The best way to control Claude Code, Codex and more from any computer or phone
| 1 | import * as dataplex from "@distilled.cloud/gcp/dataplex_v1"; |
| 2 | import { Unowned } from "alchemy/AdoptPolicy"; |
| 3 | import { isResolved } from "alchemy/Diff"; |
| 4 | import * as Provider from "alchemy/Provider"; |
| 5 | import { Resource } from "alchemy/Resource"; |
| 6 | import { tagRecord } from "alchemy/Tags"; |
| 7 | import * as Data from "effect/Data"; |
| 8 | import * as Effect from "effect/Effect"; |
| 9 | import * as Schedule from "effect/Schedule"; |
| 10 | import { GcpEnvironment } from "../Environment.ts"; |
| 11 | import { createInternalLabels, diffLabels, hasAlchemyLabels, toLabels } from "../Labels.ts"; |
| 12 | import type { Providers } from "../Providers.ts"; |
| 13 | import { waitForOperation } from "./operations.ts"; |
| 14 | import { |
| 15 | DataplexNotResolved, |
| 16 | expandParent, |
| 17 | fingerprint, |
| 18 | hasAlchemyLabelMap, |
| 19 | isPendingState, |
| 20 | listChildResources, |
| 21 | listLakes, |
| 22 | listTasks, |
| 23 | normalizeLocation, |
| 24 | parseResourceName, |
| 25 | replaceIfChanged, |
| 26 | toPhysicalRfc1035, |
| 27 | userLabels, |
| 28 | } from "./shared.ts"; |
| 29 | |
| 30 | const DEFAULT_TRIGGER = "ON_DEMAND"; |
| 31 | |
| 32 | export type TaskTriggerSpec = { |
| 33 | /** |
| 34 | * Trigger type. Immutable — changing it replaces the task. |
| 35 | * @default "ON_DEMAND" |
| 36 | */ |
| 37 | type?: dataplex.GoogleCloudDataplexV1TaskTriggerSpecTypeEnum | (string & {}); |
| 38 | /** First run after this RFC3339 timestamp. */ |
| 39 | startTime?: string; |
| 40 | /** Temporarily disable RECURRING execution. */ |
| 41 | disabled?: boolean; |
| 42 | /** Retry attempts before aborting. */ |
| 43 | maxRetries?: number; |
| 44 | /** Cron schedule. Required for `RECURRING`. */ |
| 45 | schedule?: string; |
| 46 | }; |
| 47 | |
| 48 | export type TaskExecutionSpec = { |
| 49 | /** Arguments interpolated before passing to the driver. */ |
| 50 | args?: Record<string, string>; |
| 51 | /** |
| 52 | * Service account used to execute the task. If omitted, the project |
| 53 | * Compute Engine default service account is used. |
| 54 | */ |
| 55 | serviceAccount?: string; |
| 56 | /** Project in which jobs run. Defaults to the lake project. */ |
| 57 | project?: string; |
| 58 | /** Maximum job duration (e.g. `"3600s"`). */ |
| 59 | maxJobExecutionLifetime?: string; |
| 60 | /** Cloud KMS key used to encrypt job data. */ |
| 61 | kmsKey?: string; |
| 62 | }; |
| 63 | |
| 64 | export type TaskInfrastructureSpec = { |
| 65 | /** Dataproc Serverless executor counts. */ |
| 66 | batch?: { |
| 67 | executorsCount?: number; |
| 68 | maxExecutorsCount?: number; |
| 69 | }; |
| 70 | /** Container image runtime. */ |
| 71 | containerImage?: { |
| 72 | image?: string; |
| 73 | javaJars?: string[]; |
| 74 | pythonPackages?: string[]; |
| 75 | properties?: Record<string, string>; |
| 76 | }; |
| 77 | /** VPC network used to run the job. */ |
| 78 | vpcNetwork?: { |
| 79 | network?: string; |
| 80 | subNetwork?: string; |
| 81 | networkTags?: string[]; |
| 82 | }; |
| 83 | }; |
| 84 | |
| 85 | export type TaskSparkConfig = { |
