├── .gitignore ├── LICENSE ├── README.md ├── requirements.txt ├── setup.py └── src └── pydags ├── __init__.py ├── cache.py ├── pipeline.py ├── serialization.py └── stage.py /.gitignore: -------------------------------------------------------------------------------- 1 | .idea/ 2 | 3 | # Byte-compiled / optimized / DLL files 4 | __pycache__/ 5 | *.py[cod] 6 | *$py.class 7 | 8 | # C extensions 9 | *.so 10 | 11 | # Distribution / packaging 12 | .Python 13 | build/ 14 | develop-eggs/ 15 | dist/ 16 | downloads/ 17 | eggs/ 18 | .eggs/ 19 | lib/ 20 | lib64/ 21 | parts/ 22 | sdist/ 23 | var/ 24 | wheels/ 25 | pip-wheel-metadata/ 26 | share/python-wheels/ 27 | *.egg-info/ 28 | .installed.cfg 29 | *.egg 30 | MANIFEST 31 | 32 | # PyInstaller 33 | # Usually these files are written by a python script from a template 34 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 35 | *.manifest 36 | *.spec 37 | 38 | # Installer logs 39 | pip-log.txt 40 | pip-delete-this-directory.txt 41 | 42 | # Unit test / coverage reports 43 | htmlcov/ 44 | .tox/ 45 | .nox/ 46 | .coverage 47 | .coverage.* 48 | .cache 49 | nosetests.xml 50 | coverage.xml 51 | *.cover 52 | *.py,cover 53 | .hypothesis/ 54 | .pytest_cache/ 55 | 56 | # Translations 57 | *.mo 58 | *.pot 59 | 60 | # Django stuff: 61 | *.log 62 | local_settings.py 63 | db.sqlite3 64 | db.sqlite3-journal 65 | 66 | # Flask stuff: 67 | instance/ 68 | .webassets-cache 69 | 70 | # Scrapy stuff: 71 | .scrapy 72 | 73 | # Sphinx documentation 74 | docs/_build/ 75 | 76 | # PyBuilder 77 | target/ 78 | 79 | # Jupyter Notebook 80 | .ipynb_checkpoints 81 | 82 | # IPython 83 | profile_default/ 84 | ipython_config.py 85 | 86 | # pyenv 87 | .python-version 88 | 89 | # pipenv 90 | # According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control. 91 | # However, in case of collaboration, if having platform-specific dependencies or dependencies 92 | # having no cross-platform support, pipenv may install dependencies that don't work, or not 93 | # install all needed dependencies. 94 | #Pipfile.lock 95 | 96 | # PEP 582; used by e.g. github.com/David-OConnor/pyflow 97 | __pypackages__/ 98 | 99 | # Celery stuff 100 | celerybeat-schedule 101 | celerybeat.pid 102 | 103 | # SageMath parsed files 104 | *.sage.py 105 | 106 | # Environments 107 | .env 108 | .venv 109 | env/ 110 | venv/ 111 | ENV/ 112 | env.bak/ 113 | venv.bak/ 114 | 115 | # Spyder project settings 116 | .spyderproject 117 | .spyproject 118 | 119 | # Rope project settings 120 | .ropeproject 121 | 122 | # mkdocs documentation 123 | /site 124 | 125 | # mypy 126 | .mypy_cache/ 127 | .dmypy.json 128 | dmypy.json 129 | 130 | # Pyre type checker 131 | .pyre/ 132 | -------------------------------------------------------------------------------- /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 | 635 | Copyright (C) 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 | Copyright (C) 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 | # pydags 2 | 3 | **pydags** is a Python package to facilitate the creation and running of 4 | lightweight DAG-based workloads locally. Whereas technologies like Airflow, 5 | Kubeflow, and Luigi are more heavyweight, enterprise-level workflow managers, 6 | **pydags** is instead an extensible, simple, lightweight alternative tailored 7 | for local development and execution. There are no dependencies on Docker, 8 | Kubernetes, or any other such technologies, and it is use-case agnostic (unlike 9 | Kubeflow/Luigi). 10 | 11 | ## Terminology 12 | 13 | ### Stage 14 | 15 | A *Stage* is synonymous to a node in a DAG. 16 | 17 | ### Pipeline 18 | 19 | A *Pipeline* is synonymous to a DAG, and is comprised of 1 or more *Stages* in 20 | a DAG-like structure. 21 | 22 | ## Get Started 23 | 24 | ### External Dependencies 25 | 26 | **pydags** requires multiple non-Python dependencies in order to function properly. 27 | These include Redis (for internal operation) and GraphViz (for visualization). To install these, run the following command in terminal: 28 | 29 | ```bash 30 | sudo apt-get install redis graphviz 31 | ``` 32 | 33 | Installing redis with this command on most *nix systems will result in the 34 | Redis server starting automatically. You can verify this by running the 35 | `redis-server` command, which should result in an `Address already in use` 36 | message, or similar. 