francis Posts - Answer Overflow

francis

Explore posts from servers

•Created by francis on 9/23/2024 in #questions

Is there currently a known issue with arktype + trpc compatibility?

This is with latest RC.12

TypeError: Cannot read properties of undefined (reading 'traverse')
    at assert (/node_modules/@ark/schema/out/roots/root.js:80:29)
    at inputMiddleware (/node_modules/@trpc/server/dist/index.mjs:278:33)
    at callRecursive (/node_modules/@trpc/server/dist/index.mjs:452:38)
    at Object.next (/node_modules/@trpc/server/dist/index.mjs:461:32)
    at eval (/src/lib/trpc/middleware.ts:21:15)
    at callRecursive (/node_modules/@trpc/server/dist/index.mjs:452:38)
    at resolve (/node_modules/@trpc/server/dist/index.mjs:482:30)
    at callProcedure (/node_modules/@trpc/server/dist/config-d5fdbd39.mjs:155:12)
    at inputToProcedureCall (/node_modules/@trpc/server/dist/resolveHTTPResponse-2fc435bb.mjs:46:28)`

TypeError: Cannot read properties of undefined (reading 'traverse')
    at assert (/node_modules/@ark/schema/out/roots/root.js:80:29)
    at inputMiddleware (/node_modules/@trpc/server/dist/index.mjs:278:33)
    at callRecursive (/node_modules/@trpc/server/dist/index.mjs:452:38)
    at Object.next (/node_modules/@trpc/server/dist/index.mjs:461:32)
    at eval (/src/lib/trpc/middleware.ts:21:15)
    at callRecursive (/node_modules/@trpc/server/dist/index.mjs:452:38)
    at resolve (/node_modules/@trpc/server/dist/index.mjs:482:30)
    at callProcedure (/node_modules/@trpc/server/dist/config-d5fdbd39.mjs:155:12)
    at inputToProcedureCall (/node_modules/@trpc/server/dist/resolveHTTPResponse-2fc435bb.mjs:46:28)`

This occurs when I do .input(type(anything).assert). If I do something such as .input((i) => type(anything).assert(i)), it works fine. How on earth is it ending up with this being undefined when it is run this way?

4 replies

Aarktype

•Created by francis on 9/23/2024 in #questions

How would you define an object with optional keys, but require at least one of the keys to be filled

An example of a very basic schema:

const schema = type({ a: type.string.optional(), b: type.string.optional() });

const schema = type({ a: type.string.optional(), b: type.string.optional() });

I'd like this to accept values {a: 'foo'}, { b: 'foo' }, and { a: 'foo', b: 'bar' }, but fail on {}. It would be a lovely bonus for the type to be inferred as { a: string, b?: string } | { a?: string, b: string } as well. Is there a way to do this without explicitly specifying that inferred type as an arktype union using .or?

15 replies

Aarktype

•Created by francis on 9/21/2024 in #questions

Is there a recommended way to transform a type union into a discriminated union by adding a tag?

I'm not sure I'm describing this well, but I have a situation with two types that I am .or-ing together. It all works, but I'm interacting with a library that needs a discriminated union to function properly for type narrowing, and I am not sure how to add a discriminated union tag based on which type in the Or resolved (if that makes sense?) Here's a simple example to demonstrate:

const first = type({ key: "string" });
const second = type({ key: "number" });
const combined = first.or(second);

const first = type({ key: "string" });
const second = type({ key: "number" });
const combined = first.or(second);

I'd like to have this be resolve to e.g. { key: string, tag: 'first' } | { key: number, tag: 'second' } based on which branch in the union was followed. I've tried it with this:

const first = type({ key: "string" }).pipe((o) => ({ ...o, _tag: "first" }) as const);
const second = type({ key: "number" }).pipe((o) => ({ ...o, _tag: "second" }) as const);
const combined = first.or(second);

const first = type({ key: "string" }).pipe((o) => ({ ...o, _tag: "first" }) as const);
const second = type({ key: "number" }).pipe((o) => ({ ...o, _tag: "second" }) as const);
const combined = first.or(second);

and it works - but seems inelegant. Is there a more idiomatic solution?

