将 TaskEither<E, A>[] 转换为 TaskEither<E[],A[]>

Question

我仍在学习和使用 fp-ts，但无法弄清楚。我有一些 API 个调用，我想将所有成功的响应和所有错误收集到数组中。

所以，我尝试使用 array.sequence:

TE.map(schedules =>
  array.sequence(TE.taskEither)(
    schedules.map(({ Program, ...schedule }) =>
      pipe(
        createProgramIfNotExist(Program),
        TE.map(createdProgram =>
          setRecordingSchedules(programsClient, { ...schedule, ProgramId: createdProgram.Id }),
        ),
        TE.flatten,
      ),
    ),
  ),
),
TE.flatten

这对响应工作正常，但我只收到来自 API 调用的最后一个错误。有什么办法可以把所有的错误收集到一个数组中吗？

下面我写了进行 API 调用的函数，以防万一我在那里遇到问题。

export const setRecordingSchedules = (
  fetcher: AxiosInstance,
  config: RecordingConfig,
): TE.TaskEither<Error, [ReturnType<typeof Schedule.codec.encode>, number]> => {
  const url = `/programs/${config.ProgramId}/recordingschedules`;
  return pipe(Schedule.codec.encode({ ...config, AutoPodcastConfig }), body =>
    pipe(
      TE.tryCatch(
        () => handleRateLimit(() => fetcher.put(url, body)),
        err => raiseUpdateError(unknownToError(err)),
      ),
      TE.map(({ status }) => [body, status]),
    ),
  );
};

export const createRecordingSchedule = (
  fetcher: AxiosInstance,
  program: Program.Type,
): TE.TaskEither<Error, Program.Type> =>
  pipe(
    Program.codec.encode(program),
    body =>
      pipe(
        TE.tryCatch(
          () => handleRateLimit(() => fetcher.post('/programs', body)),
          err => raiseCreateError(unknownToError(err)),
        ),
        TE.map(({ data }) =>
          pipe(
            Program.codec.decode({ ...data, Network: program.Network }),
            E.bimap(
              errors => ReportValidationError(errors, { ...data, Network: program.Network }),
              decoded => decoded,
            ),
            TE.fromEither,
          ),
        ),
      ),
    TE.flatten,
  );

Answer 1

思路是转换TaskEither<E, A>[] -> Task<Either<E, A>[]>，然后给一个合并函数Either<E, A>[] -> Either<E[], A[]>就可以实现你想要的了。合并功能也是分几步完成的：

为 Either<E[], A[]>
Either<E, A> -> Either<E[], A[]> - 将值或错误简单地包装到单例数组中
foldMap Either<E, A>[] 通过映射上述包装器并使用上述 monoid

import * as E from 'fp-ts/lib/Either';
import * as A from 'fp-ts/lib/Array';

// type Error = ...
// type Result = ...

// Make a validation monoid for `Either<E[], A[]>`
const validationMonoid = E.getValidationMonoid(A.getMonoid<Error>(), A.getMonoid<Result>());

// `Either<E, A> -> Either<E[], A[]>` - trivial wrap of value or error into singleton arrays
const validationWrap = E.bimap(x => [x], x => [x]);

// `foldMap` `Either<E, A>[]` by mapping with the above wrapper and folding using above monoid
const validationMerge = E.foldMap(validationMonoid)(validationWrap);

现在你可以做

const tasks: TaskEither<Error, Result>[] = ...;

const aggregatedResults: TaskEither<Error[], Result[]> = pipe(
  // convert `TaskEither<E, A>[] -> Task<Either<E, A>[]>`
  array.sequence(T.task)(tasks),
  // and then apply the merge
  T.map(validationMerge)
);

Answer 2

解决方案是使用 traverse。想法是获取验证任务幺半群并将每个 E 合并到一个数组中。

array.traverse(TE.getTaskValidation(A.getMonoid<E>()))(xs, TE.mapLeft(A.of))

完整的工作解决方案：

import { array, getMonoid, of as arrayOf } from 'fp-ts/lib/Array';

TE.chain(schedules =>
  array.sequence(TE.taskEither)(
    schedules.map(({ Program, ...schedule }) =>
      pipe(
        createProgramIfNotExist(Program),
        TE.map(createdProgram =>
          setRecordingSchedules(programsClient, { ...schedule, ProgramId: createdProgram.Id }),
        ),
      ),
    ),
    xs => array.traverse(TE.getTaskValidation(getMonoid<E>()))(xs, TE.mapLeft(arrayOf)),
  ),
),

将 TaskEither<E, A>[] 转换为 TaskEither<E[],A[]>

Convert TaskEither<E, A>[] to TaskEither<E[],A[]>

typescript

functional-programming

fp-ts