | state the exclusion of warranty; and each file should have at least 632 | the "copyright" line and a pointer to where the full notice is found. 633 | 634 | {one line to give the program's name and a brief idea of what it does.} 635 | Copyright (C) {year} {name of author} 636 | 637 | This program is free software: you can redistribute it and/or modify 638 | it under the terms of the GNU General Public License as published by 639 | the Free Software Foundation, either version 3 of the License, or 640 | (at your option) any later version. 641 | 642 | This program is distributed in the hope that it will be useful, 643 | but WITHOUT ANY WARRANTY; without even the implied warranty of 644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 645 | GNU General Public License for more details. 646 | 647 | You should have received a copy of the GNU General Public License 648 | along with this program. If not, see . 649 | 650 | Also add information on how to contact you by electronic and paper mail. 651 | 652 | If the program does terminal interaction, make it output a short 653 | notice like this when it starts in an interactive mode: 654 | 655 | {project} Copyright (C) {year} {fullname} 656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 657 | This is free software, and you are welcome to redistribute it 658 | under certain conditions; type `show c' for details. 659 | 660 | The hypothetical commands `show w' and `show c' should show the appropriate 661 | parts of the General Public License. Of course, your program's commands 662 | might be different; for a GUI interface, you would use an "about box". 663 | 664 | You should also get your employer (if you work as a programmer) or school, 665 | if any, to sign a "copyright disclaimer" for the program, if necessary. 666 | For more information on this, and how to apply and follow the GNU GPL, see 667 | . 668 | 669 | The GNU General Public License does not permit incorporating your program 670 | into proprietary programs. If your program is a subroutine library, you 671 | may consider it more useful to permit linking proprietary applications with 672 | the library. If this is what you want to do, use the GNU Lesser General 673 | Public License instead of this License. But first, please read 674 | . 675 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | 2 | ## hash_index 3 | 4 | A fast hash table template for array indexes that keeps the hash keys separated from the mapped values. 5 | 6 | `hash_index` is based on a similar data structure used by the DOOM 3 game from id Software. 7 | I have adapted it to be usable as a standalone template class and also added a few 8 | extra customization options and additional helper methods. You can find the original 9 | `idHashIndex` class [in here](https://github.com/id-Software/DOOM-3-BFG/blob/master/neo/idlib/containers/HashIndex.h). 10 | 11 | ---- 12 | 13 | Basic usage example: 14 | 15 | ```cpp 16 | #include "hash_index.hpp" 17 | #include 18 | #include 19 | #include 20 | 21 | struct Player 22 | { 23 | std::string name; 24 | int level; 25 | int health; 26 | // ... 27 | }; 28 | 29 | class PlayerRegistry 30 | { 31 | public: 32 | const Player * findPlayerByName(const std::string & name) const 33 | { 34 | const auto p_index = hash_idx.find(name_hasher(name), name, player_list, 35 | [](const std::string & name, const Player & player) 36 | { 37 | return name == player.name; 38 | }); 39 | if (p_index == hash_idx.null_index) 40 | { 41 | return nullptr; 42 | } 43 | return &player_list[p_index]; 44 | } 45 | 46 | void addPlayer(std::string name, const int lvl, const int hp) 47 | { 48 | hash_idx.insert(name_hasher(name), player_list.size()); 49 | player_list.push_back({ std::move(name), lvl, hp }); 50 | } 51 | 52 | private: 53 | hash_index<> hash_idx; 54 | std::vector player_list; 55 | std::hash name_hasher; 56 | }; 57 | 58 | int main() 59 | { 60 | PlayerRegistry player_reg; 61 | 62 | player_reg.addPlayer("Timmy", 10, 100); 63 | player_reg.addPlayer("K1ll3r", 63, 97 ); 64 | player_reg.addPlayer("1337john", 82, 100); 65 | 66 | assert(player_reg.findPlayerByName("Timmy") != nullptr); 67 | assert(player_reg.findPlayerByName("K1ll3r") != nullptr); 68 | assert(player_reg.findPlayerByName("1337john") != nullptr); 69 | 70 | assert(player_reg.findPlayerByName("bobby") == nullptr); 71 | assert(player_reg.findPlayerByName("johnny") == nullptr); 72 | } 73 | ``` 74 | 75 | You can find more documentation in the `hash_index.hpp` header file. 76 | 77 | ### License 78 | 79 | `hash_index` is released under the terms of the GPL version 3 80 | to comply with the original license used by the code this was based on. 81 | 82 | -------------------------------------------------------------------------------- /benchmarks.cpp: -------------------------------------------------------------------------------- 1 | 2 | // ================================================================================================ 3 | // -*- C++ -*- 4 | // File: benchmarks.cpp 5 | // Author: Guilherme R. Lampert 6 | // Created on: 03/05/16 7 | // Brief: Benchmark tests comparing hash_index with std::map and std::unordered_map. 8 | // ================================================================================================ 9 | 10 | // Compiles with: 11 | // c++ -std=c++11 -Wall -Wextra -Weffc++ -pedantic -O3 benchmarks.cpp -o hash_idx_bench 12 | // 13 | // Asm listing: 14 | // c++ -std=c++11 -S -mllvm --x86-asm-syntax=intel benchmarks.cpp 15 | 16 | #include "hash_index.hpp" 17 | #include 18 | #include 19 | 20 | #include 21 | #include 22 | #include 23 | #include 24 | #include 25 | #include 26 | #include 27 | 28 | // ======================================================== 29 | // Test support code: 30 | // ======================================================== 31 | 32 | using KeyType = std::string; 33 | using ValType = std::pair; 34 | 35 | using Clock = std::chrono::high_resolution_clock; 36 | using TimeUnit = std::chrono::nanoseconds; 37 | using Times = std::vector; 38 | 39 | static const char * TimeUnitSuffix = " ns"; 40 | 41 | // Tells the compiler that all previous changes to memory shall be visible. 42 | inline void clobber_memory() { asm volatile ( "" : : : "memory" ); } 43 | 44 | // Tells the compiler that the value of 'var' shall be visible. 