37 | 38 | ### Install pydags 39 | 40 | To install **pydags**, simply run the following command: 41 | 42 | ```bash 43 | pip install pydags 44 | ``` 45 | 46 | ## Example Usage 47 | 48 | ### Expressing a DAG 49 | 50 | Below is a simple example that aims to simply demonstrate how to specify DAGs 51 | in **pydags**. In this case, each stage of the pipeline is a Python function 52 | decorated with the `@stage` decorator. 53 | 54 | ```python 55 | from pydags.pipeline import Pipeline 56 | from pydags.stage import stage 57 | 58 | 59 | @stage 60 | def stage_1(): 61 | print('Running stage 1') 62 | 63 | @stage 64 | def stage_2(): 65 | print('Running stage 2') 66 | 67 | @stage 68 | def stage_3(): 69 | print('Running stage 3') 70 | 71 | @stage 72 | def stage_4(): 73 | print('Running stage 4') 74 | 75 | @stage 76 | def stage_5(): 77 | print('Running stage 5') 78 | 79 | @stage 80 | def stage_6(): 81 | print('Running stage 6') 82 | 83 | def build_pipeline(): 84 | stage1 = stage_1() 85 | stage2 = stage_2() 86 | stage3 = stage_3().after(stage2) 87 | stage4 = stage_4().after(stage2) 88 | stage5 = stage_5().after(stage1) 89 | stage6 = stage_6().after(stage3).after(stage4).after(stage5) 90 | 91 | pipeline = Pipeline() 92 | 93 | pipeline.add_stages([ 94 | stage1, stage2, stage3, 95 | stage4, stage5, stage6 96 | ]) 97 | 98 | return pipeline 99 | 100 | 101 | pipeline = build_pipeline() 102 | 103 | pipeline.visualize() 104 | 105 | pipeline.start() 106 | ``` 107 | 108 | Stages of the pipeline that can be run in parallel (in the above case, stages 1 109 | and 2, and stages 3, 4, 5) will only be run in parallel if you set the 110 | `num_cores` argument of the `.start()` method to a positive integer 111 | (representing the number of cores to distribute computation across). For 112 | example, if you want to parallelize the execution of such nodes that can be run 113 | in parallel, then simply replace `pipeline.start()` with 114 | `pipeline.start(num_cores=8)` (to use 8 cores). 115 | 116 | ### A Simple ML Pipeline 117 | 118 | Below is an example of a simple ML pipeline consisting of 3 stages: 1) data 119 | download, 2) preprocessing, 3) model training. All 3 stages are subclasses of 120 | `RedisStage`, and thus inherit the functionality to read from and write to a 121 | Redis server. The data is thus passed between the stages using Redis. You may 122 | subclass DiskCacheStage, or implement your own cache/storage backend, if Redis 123 | is not suited to your use case. 124 | 125 | Please note that you will need to install the additional `scikit-learn` 126 | dependency to run this example. 127 | 128 | ```python 129 | import redis 130 | from sklearn.datasets import load_wine 131 | from sklearn.ensemble import RandomForestClassifier 132 | from sklearn.model_selection import train_test_split 133 | from sklearn.preprocessing import MinMaxScaler 134 | 135 | from pydags.pipeline import Pipeline 136 | from pydags.stage import RedisStage 137 | 138 | 139 | class DataIngestor(RedisStage): 140 | @staticmethod 141 | def download_data(): 142 | data = load_wine() 143 | features = data['data'] 144 | targets = data['target'] 145 | return features, targets 146 | 147 | def run(self, *args, **kwargs): 148 | features, targets = self.download_data() 149 | self.write('features', features) 150 | self.write('targets', targets) 151 | 152 | 153 | class DataPreprocessor(RedisStage): 154 | @staticmethod 155 | def normalize(features): 156 | return MinMaxScaler().fit_transform(features) 157 | 158 | @staticmethod 159 | def split(features, targets): 160 | return train_test_split(features, targets, test_size=0.2) 161 | 162 | def run(self, *args, **kwargs): 163 | features = self.read('features') 164 | targets = self.read('targets') 165 | 166 | xtr, xte, ytr, yte = self.split(features, targets) 167 | 168 | xtr = self.normalize(xtr) 169 | xte = self.normalize(xte) 170 | 171 | data = { 172 | 'xtr': xtr, 'xte': xte, 173 | 'ytr': ytr, 'yte': yte 174 | } 175 | 176 | self.write('preprocessed_data', data) 177 | 178 | 179 | class ModelTrainer(RedisStage): 180 | def __init__(self, *args, **kwargs): 181 | super(ModelTrainer, self).