| 86 | /** Cloud Storage URI of the jar containing the main class. */ |
| 87 | mainJarFileUri?: string; |
| 88 | /** Driver main class. */ |
| 89 | mainClass?: string; |
| 90 | /** Cloud Storage URI of the main Python file. */ |
| 91 | pythonScriptFile?: string; |
| 92 | /** Cloud Storage URI of a SQL query file. */ |
| 93 | sqlScriptFile?: string; |
| 94 | /** Inline SQL query text. */ |
| 95 | sqlScript?: string; |
| 96 | /** Working-directory files. */ |
| 97 | fileUris?: string[]; |
| 98 | /** Archives extracted into the working directory. */ |
| 99 | archiveUris?: string[]; |
| 100 | /** Infrastructure used to run the Spark job. */ |
| 101 | infrastructureSpec?: TaskInfrastructureSpec; |
| 102 | }; |
| 103 | |
| 104 | export type TaskNotebookConfig = { |
| 105 | /** Path to the input notebook. */ |
| 106 | notebook: string; |
| 107 | /** Infrastructure used to run the notebook. */ |
| 108 | infrastructureSpec?: TaskInfrastructureSpec; |
| 109 | /** Working-directory files. */ |
| 110 | fileUris?: string[]; |
| 111 | /** Archives extracted into the working directory. */ |
| 112 | archiveUris?: string[]; |
| 113 | }; |
| 114 | |
| 115 | export type LakesTaskProps = { |
| 116 | /** |
| 117 | * Parent lake. Full name |
| 118 | * `projects/{project}/locations/{location}/lakes/{lake}` or the lake id |
| 119 | * (combined with `location`). Immutable — changing it replaces the task. |
| 120 | */ |
| 121 | lake: string; |
| 122 | /** |
| 123 | * Region used when `lake` is a bare id. |
| 124 | * @default the stack's GCP region (`GCP.Region`, else the profile region, else `us-central1`) |
| 125 | */ |
| 126 | location?: string; |
| 127 | /** |
| 128 | * Task id. Immutable — changing it replaces the task. |
| 129 | */ |
| 130 | taskId?: string; |
| 131 | /** |
| 132 | * Human-readable description. |
| 133 | */ |
| 134 | description?: string; |
| 135 | /** |
| 136 | * User-friendly display name. |
| 137 | */ |
| 138 | displayName?: string; |
| 139 | /** |
| 140 | * User labels. Alchemy ownership labels are merged in automatically. |
| 141 | */ |
| 142 | labels?: Record<string, string>; |
| 143 | /** |
| 144 | * How often and when the task runs. |
| 145 | * @default { type: "ON_DEMAND" } |
| 146 | */ |
| 147 | triggerSpec?: TaskTriggerSpec; |
| 148 | /** |
| 149 | * How the task is executed (service account, args, encryption). |
| 150 | */ |
| 151 | executionSpec?: TaskExecutionSpec; |
| 152 | /** |
| 153 | * Spark workload. Mutually exclusive with `notebook`. |
| 154 | */ |
| 155 | spark?: TaskSparkConfig; |
| 156 | /** |
| 157 | * Notebook workload. Mutually exclusive with `spark`. |
| 158 | */ |
| 159 | notebook?: TaskNotebookConfig; |
| 160 | }; |
| 161 | |
| 162 | export type LakesTask = Resource< |
| 163 | "GCP.Dataplex.LakesTask", |
| 164 | LakesTaskProps, |
| 165 | { |
| 166 | /** Full resource name `.../lakes/{lake}/tasks/{task}`. */ |
| 167 | name: string; |
| 168 | /** Task id. */ |
| 169 | taskId: string; |
| 170 | /** Parent lake resource name. */ |
| 171 | lake: string; |
| 172 | /** Project id. */ |
| 173 | project: string; |
| 174 | /** Location id. */ |
| 175 | location: string; |
| 176 | /** Description. */ |
| 177 | description: string | undefined; |