45 | inline void use_variable(void * var) { asm volatile ( "" : : "rm"(var) : ); } 46 | 47 | // ======================================================== 48 | 49 | static KeyType make_random_key() 50 | { 51 | // Called inside the test loops, so we don't want to create a new engine each time. 52 | static std::random_device rand_dev; 53 | static std::mt19937 rand_engine{ rand_dev() }; 54 | 55 | static std::uniform_int_distribution dist_A_Z{ 'A', 'Z' }; 56 | static std::uniform_int_distribution dist_0_9{ '0', '9' }; 57 | 58 | constexpr int key_length = 8; 59 | KeyType key; 60 | 61 | for (int i = 0; i < key_length / 2; ++i) // Half letter from A-Z 62 | { 63 | key.push_back(static_cast(dist_A_Z(rand_engine))); 64 | } 65 | for (int i = 0; i < key_length / 2; ++i) // Half numbers from 0-9 66 | { 67 | key.push_back(static_cast(dist_0_9(rand_engine))); 68 | } 69 | 70 | return key; 71 | } 72 | 73 | static std::vector make_random_key_vector(const long size) 74 | { 75 | std::vector keys; 76 | keys.reserve(size); 77 | for (long i = 0; i < size; ++i) 78 | { 79 | keys.push_back(make_random_key()); 80 | } 81 | return keys; 82 | } 83 | 84 | static void print_test_stats(const Times & times) 85 | { 86 | // Take the median, minimum and maximum: 87 | auto sum = TimeUnit::zero(); 88 | auto minimum = TimeUnit::max(); 89 | auto maximum = TimeUnit::min(); 90 | 91 | for (auto t : times) 92 | { 93 | if (t < minimum) 94 | { 95 | minimum = t; 96 | } 97 | else if (t > maximum) 98 | { 99 | maximum = t; 100 | } 101 | sum += t; 102 | } 103 | 104 | const std::uint64_t mean = sum.count() / times.size(); 105 | const std::uint64_t min = minimum.count(); 106 | const std::uint64_t max = maximum.count(); 107 | 108 | // Now we can print: 109 | std::cout << "\n"; 110 | std::cout << "average time taken...: " << mean << TimeUnitSuffix << "\n"; 111 | std::cout << "lowest time sample...: " << min << TimeUnitSuffix << "\n"; 112 | std::cout << "largest time sample..: " << max << TimeUnitSuffix << "\n"; 113 | std::cout << "\n"; 114 | } 115 | 116 | // ======================================================== 117 | // Key/Value insertion: 118 | // ======================================================== 119 | 120 | template 121 | static void test_insertion_std(const char * const map_type_name, const long num_iterations) 122 | { 123 | std::cout << "\n"; 124 | std::cout << "testing insertions on " << map_type_name << "\n"; 125 | std::cout << num_iterations << " iterations\n"; 126 | 127 | StdMapType test_map; 128 | 129 | Times times; 130 | times.reserve(num_iterations); 131 | 132 | for (long i = 0; i < num_iterations; ++i) 133 | { 134 | const KeyType key{ make_random_key() }; 135 | const ValType val{ i, key }; 136 | 137 | clobber_memory(); 138 | 139 | const auto start = Clock::now(); 140 | 141 | test_map.insert(std::make_pair(key, val)); 142 | 143 | const auto end = Clock::now(); 144 | 145 | use_variable(&test_map); 146 | 147 | times.push_back(end - start); 148 | } 149 | 150 | assert(long(test_map.size()) == num_iterations); 151 | 152 | print_test_stats(times); 153 | std::cout << "----------------------------------\n"; 154 | } 155 | 156 | static void test_insertion_map(const long num_iterations) 157 | { 158 | using MapType = std::map; 159 | test_insertion_std("std::map", num_iterations); 160 | } 161 | 162 | static void test_insertion_unordered_map(const long num_iterations) 163 | { 164 | using MapType = std::unordered_map; 165 | test_insertion_std("std::unordered_map", num_iterations); 166 | } 167 | 168 | static void test_insertion_hash_index(const long num_iterations) 169 | { 170 | std::cout << "\n"; 171 | std::cout << "testing insertions on hash_index + std::vector\n"; 172 | std::cout << num_iterations << " iterations\n"; 173 | 174 | hash_index<> hash_idx; 175 | std::vector values; 176 | 177 | Times times; 178 | times.reserve(num_iterations); 179 | 180 | for (long i = 0; i < num_iterations; ++i) 181 | { 182 | const KeyType key{ make_random_key() }; 183 | const ValType val{ i, key }; 184 | 185 | clobber_memory(); 186 | 187 | // hash_index doesn't bind the values to the keys 188 | // like std::map/unordered_map, so adding to the 189 | // hash_index also implies adding to the value store. 190 | // We also measure the vector::push_back time by design. 191 | // We're also assuming a full copy of the value for 192 | // the worst case usage where a move is not possible. 193 | 194 | const auto start = Clock::now(); 195 | 196 | values.push_back(val); 197 | hash_idx.insert(std::hash{}(key), values.size() - 1); 198 | 199 | const auto end = Clock::now(); 200 | 201 | use_variable(&hash_idx); 202 | use_variable(&values); 203 | 204 | times.push_back(end - start); 205 | } 206 | 207 | assert(long(values.size()) == num_iterations); 208 | 209 | print_test_stats(times); 210 | std::cout << "----------------------------------\n"; 211 | } 212 | 213 | // ======================================================== 214 | // Erasing by key: 215 | // ======================================================== 216 | 217 | template 218 | static void test_erasure_std(const char * const map_type_name, const long num_iterations) 219 | { 220 | std::cout << "\n"; 221 | std::cout << "testing erasures on " << map_type_name << "\n"; 222 | std::cout << num_iterations << " iterations\n"; 223 | 224 | StdMapType test_map; 225 | const auto keys = make_random_key_vector(num_iterations); 226 | 227 | // Fill up the test map: 228 | for (long i = 0; i < num_iterations; ++i) 229 | { 230 | const ValType val{ i, keys[i] }; 231 | test_map.insert(std::make_pair(keys[i], val)); 232 | } 233 | 234 | // Now attempt to erase each key, then measure: 235 | Times times; 236 | times.reserve(num_iterations); 237 | for (long i = 0; i < num_iterations; ++i) 238 | { 239 | clobber_memory(); 240 | 241 | const auto start = Clock::now(); 242 | 243 | test_map.erase(keys[i]); 244 | 245 | const auto end = Clock::now(); 246 | 247 | use_variable(&test_map); 248 | 249 | times.push_back(end - start); 250 | } 251 | 252 | assert(test_map.empty()); 253 | 254 | print_test_stats(times); 255 | std::cout << "----------------------------------\n"; 256 | } 257 | 258 | static void test_erasure_map(const long num_iterations) 259 | { 260 | using MapType = std::map; 261 | test_erasure_std("std::map", num_iterations); 262 | } 263 | 264 | static void test_erasure_unordered_map(const long num_iterations) 265 | { 266 | using MapType = std::unordered_map; 267 | test_erasure_std("std::unordered_map", num_iterations); 268 | } 269 | 270 | static void test_erasure_hash_index(const long num_iterations) 271 | { 272 | std::cout << "\n"; 273 | std::cout << "testing erasures on hash_index + std::vector\n"; 274 | std::cout << num_iterations << " iterations\n"; 275 | 276 | hash_index<> hash_idx; 277 | const auto keys = make_random_key_vector(num_iterations); 278 | 279 | // Fill up the hash_idx: 280 | for (long i = 0; i < num_iterations; ++i) 281 | { 282 | hash_idx.insert(std::hash{}(keys[i]), i); 283 | } 284 | assert(long(hash_idx.index_chain_size()) >= num_iterations); 285 | 286 | // Now attempt to erase each key, then measure: 287 | Times times; 288 | times.reserve(num_iterations); 289 | for (long i = 0; i < num_iterations; ++i) 290 | { 291 | clobber_memory(); 292 | 293 | const auto start = Clock::now(); 294 | 295 | // We are not removing a value from the value store, 296 | // so hash_idx::erase() will always beat the other standard 297 | // maps. In a real life use case, you'd probably also want 298 | // to erase a value from a vector, which will then take 299 | // linear time by itself on the number of elements shifted. 