__init__(*args, **kwargs) 182 | 183 | self.model = None 184 | 185 | def train_model(self, xtr, ytr): 186 | self.model = RandomForestClassifier().fit(xtr, ytr) 187 | 188 | def test_model(self, xte, yte): 189 | acc = self.model.score(xte, yte) 190 | return acc 191 | 192 | def run(self, *args, **kwargs): 193 | preprocessed_data = self.read('preprocessed_data') 194 | 195 | xtr = preprocessed_data['xtr'] 196 | xte = preprocessed_data['xte'] 197 | ytr = preprocessed_data['ytr'] 198 | yte = preprocessed_data['yte'] 199 | 200 | self.train_model(xtr, ytr) 201 | 202 | acc = self.test_model(xte, yte) 203 | 204 | print('Accuracy:', acc) 205 | 206 | 207 | def build_pipeline(): 208 | redis_instance = redis.Redis(host='localhost', port=6379, db=0) 209 | 210 | data_ingestor = DataIngestor(redis_instance=redis_instance) 211 | 212 | data_preprocessor = DataPreprocessor(redis_instance=redis_instance).after(data_ingestor) 213 | 214 | model_trainer = ModelTrainer(redis_instance=redis_instance).after(data_preprocessor) 215 | 216 | pipeline = Pipeline() 217 | pipeline.add_stages([data_ingestor, data_preprocessor, model_trainer]) 218 | 219 | return pipeline 220 | 221 | 222 | p = build_pipeline() 223 | 224 | p.visualize() 225 | 226 | p.start() 227 | ``` 228 | -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | pydot==1.4.2 2 | networkx==2.5 3 | redis==3.5.3 4 | diskcache==5.2.1 5 | matplotlib>=3.3.3 6 | dill==0.3.3 7 | cloudpickle==1.6.0 -------------------------------------------------------------------------------- /setup.py: -------------------------------------------------------------------------------- 1 | import setuptools 2 | 3 | with open("README.md", "r", encoding="utf-8") as fh: 4 | long_description = fh.read() 5 | 6 | setuptools.setup( 7 | name="pydags", 8 | version="0.1.1", 9 | author="David Torpey", 10 | author_email="torpey.david93@gmail.com", 11 | description="A simple, lightweight, extensible DAG framework for Python", 12 | long_description=long_description, 13 | long_description_content_type="text/markdown", 14 | url="https://github.com/DavidTorpey/pydags", 15 | project_urls={ 16 | "Bug Tracker": "https://github.com/DavidTorpey/pydags/issues", 17 | }, 18 | classifiers=[ 19 | "Programming Language :: Python :: 3", 20 | ], 21 | package_dir={"": "src"}, 22 | packages=setuptools.find_packages(where="src"), 23 | python_requires=">=3.6", 24 | license='GPLv3', 25 | install_requires=open('requirements.txt').read().splitlines() 26 | ) -------------------------------------------------------------------------------- /src/pydags/__init__.py: -------------------------------------------------------------------------------- 1 | """ 2 | This initialisation script it called the first time pydags is imported. 3 | Currently, the only purpose of the script is to set up the root logger. 4 | """ 5 | 6 | import logging 7 | 8 | logging.basicConfig( 9 | format='pydags: %(name)s - %(levelname)s - %(message)s', level=logging.INFO 10 | ) 11 | -------------------------------------------------------------------------------- /src/pydags/cache.py: -------------------------------------------------------------------------------- 1 | """ 2 | This module contains the base classes and implementations for caches to be used 3 | by both the pipeline internals during DAG execution, as well as for users to 4 | potentially subclass. Any cache must implement as least 'read', 'write', and 5 | 'delete' methods. 6 | 7 | The two caches implemented here are a in-memory cache powered by Redis, and a 8 | local disk cache. 9 | 10 | The Redis cache is used by the pipeline to read/write data pertaining to the 11 | execution of the pipeline and its constituent stages (e.g. stages that are in 12 | progress, completed, etc.). 13 | 14 | The disk cache is not used by the pipeline, but it, along with the Redis cache, 15 | can be subclassed by a user to inherent the associated caching functionality. 16 | The most common use case for doing so is to pass data throughout your DAG. 17 | 18 | References: 19 | https://pypi.org/project/diskcache/ 20 | https://redis.io/ 21 | """ 22 | 23 | from abc import ABC, abstractmethod 24 | 25 | import diskcache 26 | import redis 27 | 28 | 29 | class Cache(ABC): 30 | """ 31 | Abstract base class for all cache implementations. ALl subclasses must 32 | implement a read, write, and delete method. 33 | """ 34 | 35 | @abstractmethod 36 | def read(self, *args, **kwargs): 37 | ... 38 | 39 | @abstractmethod 40 | def write(self, *args, **kwargs): 41 | ... 42 | 43 | @abstractmethod 44 | def delete(self, *args, **kwargs): 45 | ... 