| 178 | /** User-friendly display name. */ |
| 179 | displayName: string | undefined; |
| 180 | /** User labels (Alchemy ownership labels stripped). */ |
| 181 | labels: Record<string, string>; |
| 182 | /** Trigger type. */ |
| 183 | triggerType: string | undefined; |
| 184 | /** Execution service account. */ |
| 185 | serviceAccount: string | undefined; |
| 186 | /** Lifecycle state. */ |
| 187 | state: string | undefined; |
| 188 | /** Server-assigned uid. */ |
| 189 | uid: string | undefined; |
| 190 | /** RFC3339 creation timestamp. */ |
| 191 | createTime: string | undefined; |
| 192 | /** RFC3339 last-update timestamp. */ |
| 193 | updateTime: string | undefined; |
| 194 | }, |
| 195 | never, |
| 196 | Providers |
| 197 | >; |
| 198 | |
| 199 | /** |
| 200 | * A Dataplex lake task — a Spark or notebook job scheduled against a lake. |
| 201 | * |
| 202 | * Changing `lake`, `taskId`, `location`, or trigger type replaces the |
| 203 | * task. Description, labels, execution spec, and Spark/notebook config |
| 204 | * update in place. |
| 205 | * |
| 206 | * ### Creating a Task |
| 207 | * **Example:** On-demand Spark SQL |
| 208 | * ```typescript |
| 209 | * const task = yield* GCP.Dataplex.LakesTask("SelectOne", { |
| 210 | * lake: lake.name, |
| 211 | * executionSpec: { serviceAccount }, |
| 212 | * spark: { sqlScript: "SELECT 1" }, |
| 213 | * }); |
| 214 | * ``` |
| 215 | * |
| 216 | * **Example:** Recurring notebook |
| 217 | * ```typescript |
| 218 | * const task = yield* GCP.Dataplex.LakesTask("Nightly", { |
| 219 | * lake: lake.name, |
| 220 | * triggerSpec: { type: "RECURRING", schedule: "0 2 * * *" }, |
| 221 | * executionSpec: { serviceAccount }, |
| 222 | * notebook: { notebook: "gs://bucket/job.ipynb" }, |
| 223 | * }); |
| 224 | * ``` |
| 225 | * |
| 226 | * @resource |
| 227 | * @category Dataplex |
| 228 | */ |
| 229 | export const LakesTask = Resource<LakesTask>("GCP.Dataplex.LakesTask"); |
| 230 | |
| 231 | export class LakesTaskNotResolved extends Data.TaggedError("GCP.Dataplex.LakesTaskNotResolved")<{ |
| 232 | resourceName: string; |
| 233 | }> {} |
| 234 | |
| 235 | export class LakesTaskStillExists extends Data.TaggedError("GCP.Dataplex.LakesTaskStillExists")<{ |
| 236 | resourceName: string; |
| 237 | }> {} |
| 238 | |
| 239 | const lakeOf = (lake: string, project: string, location: string) => |
| 240 | expandParent(lake, project, location, "lakes"); |
| 241 | |
| 242 | const resourceName = (lake: string, taskId: string) => `${lake}/tasks/${taskId}`; |
| 243 | |
| 244 | const triggerBody = ( |
| 245 | spec: TaskTriggerSpec | undefined, |
| 246 | ): dataplex.GoogleCloudDataplexV1TaskTriggerSpec => ({ |
| 247 | type: (spec?.type ?? DEFAULT_TRIGGER).toUpperCase(), |
| 248 | startTime: spec?.startTime, |
| 249 | disabled: spec?.disabled, |
| 250 | maxRetries: spec?.maxRetries, |
| 251 | schedule: spec?.schedule, |
| 252 | }); |
| 253 | |
| 254 | const executionBody = ( |
| 255 | spec: TaskExecutionSpec | undefined, |
| 256 | ): dataplex.GoogleCloudDataplexV1TaskExecutionSpec => ({ |
| 257 | args: spec?.args, |
| 258 | serviceAccount: spec?.serviceAccount, |
| 259 | project: spec?.project, |