300 | hash_idx.erase(std::hash{}(keys[i]), i); 301 | 302 | const auto end = Clock::now(); 303 | 304 | use_variable(&hash_idx); 305 | 306 | times.push_back(end - start); 307 | } 308 | 309 | print_test_stats(times); 310 | std::cout << "----------------------------------\n"; 311 | } 312 | 313 | // ======================================================== 314 | // Lookup by key: 315 | // ======================================================== 316 | 317 | template 318 | static void test_lookup_std(const char * const map_type_name, const long num_iterations) 319 | { 320 | std::cout << "\n"; 321 | std::cout << "testing lookup on " << map_type_name << "\n"; 322 | std::cout << num_iterations << " iterations\n"; 323 | 324 | StdMapType test_map; 325 | auto keys = make_random_key_vector(num_iterations); 326 | 327 | // Fill up the test map: 328 | for (long i = 0; i < num_iterations; ++i) 329 | { 330 | const ValType val{ i, keys[i] }; 331 | test_map.insert(std::make_pair(keys[i], val)); 332 | } 333 | 334 | // Scramble the keys so lookup is in a random order from the insertion: 335 | std::shuffle(std::begin(keys), std::end(keys), std::mt19937{}); 336 | 337 | // Now attempt to lookup each key, then measure: 338 | Times times; 339 | times.reserve(num_iterations); 340 | for (long i = 0; i < num_iterations; ++i) 341 | { 342 | clobber_memory(); 343 | 344 | const auto start = Clock::now(); 345 | 346 | auto iter = test_map.find(keys[i]); 347 | 348 | const auto end = Clock::now(); 349 | 350 | assert(iter != std::end(test_map)); 351 | 352 | use_variable(&iter); 353 | use_variable(&test_map); 354 | 355 | times.push_back(end - start); 356 | } 357 | 358 | print_test_stats(times); 359 | std::cout << "----------------------------------\n"; 360 | } 361 | 362 | static void test_lookup_map(const long num_iterations) 363 | { 364 | using MapType = std::map; 365 | test_lookup_std("std::map", num_iterations); 366 | } 367 | 368 | static void test_lookup_unordered_map(const long num_iterations) 369 | { 370 | using MapType = std::unordered_map; 371 | test_lookup_std("std::unordered_map", num_iterations); 372 | } 373 | 374 | static void test_lookup_hash_index(const long num_iterations) 375 | { 376 | std::cout << "\n"; 377 | std::cout << "testing lookup on hash_index + std::vector\n"; 378 | std::cout << num_iterations << " iterations\n"; 379 | 380 | hash_index<> hash_idx; 381 | std::vector values; 382 | auto keys = make_random_key_vector(num_iterations); 383 | 384 | // hash_index::find() will need to perform comparisons 385 | // with the lookup key and an item in the value store, 386 | // in case some keys hash to the same bucket. Since the 387 | // key and value types differ, we need this helper predicate. 388 | auto find_predicate = [](const KeyType & key, const ValType & item) 389 | { 390 | return key == item.second; 391 | }; 392 | 393 | // Fill up the test map: 394 | values.reserve(num_iterations); 395 | for (long i = 0; i < num_iterations; ++i) 396 | { 397 | values.push_back({ i, keys[i] }); 398 | hash_idx.insert(std::hash{}(keys[i]), values.size() - 1); 399 | } 400 | assert(long(hash_idx.index_chain_size()) >= num_iterations); 401 | 402 | // Scramble the keys so lookup is in a random order from the insertion: 403 | std::shuffle(std::begin(keys), std::end(keys), std::mt19937{}); 404 | 405 | // Now attempt to lookup each key, then measure: 406 | Times times; 407 | times.reserve(num_iterations); 408 | for (long i = 0; i < num_iterations; ++i) 409 | { 410 | clobber_memory(); 411 | 412 | const auto start = Clock::now(); 413 | 414 | auto index = hash_idx.find(std::hash{}(keys[i]), keys[i], values, find_predicate); 415 | 416 | const auto end = Clock::now(); 417 | 418 | assert(index != hash_idx.null_index); 419 | 420 | use_variable(&index); 421 | use_variable(&hash_idx); 422 | use_variable(&values); 423 | 424 | times.push_back(end - start); 425 | } 426 | 427 | print_test_stats(times); 428 | std::cout << "----------------------------------\n"; 429 | } 430 | 431 | // ======================================================== 432 | // main(): 433 | // ======================================================== 434 | 435 | int main(int argc, const char * argv[]) 436 | { 437 | long num_iterations = 1024; // default value if none provided via cmdline 438 | 439 | if (argc > 1) 440 | { 441 | try 442 | { 443 | num_iterations = std::stol(argv[1]); 444 | } 445 | catch (...) 446 | { 447 | std::cerr << "\nArgument must be a positive integer number!\n"; 448 | std::cerr << "Usage: " << argv[0] << " \n\n"; 449 | return EXIT_FAILURE; 450 | } 451 | } 452 | 453 | // insert() method: 454 | test_insertion_map(num_iterations); 455 | test_insertion_unordered_map(num_iterations); 456 | test_insertion_hash_index(num_iterations); 457 | 458 | // erase() method: 459 | test_erasure_map(num_iterations); 460 | test_erasure_unordered_map(num_iterations); 461 | test_erasure_hash_index(num_iterations); 462 | 463 | // find() method: 464 | test_lookup_map(num_iterations); 465 | test_lookup_unordered_map(num_iterations); 466 | test_lookup_hash_index(num_iterations); 467 | } 468 | 469 | -------------------------------------------------------------------------------- /hash_index.hpp: -------------------------------------------------------------------------------- 1 | 2 | // ================================================================================================ 3 | // -*- C++ -*- 4 | // File: hash_index.hpp 5 | // Author: Guilherme R. Lampert 6 | // Created on: 03/05/16 7 | // 8 | // About: 9 | // hash_index, a fast hash table template for array indexes. 10 | // 11 | // License: 12 | // hash_index is work derived from a similar class found on the source code release of 13 | // DOOM 3 BFG by id Software, available at , 14 | // and therefore is released under the GNU General Public License version 3 to comply 15 | // with the original work. See the accompanying LICENSE file for full disclosure. 16 | // 17 | // ================================================================================================ 18 | 19 | #ifndef HASH_INDEX_HPP 20 | #define HASH_INDEX_HPP 21 | 22 | // Defining this before including the file prevents pulling the Standard headers. 23 | // Useful to be able to place this file inside a user-defined namespace or to simply 24 | // avoid redundant inclusions. User is responsible for providing all the necessary 25 | // Standard headers before #including this one. 26 | #ifndef HASH_INDEX_NO_STD_INCLUDES 27 | #include 28 | #include 29 | #include 30 | #include 31 | #include 32 | #endif // HASH_INDEX_NO_STD_INCLUDES 33 | 34 | // Hook to allow providing a custom assert() before including this file. 35 | #ifndef HASH_INDEX_ASSERT 36 | #ifndef HASH_INDEX_NO_STD_INCLUDES 37 | #include 38 | #endif // HASH_INDEX_NO_STD_INCLUDES 39 | #define HASH_INDEX_ASSERT assert 40 | #endif // HASH_INDEX_ASSERT 41 | 42 | // 43 | // ----------------------- 44 | // hash_index<> template 45 | // ----------------------- 46 | // 47 | // Brief: 48 | // Provides an efficient way of indexing into arrays by a key 49 | // other than the integer element index. hash_index<> allows 50 | // associating a hash key (returned by a function like std::hash) 51 | // with an integer index of an external array where the mapped 52 | // value is stored. 