46 | 47 | 48 | class InvalidCacheTypeException(Exception): 49 | pass 50 | 51 | 52 | class InvalidKeyTypeException(Exception): 53 | pass 54 | 55 | 56 | class InvalidValueTypeException(Exception): 57 | pass 58 | 59 | 60 | class RedisCache(Cache): 61 | """ 62 | Implementation for an in-memory cache to be used by the Pipeline 63 | implementation itself, as well as potentially by users who wish to inherit 64 | this functionality. The in-memory caching technology used is the key-value 65 | store Redis. 66 | 67 | The use of the cache assumes redis has been installed and is running. 68 | """ 69 | 70 | def __init__(self, redis_instance: redis.Redis): 71 | if not isinstance(redis_instance, redis.Redis): 72 | raise InvalidCacheTypeException('Please ensure redis_instance is of type redis.Redis') 73 | 74 | self.redis_instance = redis_instance 75 | 76 | def read(self, k: str) -> bytes: 77 | """Read a value from Redis given the associated string key.""" 78 | if not isinstance(k, str): 79 | raise InvalidKeyTypeException('Please ensure key is a string') 80 | 81 | return self.redis_instance.get(k) 82 | 83 | def write(self, k: str, v: bytes) -> None: 84 | """Write a value to Redis given a key-value pair.""" 85 | if not isinstance(k, str): 86 | raise InvalidKeyTypeException('Please ensure key is a string') 87 | 88 | if not isinstance(v, (str, bytes)): 89 | raise InvalidValueTypeException('Please ensure value is of type string or bytes') 90 | 91 | self.redis_instance.set(k, v) 92 | 93 | def delete(self, k: str) -> None: 94 | """Delete a value from Redis given the associated string key.""" 95 | if not isinstance(k, str): 96 | raise InvalidKeyTypeException('Please ensure key is a string') 97 | 98 | self.redis_instance.delete(k) 99 | 100 | 101 | class DiskCache(Cache): 102 | """ 103 | Implementation for a disk cache to be used by users who wish to inherit 104 | this functionality. We use the diskcache Python package from pypi to 105 | implement this caching feature. 106 | """ 107 | 108 | def __init__(self, disk_cache: diskcache.Cache): 109 | if not isinstance(disk_cache, diskcache.Cache): 110 | raise InvalidCacheTypeException('Please ensure disk_cache is of type diskcache.Cache') 111 | 112 | self.disk_cache = disk_cache 113 | 114 | def read(self, k: str) -> bytes: 115 | """Read a value from the disk cache given the associated string key.""" 116 | if not isinstance(k, str): 117 | raise InvalidKeyTypeException('Please ensure key is a string') 118 | 119 | return self.disk_cache[k] 120 | 121 | def write(self, k: str, v: bytes) -> None: 122 | """Write a value to the disk cache given a key-value pair.""" 123 | if not isinstance(k, str): 124 | raise InvalidKeyTypeException('Please ensure key is a string') 125 | 126 | if not isinstance(v, (str, bytes)): 127 | raise InvalidValueTypeException('Please ensure value is of type string or bytes') 128 | 129 | self.disk_cache[k] = v 130 | 131 | def delete(self, k: str) -> None: 132 | """Delete a value from disk cache given the associated string key.""" 133 | if not isinstance(k, str): 134 | raise InvalidKeyTypeException('Please ensure key is a string') 135 | 136 | self.disk_cache.delete(k) 137 | -------------------------------------------------------------------------------- /src/pydags/pipeline.py: -------------------------------------------------------------------------------- 1 | """ 2 | This module contains the implementation of the 'Pipeline': the entity 3 | synonymous with a DAG, which contains one or more 'Stages'. The pipeline 4 | models the associated DAG as a networkx.DiGraph object, and ensure the DAG 5 | property remains satisfied as nodes/stages are added. 6 | 7 | The execution of the pipeline can be run serial mode or in parallel mode. In 8 | serial mode, the stages are executed in the correct order (as defined by the 9 | interdependencies between stages), but are run in sequence. In this mode, 10 | stages that can be run in parallel are not. In parallel mode, stages that can 11 | be run in parallel are, and are split across the number of cores specific upon 12 | pipeline execution. 