| 260 | maxJobExecutionLifetime: spec?.maxJobExecutionLifetime, |
| 261 | kmsKey: spec?.kmsKey, |
| 262 | }); |
| 263 | |
| 264 | const toAttrs = (task: dataplex.GoogleCloudDataplexV1Task, project: string) => { |
| 265 | const name = task.name ?? ""; |
| 266 | const parsed = parseResourceName(name, "tasks"); |
| 267 | return { |
| 268 | name, |
| 269 | taskId: parsed.id, |
| 270 | lake: parsed.parent, |
| 271 | project: parsed.project || project, |
| 272 | location: parsed.location, |
| 273 | description: task.description, |
| 274 | displayName: task.displayName, |
| 275 | labels: userLabels(task.labels), |
| 276 | triggerType: task.triggerSpec?.type, |
| 277 | serviceAccount: task.executionSpec?.serviceAccount, |
| 278 | state: task.state, |
| 279 | uid: task.uid, |
| 280 | createTime: task.createTime, |
| 281 | updateTime: task.updateTime, |
| 282 | }; |
| 283 | }; |
| 284 | |
| 285 | const getByName = (name: string) => |
| 286 | name.length === 0 || name.includes("//") |
| 287 | ? Effect.undefined |
| 288 | : dataplex |
| 289 | .getProjectsLocationsLakesTasks({ name }) |
| 290 | .pipe(Effect.catchTag("NotFound", () => Effect.undefined)); |
| 291 | |
| 292 | const waitUntilReady = (name: string) => |
| 293 | getByName(name).pipe( |
| 294 | Effect.filterOrFail( |
| 295 | (task): task is dataplex.GoogleCloudDataplexV1Task => task !== undefined, |
| 296 | () => new LakesTaskNotResolved({ resourceName: name }), |
| 297 | ), |
| 298 | Effect.filterOrFail( |
| 299 | (task) => !isPendingState(task.state), |
| 300 | () => new DataplexNotResolved({ resourceName: name }), |
| 301 | ), |
| 302 | Effect.retry({ |
| 303 | while: (error) => |
| 304 | error._tag === "GCP.Dataplex.LakesTaskNotResolved" || |
| 305 | error._tag === "GCP.Dataplex.NotResolved" || |
| 306 | error._tag === "TooManyRequests", |
| 307 | times: 10, |
| 308 | schedule: Schedule.spaced("10 seconds"), |
| 309 | }), |
| 310 | ); |
| 311 | |
| 312 | const waitUntilGone = (name: string) => |
| 313 | getByName(name).pipe( |
| 314 | Effect.flatMap((task) => |
| 315 | task === undefined |
| 316 | ? Effect.void |
| 317 | : Effect.fail(new LakesTaskStillExists({ resourceName: name })), |
| 318 | ), |
| 319 | Effect.retry({ |
| 320 | while: (error) => error._tag === "GCP.Dataplex.LakesTaskStillExists", |
| 321 | times: 10, |
| 322 | schedule: Schedule.spaced("3 seconds"), |
| 323 | }), |
| 324 | ); |
| 325 | |
| 326 | export const LakesTaskProvider = () => |
| 327 | Provider.succeed(LakesTask, { |
| 328 | stables: ["name", "taskId", "lake", "project", "location", "uid", "createTime"], |
| 329 | |
| 330 | diff: Effect.fn(function* ({ news, olds, output }) { |
| 331 | if (!isResolved(news)) return undefined; |
| 332 | const env = yield* GcpEnvironment.current; |
| 333 | const previousId = olds?.taskId ?? output?.taskId; |
| 334 | const nextId = news.taskId ?? previousId; |
| 335 | const previousLake = olds?.lake ?? output?.lake; |
| 336 | const nextLake = news.lake ?? previousLake; |
| 337 | const previousLocation = normalizeLocation(olds?.location ?? output?.location, env.region); |
| 338 | const nextLocation = normalizeLocation( |
| 339 | news.location ?? olds?.location ?? output?.location, |
| 340 | env.region, |
| 341 | ); |