53 | // 54 | // Template arguments: 55 | // IndexType is the type used to store the value array indexes. 56 | // It must be an integer type. By default set to unsigned int 57 | // to save memory on platforms where that type is 32-bits wide, 58 | // since most of the time the full 64-bits indexing range is 59 | // not required. 60 | // 61 | // KeyType is the type of a hash key/hash-code used to 62 | // lookup into hash_index. This type is only used in the 63 | // public interface. By default set to std::size_t to match 64 | // the return value of std::hash. 65 | // 66 | // SizeType is used internally to store array sizes and is also 67 | // visible in the public interface on functions that deal in sizes. 68 | // Set to std::size_t by default to match the Standard library but 69 | // could also be replaced by a 32-bits type to save memory if the 70 | // full 64-bits range is not required. 71 | // 72 | // Allocator is the Standard-compatible allocator used to alloc 73 | // and dealloc the underlaying arrays used by hash_index<>. 74 | // We inherit privately from it to take advantage of the Empty 75 | // Base Class Optimization (EBO) if it is an empty class, so 76 | // if providing a custom one, make sure it is inheritable 77 | // (i.e., not final). By default std::allocator is used. The 78 | // only methods we require are allocate() and deallocate(). 79 | // 80 | // All the template parameters are optional and you can simply 81 | // declare an instance of this class as 'hash_index<>' if the 82 | // defaults described above fit your needs. 83 | // 84 | // ---------------- 85 | // Usage examples 86 | // ---------------- 87 | // 88 | // struct Thing 89 | // { 90 | // std::string name; 91 | // ... 92 | // }; 93 | // 94 | // hash_index<> hash_idx; 95 | // std::vector values; 96 | // 97 | // Insertion: 98 | // 99 | // Thing t = make_thing(); 100 | // values.push_back(t); 101 | // hash_idx.insert(std::hash{}(t.name), values.size() - 1); 102 | // 103 | // Lookup: 104 | // 105 | // std::string key = ...; 106 | // const auto index = hash_idx.find(std::hash{}(key), key, values, 107 | // [](const std::string & key, const Thing & item) 108 | // { 109 | // return key == item.name; 110 | // }); 111 | // index == hash_index<>::null_index if not found, index into 'values[]' otherwise. 112 | // 113 | // Removal: 114 | // 115 | // std::string key = ...; 116 | // unsigned int index = ...; 117 | // hash_idx.erase(std::hash{}(key), index); 118 | // 119 | template 120 | < 121 | typename IndexType = unsigned int, 122 | typename KeyType = std::size_t, 123 | typename SizeType = std::size_t, 124 | typename Allocator = std::allocator 125 | > 126 | class hash_index final 127 | : private Allocator // Take advantage of EBO for the default empty std::allocator 128 | { 129 | public: 130 | 131 | static_assert(std::is_integral::value, "Integer type required for IndexType!"); 132 | static_assert(std::is_integral::value, "Integer type required for KeyType!"); 133 | static_assert(std::is_integral::value, "Integer type required for SizeType!"); 134 | 135 | // 136 | // index_type / null_index: 137 | // 138 | // You can use any signed or unsigned integral type for the underlaying 139 | // index type of hash_index. By default index_type is an unsigned int, so 140 | // on most platforms each index will only take 32-bits. If you need the 141 | // full 64-bits indexing range, then instantiate the class with int64_t 142 | // or equivalent. The bit pattern ~0 (0xFFFFF...) is reserved as an 143 | // internal sentinel value to indicate an unused hash bucket, so this 144 | // value cannot be used as an index. 145 | // 146 | using index_type = IndexType; 147 | static constexpr index_type null_index = ~static_cast(0); 148 | 149 | // 150 | // key_type: 151 | // 152 | // Integral type of the hashed key value. By default it is set to std::size_t 153 | // to match the type returned by std::hash, but you can change it to any other 154 | // integral type, signed or unsigned, that better fits your hash function. 155 | // 156 | using key_type = KeyType; 157 | 158 | // 159 | // size_type: 160 | // 161 | // Type used to represent the internal array counts and sizes. 162 | // Set to std::size_t by default to be consistent with the C++ 163 | // Standard Library, but it is rare the case where you actually 164 | // have a hash_index big enough to overflow a smaller integer type, 165 | // so this can probably be replaced by a 32-bits unsigned integer 166 | // if you care about saving space on each individual class instance. 167 | // 168 | using size_type = SizeType; 169 | 170 | // 171 | // default_initial_size / default_granularity: 172 | // 173 | // Default values used by the non-parameterized constructor 174 | // to set the initial number of hash buckets and size of the 175 | // index chain. Larger sizes reduce collisions and reallocations 176 | // but will of course consume more memory. 177 | // 178 | static constexpr size_type default_initial_size = 1024; 179 | static constexpr size_type default_granularity = 1024; 180 | 181 | // 182 | // Constructors-destructor / copy-assignment: 183 | // 184 | 185 | hash_index() 186 | { 187 | internal_init(default_initial_size, default_initial_size); 188 | } 189 | 190 | hash_index(const size_type initial_hash_buckets_size, 191 | const size_type initial_index_chain_size) 192 | { 193 | internal_init(initial_hash_buckets_size, initial_index_chain_size); 194 | } 195 | 196 | ~hash_index() 197 | { 198 | clear_and_free(); 199 | } 200 | 201 | hash_index(const hash_index & other) 202 | { 203 | // 'other' is empty/never allocated: 204 | if (other.m_lookup_mask == 0) 205 | { 206 | m_hash_buckets = m_invalid_index_dummy; 207 | m_index_chain = m_invalid_index_dummy; 208 | } 209 | else // Copy data from 'other': 210 | { 211 | HASH_INDEX_ASSERT(other.is_allocated()); 212 | HASH_INDEX_ASSERT(is_power_of_two(other.m_hash_buckets_size)); 213 | 214 | m_hash_buckets = Allocator::allocate(other.m_hash_buckets_size); 215 | m_index_chain = Allocator::allocate(other.m_index_chain_size); 216 | 217 | std::copy(other.m_hash_buckets, other.m_hash_buckets + other.m_hash_buckets_size, m_hash_buckets); 218 | std::copy(other.m_index_chain, other.m_index_chain + other.m_index_chain_size, m_index_chain); 219 | } 220 | 221 | m_hash_buckets_size = other.m_hash_buckets_size; 222 | m_index_chain_size = other.m_index_chain_size; 223 | m_hash_mask = other.m_hash_mask; 224 | m_lookup_mask = other.m_lookup_mask; 225 | m_granularity = other.m_granularity; 226 | } 227 | 228 | hash_index & operator = (hash_index other) 229 | { 230 | swap(*this, other); 231 | return *this; 232 | } 233 | 234 | hash_index(hash_index && other) 235 | : hash_index{} // Init via default constructor (C++11 constructor delegation) 236 | { 237 | swap(*this, other); 238 | } 239 | 240 | // Non-throwing swap() overload for hash_index so we 241 | // can enable copy-and-swap in the above constructors. 