13 | 14 | A bundle of stages in a pipeline that can be run in parallel are known as a 15 | 'Group'. In serial mode, Groups contain only the single stage currently being 16 | executed. In parallel mode, Groups contains all the nodes that can be run in 17 | parallel at that time. In both cases, groups are run within a bespoke context 18 | manager (pydags.stage.StageExecutor). It is during startup and teardown of this 19 | context manager that pertinent information relating to DAG execution (such as 20 | stages in progress, completed stages, etc.) is written to Redis. This 21 | information can then be read from Redis and used downstream. 22 | """ 23 | 24 | from io import BytesIO 25 | from multiprocessing.pool import ThreadPool 26 | import typing 27 | import logging 28 | 29 | import networkx as nx 30 | import redis 31 | import matplotlib.pyplot as plt 32 | import matplotlib.image as mpimg 33 | 34 | from .serialization import CloudPickleSerializer 35 | from .cache import RedisCache 36 | from .stage import StageExecutor, Stage 37 | 38 | 39 | class StageException(Exception): 40 | pass 41 | 42 | 43 | class InvalidStageTypeException(StageException): 44 | pass 45 | 46 | 47 | class DAGException(Exception): 48 | pass 49 | 50 | 51 | class DAGVerificationException(DAGException): 52 | pass 53 | 54 | 55 | class Pipeline(CloudPickleSerializer, RedisCache): 56 | """ 57 | Implementation of the Pipeline functionality of pydags. The goal of a 58 | Pipeline is to orchestrate the execution of the DAG's constituent stages, 59 | and write associated Stage metadata during execution. This metadata, at 60 | present, consists of the nodes in progress, and the completed nodes. 61 | 62 | pipeline: The instance of networkx.DiGraph to house the underlying DAG of 63 | the pipeline. The reason this library is used is because various 64 | important things are already implemented, including topological 65 | sorting, graph visualization, numerous graph algorithms, etc. 66 | """ 67 | 68 | pipeline = nx.DiGraph() 69 | 70 | def __init__(self, redis_cache=redis.Redis(host='localhost', port=6379, db=1)): 71 | """ 72 | We use database number 1 from Redis for the Pipeline's underlying Redis 73 | cache. 74 | """ 75 | 76 | RedisCache.__init__(self, redis_cache) 77 | 78 | def topological_sort_grouped(self) -> typing.Generator: 79 | """ 80 | Method to perform a topological sort on the DAG/pipeline. However, 81 | this sort differs from nx.topological_sort in that it is grouped. This 82 | means that stages that can be run in parallel at each level of the DAG 83 | are grouped together. This is done in the case the user wants to 84 | distribute the execution of the pipeline across cores. In that case, 85 | the stages in each group will run in parallel. The default behaviour, 86 | however, is to run the entire pipeline serially, include stages within 87 | the same paralellizable group. 88 | 89 | Returns: 90 | Generator where each element is a list of nodes in the same group. 91 | """ 92 | 93 | logging.info('Computing a grouped topological sort of the pipeline DAG') 94 | indegree_map = {v: d for v, d in self.pipeline.in_degree() if d > 0} 95 | zero_indegree = [v for v, d in self.pipeline.in_degree() if d == 0] 96 | while zero_indegree: 97 | yield zero_indegree 98 | new_zero_indegree = [] 99 | for v in zero_indegree: 100 | for _, child in self.pipeline.edges(v): 101 | indegree_map[child] -= 1 102 | if not indegree_map[child]: 103 | new_zero_indegree.append(child) 104 | zero_indegree = new_zero_indegree 105 | 106 | def add_stage(self, stage: Stage) -> None: 107 | """ 108 | Method to add a stage to the pipeline. Stages are not added if they 109 | already exist in the DAG (although networkx accommodates for this). 110 | A stage is defined according to its name (usually the user-defined 111 | class or function name), and an attribute called 'stage_wrapper', 112 | which is the actual instance of a subclass of BaseStage. This object 113 | is used to run the associated DAG stage. 114 | 115 | Additionally, we add edges in the DAG between a stage and its preceding 116 | stages (as defined by the user). 