| 342 | const previousTrigger = ( |
| 343 | olds?.triggerSpec?.type ?? |
| 344 | output?.triggerType ?? |
| 345 | DEFAULT_TRIGGER |
| 346 | ).toUpperCase(); |
| 347 | const nextTrigger = (news.triggerSpec?.type ?? previousTrigger).toUpperCase(); |
| 348 | if ( |
| 349 | replaceIfChanged(previousId, nextId) || |
| 350 | replaceIfChanged(previousLake, nextLake) || |
| 351 | (output !== undefined && previousLocation !== nextLocation) || |
| 352 | previousTrigger !== nextTrigger |
| 353 | ) { |
| 354 | return { |
| 355 | action: "replace" as const, |
| 356 | deleteFirst: |
| 357 | previousLake === nextLake && |
| 358 | previousLocation === nextLocation && |
| 359 | previousId !== undefined && |
| 360 | nextId === previousId, |
| 361 | }; |
| 362 | } |
| 363 | return undefined; |
| 364 | }), |
| 365 | |
| 366 | read: Effect.fn(function* ({ id, olds, output }) { |
| 367 | const env = yield* GcpEnvironment.current; |
| 368 | const location = normalizeLocation(olds?.location ?? output?.location, env.region); |
| 369 | const lake = lakeOf(olds?.lake ?? output?.lake ?? "", env.project, location); |
| 370 | const taskId = yield* toPhysicalRfc1035(id, olds?.taskId, output?.taskId); |
| 371 | const name = output?.name ?? resourceName(lake, taskId); |
| 372 | const existing = yield* getByName(name); |
| 373 | if (existing === undefined) return undefined; |
| 374 | const attrs = toAttrs(existing, env.project); |
| 375 | return (yield* hasAlchemyLabels(id, tagRecord(existing.labels))) ? attrs : Unowned(attrs); |
| 376 | }), |
| 377 | |
| 378 | list: () => |
| 379 | Effect.gen(function* () { |
| 380 | const env = yield* GcpEnvironment.current; |
| 381 | const lakes = yield* listLakes(env.project, env.region); |
| 382 | const tasks = yield* listChildResources(lakes, listTasks); |
| 383 | return tasks |
| 384 | .filter((task) => hasAlchemyLabelMap(task.labels)) |
| 385 | .map((task) => toAttrs(task, env.project)); |
| 386 | }), |
| 387 | |
| 388 | reconcile: Effect.fn(function* ({ id, news, output }) { |
| 389 | const env = yield* GcpEnvironment.current; |
| 390 | const location = normalizeLocation(news.location ?? output?.location, env.region); |
| 391 | const lake = lakeOf(news.lake, env.project, location); |
| 392 | const taskId = yield* toPhysicalRfc1035(id, news.taskId, output?.taskId); |
| 393 | const name = output?.name ?? resourceName(lake, taskId); |
| 394 | const desiredLabels = { |
| 395 | ...toLabels(news.labels), |
| 396 | ...(yield* createInternalLabels(id)), |
| 397 | }; |
| 398 | const triggerSpec = triggerBody(news.triggerSpec); |
| 399 | const executionSpec = executionBody(news.executionSpec); |
| 400 | |
| 401 | let current = yield* getByName(name); |
| 402 | |
| 403 | if (current === undefined) { |
| 404 | const created = yield* dataplex |
| 405 | .createProjectsLocationsLakesTasks({ |
| 406 | parent: lake, |
| 407 | taskId, |
| 408 | body: { |
| 409 | description: news.description, |
| 410 | displayName: news.displayName, |
| 411 | labels: desiredLabels, |
| 412 | triggerSpec, |
| 413 | executionSpec, |
| 414 | spark: news.spark, |
| 415 | notebook: news.notebook, |
| 416 | }, |
| 417 | }) |
| 418 | .pipe(Effect.catchTag("Conflict", () => Effect.undefined)); |