242 | friend void swap(hash_index & lhs, hash_index & rhs) noexcept 243 | { 244 | using std::swap; 245 | swap(lhs.m_hash_buckets, rhs.m_hash_buckets); 246 | swap(lhs.m_index_chain, rhs.m_index_chain); 247 | swap(lhs.m_hash_buckets_size, rhs.m_hash_buckets_size); 248 | swap(lhs.m_index_chain_size, rhs.m_index_chain_size); 249 | swap(lhs.m_hash_mask, rhs.m_hash_mask); 250 | swap(lhs.m_lookup_mask, rhs.m_lookup_mask); 251 | swap(lhs.m_granularity, rhs.m_granularity); 252 | } 253 | 254 | // 255 | // Lookup: 256 | // 257 | 258 | index_type first(const key_type key) const 259 | { 260 | // 261 | // By ensuring the size of the hash buckets array is always a PoT 262 | // we can use as fast AND (key & m_hash_mask) instead of the slow 263 | // integer modulo to index the table with the incoming hash key. 264 | // 265 | // The lookup mask is either zero for an empty table, which is 266 | // the m_invalid_index_dummy[] in that case, so we always get 267 | // zero from the last AND and return the null_index that 268 | // m_invalid_index_dummy[] holds. Otherwise, when the hash_index table 269 | // is not empty lookup mask has all bits set to 1, so the last AND 270 | // simply yields (key & m_hash_mask), which is the right hash index. 271 | // 272 | return m_hash_buckets[key & m_hash_mask & m_lookup_mask]; 273 | } 274 | 275 | index_type next(const index_type index) const 276 | { 277 | // The index chain is resized when new items are inserted 278 | // to match the largest index, so this check is mostly 279 | // for internal consistency. 280 | HASH_INDEX_ASSERT(static_cast(index) < m_index_chain_size); 281 | return m_index_chain[index & m_lookup_mask]; 282 | } 283 | 284 | template 285 | index_type find(const key_type key, const ValueType & needle, const CollectionType & collection, Predicate pred) const 286 | { 287 | for (index_type i = first(key); i != null_index; i = next(i)) 288 | { 289 | const auto & item = collection[i]; 290 | if (pred(needle, item)) 291 | { 292 | return i; 293 | } 294 | } 295 | return null_index; 296 | } 297 | 298 | template 299 | index_type find(const key_type key, const ValueType & needle, const CollectionType & collection) const 300 | { 301 | return find(key, needle, collection, std::equal_to{}); 302 | } 303 | 304 | // 305 | // Insertion / removal: 306 | // 307 | 308 | void insert(const key_type key, const index_type index) 309 | { 310 | if (!is_allocated()) 311 | { 312 | const size_type index_chain_size = ((static_cast(index) >= m_index_chain_size) ? 313 | index + 1 : m_index_chain_size); 314 | internal_allocate(m_hash_buckets_size, index_chain_size); 315 | } 316 | else if (index >= m_index_chain_size) 317 | { 318 | resize_index_chain(index + 1); 319 | } 320 | 321 | const key_type k = key & m_hash_mask; 322 | m_index_chain[index] = m_hash_buckets[k]; 323 | m_hash_buckets[k] = index; 324 | } 325 | 326 | void erase(const key_type key, const index_type index) 327 | { 328 | HASH_INDEX_ASSERT(static_cast(index) < m_index_chain_size); 329 | 330 | if (!is_allocated()) 331 | { 332 | return; 333 | } 334 | 335 | const key_type k = key & m_hash_mask; 336 | 337 | if (m_hash_buckets[k] == index) 338 | { 339 | m_hash_buckets[k] = m_index_chain[index]; 340 | } 341 | else 342 | { 343 | for (index_type i = m_hash_buckets[k]; i != null_index; i = m_index_chain[i]) 344 | { 345 | if (m_index_chain[i] == index) 346 | { 347 | m_index_chain[i] = m_index_chain[index]; 348 | break; 349 | } 350 | } 351 | } 352 | 353 | m_index_chain[index] = null_index; 354 | } 355 | 356 | // Insert an entry into the index chain and add it to the hash, increasing all indexes >= index. 357 | void insert_at_index(const key_type key, const index_type index) 358 | { 359 | if (!is_allocated()) 360 | { 361 | return; 362 | } 363 | 364 | size_type i; 365 | index_type max = index; 366 | 367 | for (i = 0; i < m_hash_buckets_size; ++i) 368 | { 369 | if (m_hash_buckets[i] >= index) 370 | { 371 | m_hash_buckets[i]++; 372 | if (m_hash_buckets[i] > max) 373 | { 374 | max = m_hash_buckets[i]; 375 | } 376 | } 377 | } 378 | 379 | for (i = 0; i < m_index_chain_size; ++i) 380 | { 381 | if (m_index_chain[i] >= index) 382 | { 383 | m_index_chain[i]++; 384 | if (m_index_chain[i] > max) 385 | { 386 | max = m_index_chain[i]; 387 | } 388 | } 389 | } 390 | 391 | if (max >= m_index_chain_size) 392 | { 393 | resize_index_chain(max + 1); 394 | } 395 | for (i = max; i > index; --i) 396 | { 397 | m_index_chain[i] = m_index_chain[i - 1]; 398 | } 399 | m_index_chain[index] = null_index; 400 | 401 | insert(key, index); 402 | } 403 | 404 | // Remove an entry from the index chain and remove it from the hash, decreasing all indexes >= index. 405 | void erase_and_remove_index(const key_type key, const index_type index) 406 | { 407 | HASH_INDEX_ASSERT(static_cast(index) < m_index_chain_size); 408 | 409 | if (!is_allocated()) 410 | { 411 | return; 412 | } 413 | 414 | erase(key, index); 415 | 416 | size_type i; 417 | index_type max = index; 418 | 419 | for (i = 0; i < m_hash_buckets_size; ++i) 420 | { 421 | if (m_hash_buckets[i] >= index) 422 | { 423 | if (m_hash_buckets[i] > max) 424 | { 425 | max = m_hash_buckets[i]; 426 | } 427 | m_hash_buckets[i]--; 428 | } 429 | } 430 | 431 | for (i = 0; i < m_index_chain_size; ++i) 432 | { 433 | if (m_index_chain[i] >= index) 434 | { 435 | if (m_index_chain[i] > max) 436 | { 437 | max = m_index_chain[i]; 438 | } 439 | m_index_chain[i]--; 440 | } 441 | } 442 | 443 | for (i = index; i < max; ++i) 444 | { 445 | m_index_chain[i] = m_index_chain[i + 1]; 446 | } 447 | m_index_chain[max] = null_index; 448 | } 449 | 450 | // 451 | // Memory management: 452 | // 453 | 454 | void clear() noexcept 455 | { 456 | if (m_hash_buckets != m_invalid_index_dummy) 457 | { 458 | const index_type fill_val = null_index; 459 | std::fill_n(m_hash_buckets, m_hash_buckets_size, fill_val); 460 | } 461 | // Clearing the index chain is not strictly necessary since 462 | // inserting new elements in the hash_index will overwrite 463 | // corresponding index chain entries. 