117 | 118 | Lastly, a check is done to ensure that after adding the stage, the DAG 119 | is still indeed a DAG. 120 | """ 121 | 122 | if not isinstance(stage, Stage): 123 | raise InvalidStageTypeException('Please ensure your stage is a subclass of pydags.stage.Stage') 124 | 125 | self.pipeline.add_node(stage.name, stage_wrapper=stage) 126 | 127 | for preceding_stage in stage.preceding_stages: 128 | self.pipeline.add_edges_from([(preceding_stage.name, stage.name)]) 129 | 130 | if not nx.is_directed_acyclic_graph(self.pipeline): 131 | raise DAGVerificationException('Pipeline is no longer a DAG!') 132 | 133 | def add_stages(self, stages: typing.List[Stage]) -> None: 134 | """ 135 | Method to add a list of stages to the pipeline. Stages are not added if 136 | they already exist in the DAG (although networkx accommodates for 137 | this). A stage is defined according to its name (usually the 138 | user-defined class or function name), and an attribute called 139 | 'stage_wrapper', which is the actual instance of a subclass of 140 | BaseStage. This object is used to run the associated DAG stage. 141 | 142 | Additionally, we add edges in the DAG between a stage and its preceding 143 | stages (as defined by the user). 144 | 145 | Lastly, a check is done to ensure that after adding the stage, the DAG 146 | is still indeed a DAG. 147 | """ 148 | 149 | for stage in stages: 150 | self.add_stage(stage) 151 | 152 | def run_stage(self, stage_name: str) -> None: 153 | """ 154 | Method to run a particular stage of the pipeline/DAG. The 155 | associated instance of BaseStage is obtained used the stage name and 156 | executed using the required 'run' method. 157 | 158 | Args: 159 | stage_name : Name of the stage in the pipeline. 160 | """ 161 | self.pipeline.nodes[stage_name]['stage_wrapper'].run() 162 | 163 | def start(self, num_cores: int = None) -> None: 164 | """ 165 | Method to execute the pipeline (and all its constituent stages). The 166 | order of execution is defined by the grouped topological sort. The 167 | stages within a group will be executed in parallel (across cores) if 168 | num_cores is a positive integer. If num_cores remains None (as per 169 | default), then the entire pipeline (including stages within the same 170 | group) will run serially. 171 | 172 | Args: 173 | num_cores [, ]: Number of cores to distribute across. 174 | """ 175 | 176 | logging.info('Serializing pipeline and writing to Redis') 177 | self.write('pipeline', self.serialize(self.pipeline)) 178 | 179 | sorted_grouped_stages = self.topological_sort_grouped() 180 | for group in sorted_grouped_stages: 181 | logging.info('Processing group: %s', group) 182 | if num_cores: 183 | pool = ThreadPool(num_cores) 184 | with StageExecutor(self.redis_instance, group) as stage_executor: 185 | stage_executor.execute(pool.map, self.run_stage, group) 186 | else: 187 | for stage in group: 188 | with StageExecutor(self.redis_instance, [stage]) as stage_executor: 189 | stage_executor.execute(self.run_stage, stage) 190 | self.delete('done') 191 | 192 | def visualize(self): 193 | """ 194 | Method to visualize the pipeline/DAG by rendering a matplotlib figure. 195 | 196 | References: 197 | https://stackoverflow.com/questions/10379448/plotting-directed-graphs-in-python-in-a-way-that-show-all-edges-separately 198 | """ 199 | 200 | drawing = nx.drawing.nx_pydot.to_pydot(self.pipeline) 201 | 202 | png_str = drawing.create_png() 203 | sio = BytesIO() 204 | sio.write(png_str) 205 | sio.seek(0) 206 | 207 | img = mpimg.imread(sio) 208 | plt.imshow(img) 209 | 210 | plt.show() 211 | -------------------------------------------------------------------------------- /src/pydags/serialization.py: -------------------------------------------------------------------------------- 1 | """ 2 | This module contains the bases classes and implementations for the different 3 | serializers in use in pydags. Serialization is important to enable reading from 4 | and writing to the various key-value caches implemented in pydags (e.g. 5 | pydags.cache.RedisCache and pydags.cache.DiskCache). Users may implement their 6 | own serializers, however, any implementation must contain as least 'serialize' 7 | and 'deserialize' methods. 8 | 9 | Various pickle-based serialization algorithms are implemented: 1) vanilla 10 | pickle, 2) dill, and 3) cloudpickle. There are limitations of each, however, 11 | the default is pickle unless more comprehensive serialization is required 12 | (such as with the Pipeline implementation, where cloudpickle is used). 13 | 14 | Some of the noted limitations include: 15 | 1. pickle 16 | - Cannot serialize decorated functions 17 | 18 | 2. dill 19 | - Cannot serialize subclasses of abstract base classes 20 | """ 21 | 22 | import pickle 23 | from abc import ABC, abstractmethod 24 | 25 | import dill 26 | import cloudpickle 27 | 28 | 29 | class Serializer(ABC): 30 | @abstractmethod 31 | def serialize(self, *args, **kwargs): 32 | ... 