| 419 | if (created !== undefined) { |
| 420 | yield* waitForOperation(created); |
| 421 | } |
| 422 | current = yield* waitUntilReady(name); |
| 423 | } |
| 424 | |
| 425 | if (current === undefined) { |
| 426 | return yield* new LakesTaskNotResolved({ resourceName: name }); |
| 427 | } |
| 428 | |
| 429 | const observedLabels = tagRecord(current.labels); |
| 430 | const { upsert, removed } = diffLabels(observedLabels, desiredLabels); |
| 431 | const labelsChanged = upsert.length > 0 || removed.length > 0; |
| 432 | const descriptionChanged = (current.description ?? "") !== (news.description ?? ""); |
| 433 | const displayNameChanged = (current.displayName ?? "") !== (news.displayName ?? ""); |
| 434 | const triggerChanged = fingerprint(current.triggerSpec) !== fingerprint(triggerSpec); |
| 435 | const executionChanged = fingerprint(current.executionSpec) !== fingerprint(executionSpec); |
| 436 | const sparkChanged = fingerprint(current.spark) !== fingerprint(news.spark); |
| 437 | const notebookChanged = fingerprint(current.notebook) !== fingerprint(news.notebook); |
| 438 | |
| 439 | if ( |
| 440 | labelsChanged || |
| 441 | descriptionChanged || |
| 442 | displayNameChanged || |
| 443 | triggerChanged || |
| 444 | executionChanged || |
| 445 | sparkChanged || |
| 446 | notebookChanged |
| 447 | ) { |
| 448 | const updateMask = [ |
| 449 | labelsChanged ? "labels" : undefined, |
| 450 | descriptionChanged ? "description" : undefined, |
| 451 | displayNameChanged ? "display_name" : undefined, |
| 452 | triggerChanged ? "trigger_spec" : undefined, |
| 453 | executionChanged ? "execution_spec" : undefined, |
| 454 | sparkChanged ? "spark" : undefined, |
| 455 | notebookChanged ? "notebook" : undefined, |
| 456 | ].filter((field): field is string => field !== undefined); |
| 457 | const operation = yield* dataplex.patchProjectsLocationsLakesTasks({ |
| 458 | name: current.name ?? name, |
| 459 | updateMask: updateMask.join(","), |
| 460 | body: { |
| 461 | description: news.description, |
| 462 | displayName: news.displayName, |
| 463 | labels: desiredLabels, |
| 464 | triggerSpec, |
| 465 | executionSpec, |
| 466 | spark: news.spark, |
| 467 | notebook: news.notebook, |
| 468 | }, |
| 469 | }); |
| 470 | yield* waitForOperation(operation); |
| 471 | current = yield* waitUntilReady(current.name ?? name); |
| 472 | } |
| 473 | |
| 474 | return toAttrs(current, env.project); |
| 475 | }), |
| 476 | |
| 477 | delete: Effect.fn(function* ({ output }) { |
| 478 | if (!output.name || output.name.includes("//")) return; |
| 479 | const operation = yield* dataplex |
| 480 | .deleteProjectsLocationsLakesTasks({ name: output.name }) |
| 481 | .pipe( |
| 482 | Effect.retry({ |
| 483 | while: (error) => error._tag === "Conflict", |
| 484 | times: 8, |
| 485 | schedule: Schedule.spaced("2 seconds"), |
| 486 | }), |
| 487 | Effect.catchTags({ |
| 488 | NotFound: () => Effect.undefined, |
| 489 | BadRequest: () => Effect.undefined, |
| 490 | }), |
| 491 | ); |
| 492 | if (operation !== undefined) { |
| 493 | yield* waitForOperation(operation, { notFoundOk: true }); |
| 494 | } |
| 495 | yield* waitUntilGone(output.name); |
| 496 | }), |
| 497 | }); |