464 | } 465 | 466 | void clear_and_resize(const size_type new_hash_buckets_size, const size_type new_index_chain_size) 467 | { 468 | HASH_INDEX_ASSERT(is_power_of_two(new_hash_buckets_size) && "Size of hash_index buckets array must be a power-of-2!"); 469 | 470 | clear_and_free(); 471 | m_hash_buckets_size = new_hash_buckets_size; 472 | m_index_chain_size = new_index_chain_size; 473 | } 474 | 475 | void clear_and_free() 476 | { 477 | if (m_hash_buckets != m_invalid_index_dummy) 478 | { 479 | Allocator::deallocate(m_hash_buckets, m_hash_buckets_size); 480 | m_hash_buckets = m_invalid_index_dummy; 481 | } 482 | if (m_index_chain != m_invalid_index_dummy) 483 | { 484 | Allocator::deallocate(m_index_chain, m_index_chain_size); 485 | m_index_chain = m_invalid_index_dummy; 486 | } 487 | m_lookup_mask = 0; 488 | } 489 | 490 | void set_granularity(const size_type new_granularity) 491 | { 492 | HASH_INDEX_ASSERT(new_granularity > 0); 493 | m_granularity = new_granularity; 494 | } 495 | 496 | void resize_index_chain(const size_type new_index_chain_size) 497 | { 498 | if (new_index_chain_size <= m_index_chain_size) 499 | { 500 | return; 501 | } 502 | 503 | size_type new_size; 504 | const auto mod = new_index_chain_size % m_granularity; 505 | 506 | if (mod == 0) 507 | { 508 | new_size = new_index_chain_size; 509 | } 510 | else 511 | { 512 | new_size = new_index_chain_size + m_granularity - mod; 513 | } 514 | 515 | if (m_index_chain == m_invalid_index_dummy) // Not allocated yet; Defer. 516 | { 517 | m_index_chain_size = new_size; 518 | return; 519 | } 520 | 521 | const auto old_index_chain_size = m_index_chain_size; 522 | const auto old_index_chain = m_index_chain; 523 | 524 | auto new_index_chain = Allocator::allocate(new_size); 525 | std::copy(old_index_chain, old_index_chain + old_index_chain_size, new_index_chain); 526 | 527 | // Newly allocated space must be filled with null_index 528 | const index_type fill_val = null_index; 529 | std::fill_n(new_index_chain + old_index_chain_size, new_size - old_index_chain_size, fill_val); 530 | 531 | Allocator::deallocate(old_index_chain, old_index_chain_size); 532 | m_index_chain = new_index_chain; 533 | m_index_chain_size = new_size; 534 | } 535 | 536 | // 537 | // Queries: 538 | // 539 | 540 | size_type compute_distribution_percentage() const 541 | { 542 | // Computes a number in the range [0,100] representing the spread over the hash table. 543 | // This method will allocate some temporary memory via std::vector. 544 | 545 | if (!is_allocated()) 546 | { 547 | return 100; 548 | } 549 | 550 | long total_items = 0; 551 | std::vector hash_items_count(m_hash_buckets_size); 552 | 553 | for (size_type i = 0; i < m_hash_buckets_size; ++i) 554 | { 555 | hash_items_count[i] = 0; 556 | for (index_type index = m_hash_buckets[i]; index != null_index; index = m_index_chain[index]) 557 | { 558 | hash_items_count[i]++; 559 | } 560 | total_items += hash_items_count[i]; 561 | } 562 | 563 | // If no items in the hash buckets... 564 | if (total_items <= 1) 565 | { 566 | return 100; 567 | } 568 | 569 | long error = 0; 570 | const long average = total_items / m_hash_buckets_size; 571 | 572 | for (size_type i = 0; i < m_hash_buckets_size; ++i) 573 | { 574 | long e = hash_items_count[i] - average; 575 | if (e < 0) { e = -e; } // absolute value of 'e' 576 | 577 | if (e > 1) 578 | { 579 | error += (e - 1); 580 | } 581 | } 582 | 583 | return 100 - (error * 100 / total_items); 584 | } 585 | 586 | size_type allocated_bytes() const noexcept 587 | { 588 | if (!is_allocated()) 589 | { 590 | return 0; 591 | } 592 | return (m_hash_buckets_size * sizeof(index_type)) + 593 | (m_index_chain_size * sizeof(index_type)); 594 | } 595 | 596 | size_type hash_buckets_size() const noexcept 597 | { 598 | return m_hash_buckets_size; 599 | } 600 | 601 | size_type index_chain_size() const noexcept 602 | { 603 | return m_index_chain_size; 604 | } 605 | 606 | size_type granularity() const noexcept 607 | { 608 | return m_granularity; 609 | } 610 | 611 | bool is_allocated() const noexcept 612 | { 613 | return (m_hash_buckets != nullptr) && 614 | (m_hash_buckets != m_invalid_index_dummy); 615 | } 616 | 617 | // 618 | // Deep comparison operators: 619 | // 620 | 621 | bool operator == (const hash_index & other) const noexcept 622 | { 623 | // Self comparison of comparison to pointer/ref to self? 624 | if (this == &other) 625 | { 626 | return true; 627 | } 628 | 629 | // Same sizes? 630 | if (m_hash_buckets_size != other.m_hash_buckets_size) { return false; } 631 | if (m_index_chain_size != other.m_index_chain_size ) { return false; } 632 | if (m_hash_mask != other.m_hash_mask ) { return false; } 633 | if (m_lookup_mask != other.m_lookup_mask ) { return false; } 634 | if (m_granularity != other.m_granularity ) { return false; } 635 | 636 | // This or other could be pointing to the m_invalid_index_dummy. 637 | if ( is_allocated() && !other.is_allocated()) { return false; } 638 | if (!is_allocated() && other.is_allocated()) { return false; } 639 | 640 | // Same sizes, but do both have the same data? 641 | for (size_type i = 0; i < m_hash_buckets_size; ++i) 642 | { 643 | if (m_hash_buckets[i] != other.m_hash_buckets[i]) 644 | { 645 | return false; 646 | } 647 | } 648 | for (size_type i = 0; i < m_index_chain_size; ++i) 649 | { 650 | if (m_index_chain[i] != other.m_index_chain[i]) 651 | { 652 | return false; 653 | } 654 | } 655 | 656 | // The same sizes and contents. 657 | return true; 658 | } 659 | 660 | bool operator != (const hash_index & other) const noexcept 661 | { 662 | return !(*this == other); 663 | } 664 | 665 | private: 666 | 667 | template 668 | static bool is_power_of_two(const IntType num) noexcept 669 | { 670 | static_assert(std::is_integral::value, "Integer type required!"); 671 | return (num > 0) && ((num & (num - 1)) == 0); 672 | } 673 | 674 | void internal_init(const size_type initial_hash_buckets_size, 675 | const size_type initial_index_chain_size) 676 | { 677 | HASH_INDEX_ASSERT(is_power_of_two(initial_hash_buckets_size) && "Size of hash_index buckets array must be a power-of-2!"); 678 | HASH_INDEX_ASSERT(!is_allocated() && "Already initialized!"); 679 | 680 | m_hash_buckets = m_invalid_index_dummy; 681 | m_index_chain = m_invalid_index_dummy; 682 | m_hash_buckets_size = initial_hash_buckets_size; 683 | m_index_chain_size = initial_index_chain_size; 684 | m_hash_mask = m_hash_buckets_size - 1; 685 | m_lookup_mask = 0; 686 | m_granularity = default_granularity; 687 | } 688 | 689 | void internal_allocate(const size_type new_hash_buckets_size, 690 | const size_type new_index_chain_size) 691 | { 692 | HASH_INDEX_ASSERT(is_power_of_two(new_hash_buckets_size) && "Size of hash_index buckets array must be a power-of-2!"); 693 | 694 | if (is_allocated()) 695 | { 696 | clear_and_free(); 697 | } 698 | 699 | m_hash_buckets = Allocator::allocate(new_hash_buckets_size); 700 | m_index_chain = Allocator::allocate(new_index_chain_size); 701 | m_hash_buckets_size = new_hash_buckets_size; 702 | m_index_chain_size = new_index_chain_size; 703 | m_hash_mask = m_hash_buckets_size - 1; 704 | m_lookup_mask = ~static_cast(0); 705 | 706 | const index_type fill_val = null_index; 707 | std::fill_n(m_hash_buckets, new_hash_buckets_size, fill_val); 708 | std::fill_n(m_index_chain, new_index_chain_size, fill_val); 709 | } 710 | 711 | // 712 | // m_hash_buckets[] is indexed directly from an incoming key ANDed with 713 | // m_hash_mask (which is just the cached result of m_hash_buckets_size - 1). 