33 | 34 | @abstractmethod 35 | def deserialize(self, *args, **kwargs): 36 | ... 37 | 38 | 39 | class SerializationException(Exception): 40 | pass 41 | 42 | 43 | class InvalidTypeForDeserializationException(SerializationException): 44 | pass 45 | 46 | 47 | class PickleSerializer(Serializer): # Cannot serialize decorated functions 48 | def serialize(self, obj: object) -> bytes: 49 | return pickle.dumps(obj) 50 | 51 | def deserialize(self, serialized_obj: bytes) -> object: 52 | if not isinstance(serialized_obj, bytes): 53 | raise InvalidTypeForDeserializationException('Please ensure the serialized object is of type bytes.') 54 | return pickle.loads(serialized_obj) 55 | 56 | 57 | class DillSerializer(Serializer): # Cannot serialize (subclasses of) abstract base classes 58 | def serialize(self, obj: object) -> bytes: 59 | return dill.dumps(obj) 60 | 61 | def deserialize(self, serialized_obj: bytes) -> object: 62 | if not isinstance(serialized_obj, bytes): 63 | raise InvalidTypeForDeserializationException('Please ensure the serialized object is of type bytes.') 64 | return dill.loads(serialized_obj) 65 | 66 | 67 | class CloudPickleSerializer(Serializer): 68 | def serialize(self, obj: object) -> bytes: 69 | return cloudpickle.dumps(obj) 70 | 71 | def deserialize(self, serialized_obj: bytes) -> object: 72 | if not isinstance(serialized_obj, bytes): 73 | raise InvalidTypeForDeserializationException('Please ensure the serialized object is of type bytes.') 74 | return cloudpickle.loads(serialized_obj) 75 | -------------------------------------------------------------------------------- /src/pydags/stage.py: -------------------------------------------------------------------------------- 1 | """ 2 | This module contains the base classes of the Stage functionality for pydags, 3 | as well as various implementations of Stages to be used in different contexts. 4 | Most implementations can be subclassed and extended by the user. All Stages 5 | must contain 1) a 'name' property and 2) 'run' method. 6 | 7 | There are two primary ways for users to define Stages in their own pipelines. 8 | The first is by decorating a function with the pydags.stage.stage decorator. 9 | The other is by subclassing Stage, RedisStage, DiskCacheStage. 10 | """ 11 | 12 | from abc import ABC, abstractmethod 13 | 14 | import diskcache 15 | import redis 16 | 17 | from .serialization import PickleSerializer 18 | from .cache import RedisCache, DiskCache 19 | 20 | 21 | class Stage(ABC): 22 | """ 23 | Base abstract class from which all stages must inherit. All subclasses must 24 | implement at least `name` and `run` methods. 25 | 26 | preceding_stages: List of preceding stages for the stage 27 | name: Name of the stage 28 | """ 29 | def __init__(self): 30 | self.preceding_stages = list() 31 | 32 | def after(self, pipeline_stage): 33 | """Method to add stages as dependencies for the current stage.""" 34 | self.preceding_stages.append(pipeline_stage) 35 | return self 36 | 37 | @property 38 | @abstractmethod 39 | def name(self) -> str: 40 | ... 41 | 42 | @abstractmethod 43 | def run(self, *args, **kwargs): 44 | ... 45 | 46 | 47 | class DecoratorStage(Stage): 48 | """ 49 | Class to wrap any user-defined function decorated with the stage decorator. 50 | 51 | stage_function: The callable defined by the user-defined function pipeline 52 | stage. 53 | args: The arguments to the user-defined function pipeline stage. 54 | kwargs: The keyword arguments to the user-defined function pipeline stage. 55 | """ 56 | 57 | def __init__(self, stage_function: callable, *args, **kwargs): 58 | super().__init__() 59 | 60 | self.stage_function = stage_function 61 | self.args = args 62 | self.kwargs = kwargs 63 | 64 | @property 65 | def name(self) -> str: 66 | """Name is given by the name of the user-defined decorated function.""" 67 | return self.stage_function.__name__ 68 | 69 | def run(self) -> None: 70 | """ 71 | Stage is run by calling the wrapped user function with its arguments. 72 | """ 73 | self.stage_function(*self.args, **self.kwargs) 74 | 75 | 76 | def stage(stage_function: callable): 77 | """ 78 | Decorator used to specify user-defined functions as pipeline stages (i.e. 79 | DAG nodes). The decorated wraps the decorated function in the 80 | DecoratorStage class, as this follows the expected format for a pipeline 81 | stage. 