714 | // The resulting index will be either the index of the requested data or 715 | // an index into m_index_chain[]. Index chain will then either have the 716 | // index of the requested data or a successive chain of indexes to itself 717 | // until the data index is found or a null_index is encountered. 718 | // See first() and next() for more details on the lookup algorithm. 719 | // 720 | // m_hash_buckets[] can be of any power-of-two size (PoT is enforced). 721 | // m_index_chain[] must match the size of the external data array and 722 | // will be resized accordingly to accommodate the largest index inserted. 723 | // 724 | // Note: When the hash_index is empty, both m_hash_buckets and m_index_chain 725 | // will point to the m_invalid_index_dummy below. This allows simplifying some 726 | // things and avoiding additional IF checks for the empty hash_index case. 727 | // 728 | index_type * m_hash_buckets = nullptr; 729 | index_type * m_index_chain = nullptr; 730 | size_type m_hash_buckets_size = 0; 731 | size_type m_index_chain_size = 0; 732 | 733 | // 734 | // As mentioned above, m_hash_mask is the cached result of m_hash_buckets_size - 1. 735 | // Since we require a power-of-two size for the hash buckets array, we can use a 736 | // much cheaper AND to modulate the incoming hash key, but we need to AND it with 737 | // the num of buckets - 1, hence this cached value. 738 | // 739 | size_type m_hash_mask = 0; 740 | 741 | // 742 | // m_lookup_mask is a bit of magic... We use it to make the methods first() and 743 | // next() branch-less. When the hash_index is empty, the buckets and index chain 744 | // will point to the m_invalid_index_dummy[] array. In this case, we always want 745 | // the incoming hash key to map to index zero, so m_lookup_mask is set to zero 746 | // when hash_index is empty. It will get ANDed with the incoming hash key in 747 | // first() and next() yielding the 0th index of m_invalid_index_dummy[] (null_index), 748 | // so we've successfully managed to avoid an extra IF there for the empty case. 749 | // But when the table is not empty, buckets and the index chain will point to 750 | // heap-allocated arrays, so ANDing m_lookup_mask with a hash key should not 751 | // alter the resulting array index. Therefore, when the table is not empty, 752 | // m_lookup_mask is set to the bit pattern ~0 (0xFFFFF... or all 1s) so that 753 | // ANDing anything with it will just yield back the original input. 754 | // 755 | size_type m_lookup_mask = 0; 756 | 757 | // 758 | // Factor used to resize the index chain on demand. 759 | // Works best if using a power-of-two, but not required. 760 | // 761 | size_type m_granularity = 0; 762 | 763 | // 764 | // The initial empty hash_index allocates no heap memory, but to simplify 765 | // handling of the empty case we still want the hash buckets and 766 | // index chain to point to some array with at least one null_index 767 | // in it. That's where this placeholder comes in. The data in it 768 | // will not be modified by the hash_index, but since the m_hash_buckets 769 | // and m_index_chain pointers are not const, this has to be kept mutable. 770 | // 771 | static index_type m_invalid_index_dummy[1]; 772 | }; 773 | 774 | // This abomination here is the m_invalid_index_dummy[] static initializer: 775 | template 776 | typename hash_index::index_type hash_index::m_invalid_index_dummy[1] = { 777 | hash_index::null_index 778 | }; 779 | 780 | #endif // HASH_INDEX_HPP 781 | -------------------------------------------------------------------------------- /tests.cpp: -------------------------------------------------------------------------------- 1 | 2 | // ================================================================================================ 3 | // -*- C++ -*- 4 | // File: tests.cpp 5 | // Author: Guilherme R. Lampert 6 | // Created on: 03/05/16 7 | // Brief: Minimal unit tests for the hash_index. 8 | // ================================================================================================ 9 | 10 | // Compiles with: 11 | // c++ -std=c++11 -Wall -Wextra -Weffc++ -pedantic -O3 tests.cpp -o hash_idx_tests 12 | 13 | #include "hash_index.hpp" 14 | 15 | #include 16 | #include 17 | #include 18 | #include 19 | #include 20 | #include 21 | 22 | // ======================================================== 23 | // Basic tests (relying on assert(), so leave it on!): 24 | // ======================================================== 25 | 26 | template 27 | static void fill_random_keys(HashIndexType * out_hash_idx, std::vector * out_hash_keys = nullptr) 28 | { 29 | constexpr std::size_t count = 1024; 30 | 31 | static std::random_device rand_dev; 32 | static std::mt19937 rand_engine{ rand_dev() }; 33 | 34 | std::hash hasher; 35 | for (std::size_t i = 0; i < count; ++i) 36 | { 37 | const auto key = hasher(i) ^ (hasher(rand_engine()) << 1u); 38 | 39 | out_hash_idx->insert(static_cast(key), 40 | static_cast(i)); 41 | 42 | if (out_hash_keys != nullptr) 43 | { 44 | out_hash_keys->push_back(key); 45 | } 46 | } 47 | } 48 | 49 | template 50 | static void test_copy_move_assign() 51 | { 52 | HashIndexType h1; 53 | assert(h1.is_allocated() == false); 54 | assert(h1.allocated_bytes() == 0); 55 | 56 | fill_random_keys(&h1); 57 | assert(h1.is_allocated() == true); 58 | assert(h1.allocated_bytes() != 0); 59 | 60 | // Copy constructor: 61 | HashIndexType h2{ h1 }; 62 | assert(h2.is_allocated() == true); 63 | assert(h2.allocated_bytes() == h1.allocated_bytes()); 64 | assert(h2.index_chain_size() == h1.index_chain_size()); 65 | assert(h2.hash_buckets_size() == h1.hash_buckets_size()); 66 | assert(h2 == h1); // Deep comparison 67 | assert(h1 == h2); // Deep comparison 68 | 69 | // Move constructor (steal h1): 70 | HashIndexType h3{ std::move(h1) }; 71 | 72 | // h3 should now be equal to h2 (which is a copy of former h1): 73 | assert(h3.is_allocated() == true); 74 | assert(h3.allocated_bytes() == h2.allocated_bytes()); 75 | assert(h3.index_chain_size() == h2.index_chain_size()); 76 | assert(h3.hash_buckets_size() == h2.hash_buckets_size()); 77 | assert(h3 == h2); 78 | assert(h2 == h3); 79 | 80 | // h1 must now be an empty hash_index: 81 | assert(h1.is_allocated() == false); 82 | assert(h1.allocated_bytes() == 0); 83 | 84 | // Assignment operator: 85 | HashIndexType h4; 86 | assert(h4.is_allocated() == false); 87 | assert(h4.allocated_bytes() == 0); 88 | 89 | h4 = h3; 90 | assert(h4.is_allocated() == true); 91 | assert(h4.allocated_bytes() != 0); 92 | 93 | assert(h4 == h3); 94 | assert(h3 == h4); 95 | } 96 | 97 | template 98 | static void test_insertion() 99 | { 100 | HashIndexType h1; 101 | std::vector keys; 102 | 103 | // fill will do the insertions: 104 | fill_random_keys(&h1, &keys); 105 | 106 | // now we just validate the inserted keys are there: 107 | for (auto k : keys) 108 | { 109 | assert(h1.first(static_cast(k)) != h1.null_index); 110 | } 111 | } 112 | 113 | template 114 | static void test_erasure() 115 | { 116 | HashIndexType h1; 117 | std::vector keys; 118 | 119 | // fill with some sample keys: 120 | fill_random_keys(&h1, &keys); 121 | 122 | // and erase them: 123 | for (std::size_t i = 0; i < keys.size(); ++i) 124 | { 125 | h1.erase(static_cast(keys[i]), 126 | static_cast(i)); 127 | } 128 | 129 | // now first() should yield null_index for all keys. 130 | for (auto k : keys) 131 | { 132 | assert(h1.first(static_cast(k)) == h1.null_index); 133 | } 134 | } 135 | 136 | template 137 | static void test_lookup() 138 | { 139 | HashIndexType h1; 140 | std::vector keys; 141 | std::vector found_keys; 142 | 143 | // fill with some sample keys: 144 | fill_random_keys(&h1, &keys); 145 | found_keys.resize(keys.size(), false); 146 | 147 | // and find them: 148 | for (std::size_t k = 0; k < keys.size(); ++k) 149 | { 150 | for (auto i = h1.first(static_cast(keys[k])); i != h1.null_index; i = h1.next(i)) 151 | { 152 | if (static_cast(i) == k) 153 | { 154 | found_keys[k] = true; 155 | break; 156 | } 157 | } 158 | } 159 | 160 | // check all found: 161 | for (std::size_t i = 0; i < found_keys.size(); ++i) 162 | { 163 | assert(found_keys[i] == true); 164 | } 165 | } 166 | 167 | template 168 | static void test_key_collisions() 169 | { 170 | HashIndexType h1; 171 | 172 | constexpr std::size_t count = 1024; 173 | constexpr std::size_t key = 0xCAFED00D; 174 | 175 | // insert a bunch of indexes at the same key/bucket: 176 | for (std::size_t i = 0; i < count; ++i) 177 | { 178 | h1.insert(static_cast(key), 179 | static_cast(i)); 180 | } 181 | 182 | // Ensure they are still reachable: 183 | std::vector found_keys(count, false); 184 | 185 | for (std::size_t k = 0; k < count; ++k) 186 | { 187 | for (auto i = h1.first(key); i != h1.null_index; i = h1.next(i)) 188 | { 189 | if (static_cast(i) == k) 190 | { 191 | found_keys[k] = true; 192 | break; 193 | } 194 | } 195 | } 196 | 197 | for (std::size_t i = 0; i < found_keys.size(); ++i) 198 | { 199 | assert(found_keys[i] == true); 200 | } 201 | } 202 | 203 | // ======================================================== 204 | // main() - Test driver: 205 | // ======================================================== 206 | 207 | // Should work all the same regardless of the integer types used. 208 | using HashIndexDefault = hash_index<>; 209 | using HashIndexSInt32 = hash_index; 210 | using HashIndexUInt32 = hash_index; 211 | using HashIndexSInt64 = hash_index; 212 | using HashIndexUInt64 = hash_index; 213 | 214 | // Calls the same test once for each specialized hash_index instantiation. 215 | #define TEST(func) \ 216 | do \ 217 | { \ 218 | std::cout << "> Testing " << #func << "...\n"; \ 219 | test_##func(); \ 220 | test_##func(); \ 221 | test_##func(); \ 222 | test_##func(); \ 223 | test_##func(); \ 224 | std::cout << "> " << #func << " test completed!\n\n"; \ 225 | } \ 226 | while (0,0) 227 | 228 | int main() 229 | { 230 | std::cout << "\nRunning unit tests for hash_index...\n\n"; 231 | 232 | TEST(copy_move_assign); 233 | TEST(insertion); 234 | TEST(erasure); 235 | TEST(lookup); 236 | TEST(key_collisions); 237 | 238 | std::cout << "All tests passed!\n\n"; 239 | } 240 | -------------------------------------------------------------------------------- /vs2015/Tests.sln: -------------------------------------------------------------------------------- 1 |  2 | Microsoft Visual Studio Solution File, Format Version 12.00 3 | # Visual Studio 14 4 | VisualStudioVersion = 14.0.25420.1 5 | MinimumVisualStudioVersion = 10.0.40219.1 6 | Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "Tests", "Tests.vcxproj", "{48A2EED8-19A3-4772-91E2-888708932EA6}" 7 | EndProject 8 | Global 9 | GlobalSection(SolutionConfigurationPlatforms) = preSolution 10 | Debug|x64 = Debug|x64 11 | Debug|x86 = Debug|x86 12 | Release|x64 = Release|x64 13 | Release|x86 = Release|x86 14 | EndGlobalSection 15 | GlobalSection(ProjectConfigurationPlatforms) = postSolution 16 | {48A2EED8-19A3-4772-91E2-888708932EA6}.Debug|x64.ActiveCfg = Debug|x64 17 | {48A2EED8-19A3-4772-91E2-888708932EA6}.Debug|x64.Build.0 = Debug|x64 18 | {48A2EED8-19A3-4772-91E2-888708932EA6}.Debug|x86.ActiveCfg = Debug|Win32 19 | {48A2EED8-19A3-4772-91E2-888708932EA6}.Debug|x86.Build.0 = Debug|Win32 20 | {48A2EED8-19A3-4772-91E2-888708932EA6}.Release|x64.ActiveCfg = Release|x64 21 | {48A2EED8-19A3-4772-91E2-888708932EA6}.Release|x64.Build.0 = Release|x64 22 | {48A2EED8-19A3-4772-91E2-888708932EA6}.Release|x86.ActiveCfg = Release|Win32 23 | {48A2EED8-19A3-4772-91E2-888708932EA6}.Release|x86.Build.0 = Release|Win32 24 | EndGlobalSection 25 | GlobalSection(SolutionProperties) = preSolution 26 | HideSolutionNode = FALSE 27 | EndGlobalSection 28 | EndGlobal 29 | -------------------------------------------------------------------------------- /vs2015/Tests.vcxproj: -------------------------------------------------------------------------------- 1 |  2 | 3 | 4 | 5 | Debug 6 | Win32 7 | 8 | 9 | Release 10 | Win32 11 | 12 | 13 | Debug 14 | x64 15 | 16 | 17 | Release 18 | x64 19 | 20 | 21 | 22 | {48A2EED8-19A3-4772-91E2-888708932EA6} 23 | Win32Proj 24 | Tests 25 | 8.1 26 | 27 | 28 | 29 | Application 30 | true 31 | v140 32 | Unicode 33 | 34 | 35 | Application 36 | false 37 | v140 38 | true 39 | Unicode 40 | 41 | 42 | Application 43 | true 44 | v140 45 | Unicode 46 | 47 | 48 | Application 49 | false 50 | v140 51 | true 52 | Unicode 53 | 54 | 55 | 56 | 57 | 58 | 59 | 60 | 61 | 62 | 63 | 64 | 65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | true 74 | 75 | 76 | true 77 | 78 | 79 | false 80 | 81 | 82 | false 83 | 84 | 85 | 86 | 87 | 88 | Level4 89 | Disabled 90 | WIN32;_DEBUG;_CONSOLE;%(PreprocessorDefinitions) 91 | 92 | 93 | Console 94 | true 95 | 96 | 97 | 98 | 99 | 100 | 101 | Level4 102 | Disabled 103 | _DEBUG;_CONSOLE;%(PreprocessorDefinitions) 104 | 105 | 106 | Console 107 | true 108 | 109 | 110 | 111 | 112 | Level4 113 | 114 | 115 | MaxSpeed 116 | true 117 | true 118 | WIN32;NDEBUG;_CONSOLE;%(PreprocessorDefinitions) 119 | 120 | 121 | Console 122 | true 123 | true 124 | true 125 | 126 | 127 | 128 | 129 | Level4 130 | 131 | 132 | MaxSpeed 133 | true 134 | true 135 | NDEBUG;_CONSOLE;%(PreprocessorDefinitions) 136 | 137 | 138 | Console 139 | true 140 | true 141 | true 142 | 143 | 144 | 145 | 146 | 147 | 148 | 149 | 150 | 151 | 152 | 153 | -------------------------------------------------------------------------------- /vs2015/Tests.vcxproj.filters: -------------------------------------------------------------------------------- 1 |  2 | 3 | 4 | 5 | {4FC737F1-C7A5-4376-A066-2A32D752A2FF} 6 | cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx 7 | 8 | 9 | 10 | 11 | Source Files 12 | 13 | 14 | 15 | 16 | Source Files 17 | 18 | 19 | --------------------------------------------------------------------------------