82 | """ 83 | def wrapper(*args, **kwargs) -> DecoratorStage: 84 | return DecoratorStage(stage_function, *args, **kwargs) 85 | 86 | return wrapper 87 | 88 | 89 | class StageExecutor(PickleSerializer, RedisCache): 90 | """ 91 | Context manager for the execution of a stage, or group of stages, of a 92 | pipeline. 93 | 94 | The setup phase (__enter__) persists relevant metadata such as the stages 95 | currently in progress to a Redis server. 96 | 97 | The teardown phase (__exit__) deletes relevant metadata from the Redis 98 | server. 99 | 100 | redis_instance: A connection to Redis. 101 | stages: The stages that are currently in progress. 102 | """ 103 | def __init__(self, redis_instance: redis.Redis, stages): 104 | RedisCache.__init__(self, redis_instance) 105 | 106 | self.pipeline = self.read('pipeline') 107 | self.stages = stages 108 | 109 | def __enter__(self): 110 | self.write('in_progress', self.serialize(self.stages)) 111 | return self 112 | 113 | def __exit__(self, exc_type, exc_val, exc_tb): 114 | completed = self.deserialize(self.read('in_progress')) 115 | current_done = self.read('done') 116 | if current_done is None: 117 | current_done = [] 118 | else: 119 | current_done = self.deserialize(current_done) 120 | current_done += completed 121 | self.write('done', self.serialize(current_done)) 122 | self.delete('in_progress') 123 | 124 | @staticmethod 125 | def execute(fn: callable, *args, **kwargs) -> None: 126 | """Execute the stage/group of stages.""" 127 | fn(*args, **kwargs) 128 | 129 | 130 | class RedisStage(Stage, PickleSerializer, RedisCache): 131 | """ 132 | Stage type to use if an in-memory cache (i.e. Redis) is required. Redis can 133 | be used to pass data between stages, or cache values to be used elsewhere 134 | downstream. It's completely up to the implementer/user, as this interface 135 | to Redis is generic, and enables the reading/writing of generic Python 136 | objects from/to Redis through pickle-based serialization. 137 | 138 | The underlying DAG of the Pipeline object requires serialisation itself as 139 | part of the inner workings of pydags. As such, the __getstate__ and 140 | __setstate__ dunder methods are overridden in order to temporarily stop 141 | Redis upon pickling, and to restart it when unpickling. 142 | """ 143 | 144 | def __init__(self, redis_instance): 145 | RedisCache.__init__(self, redis_instance=redis_instance) 146 | Stage.__init__(self) 147 | 148 | def __getstate__(self): 149 | self.redis_metadata = self.redis_instance.connection_pool.connection_kwargs 150 | self.redis_instance.close() 151 | return self 152 | 153 | def __setstate__(self, state): 154 | self.redis_instance = redis.Redis(**self.redis_metadata) 155 | 156 | def read(self, k: str) -> object: 157 | return self.deserialize(RedisCache.read(self, k)) 158 | 159 | def write(self, k: str, v: object) -> None: 160 | RedisCache.write(self, k, self.serialize(v)) 161 | 162 | @property 163 | def name(self) -> str: 164 | """ 165 | The name is the final subclass (i.e. the name class defined by the user 166 | when subclassing this class. 167 | """ 168 | return self.__class__.__name__ 169 | 170 | 171 | class DiskCacheStage(Stage, PickleSerializer, DiskCache): 172 | """ 173 | Stage type to use if a disk-based cache is required. The disk cache can be 174 | used to pass data between stages, or cache values to be used elsewhere 175 | downstream. It's completely up to the implementer/user, as this interface 176 | to the disk cache is generic, and enables the reading/writing of generic 177 | Python objects from/to the disk cache through pickle-based serialization. 178 | """ 179 | 180 | def __init__(self, cache: diskcache.Cache): 181 | DiskCache.__init__(self, disk_cache=cache) 182 | Stage.__init__(self) 183 | 184 | def read(self, k: str) -> object: 185 | return self.deserialize(DiskCache.read(self, k)) 186 | 187 | def write(self, k: str, v: object) -> None: 188 | DiskCache.write(self, k, self.serialize(v)) 189 | 190 | @property 191 | def name(self) -> str: 192 | """ 193 | The name is the final subclass (i.e. the name class defined by the user 194 | when subclassing this class. 195 | """ 196 | return self.__class__.__name__ 197 | --------------------------------------------------------------------------------