13 replies

Aarktype

•Created by francis on 9/21/2024 in #questions

Is there a way to do a case insensitive literal match?

I can pipe a string.lower to a lower-case literal but that seems complicated for this.

27 replies

Aarktype

•Created by francis on 9/21/2024 in #questions

Is there a way to perform schema parsing generically?

My reasoning is to integrate Arktype with https://effect.website/ This most simple example fails:

function genericDemo<T>(schema: Type<T>, u: unknown): T {
  return schema.assert(u);
}

function genericDemo<T>(schema: Type<T>, u: unknown): T {
  return schema.assert(u);
}

This fails as the return type of .assert is finalizeDistillation<T, _distill<T, { endpoint: "out"; }>> Is there a way to coerce this back to a T? Or is there some constraint I need to put on the T input to make this work?

31 replies

DTDrizzle Team

•Created by francis on 9/20/2024 in #help

How do you define two tables that have foreign keys to each other? drizzle-kit introspect fails

Drizzle-kit now generates a schema file with type errors - it was previously working prior to 0.24.0. In addition, the generated foreign key entry for multicolumn foreign keys has the columns in the incorrect order. (code below due to length limit)

5 replies

Aarktype

•Created by francis on 8/25/2024 in #questions

vscode intellisense errors on the same type definition in some files, but not others

e.g. the following type:

const exampleType = type({
  email: "string.email",
});

const exampleType = type({
  email: "string.email",
});

In some files, this works fine. In others, in the exact same project, this causes the following error:

Type '"string.email"' is not assignable to type '"'string' must reference a module to be accessed using dot syntax "'.
The expected type comes from property 'email' which is declared here on type 'validateObjectLiteral<{ readonly email: "string.email"; }, {}, bindThis<{ readonly email: "string.email"; }>>'

Type '"string.email"' is not assignable to type '"'string' must reference a module to be accessed using dot syntax "'.
The expected type comes from property 'email' which is declared here on type 'validateObjectLiteral<{ readonly email: "string.email"; }, {}, bindThis<{ readonly email: "string.email"; }>>'

Any ideas for what might be going on?

7 replies

DTDrizzle Team

•Created by francis on 8/8/2024 in #help

What's the recommended way to do a basic filter by a boolean column's value?

I can't figure out how to use a non-filter expression for filtering. e.g., I would expect db.select().from(appointment).where(appointment.canceled) to generate the equivalent sql select * from appointment where appointment.canceled, which works exactly how I expect. .where(() => appointment.canceled) or .where((a) => a.canceled) don't work either. I can use

.where(sql`${appointment.canceled}`)

.where(sql`${appointment.canceled}`)

but surely there must be a simpler way? I can provide this column as an argument to a filter, so why can't I use this column itself as a filter?

3 replies

Aarktype

•Created by francis on 8/7/2024 in #questions

Are there examples of how to create mapped types?

e.g. a { [K in <SomeOtherType>]: SomeValueType } type, where SomeValueType is either a static type or is computed from K. I currently have a very basic setup where I have:

type MappedEnum<T extends string> = { readonly [K in T]: T };

const kitTypes = ["test_only", "test_self_serve", "test_guided_audit"];
const kitTypeValidator = type("===", ...kitTypes);
export type KitType = typeof kitTypeValidator.infer;
export const KitType = {
  test_only: "test_only",
  test_self_serve: "test_self_serve",
  test_guided_audit: "test_guided_audit",
} as const satisfies MappedEnum<KitType>;

type MappedEnum<T extends string> = { readonly [K in T]: T };

const kitTypes = ["test_only", "test_self_serve", "test_guided_audit"];
const kitTypeValidator = type("===", ...kitTypes);
export type KitType = typeof kitTypeValidator.infer;
export const KitType = {
  test_only: "test_only",
  test_self_serve: "test_self_serve",
  test_guided_audit: "test_guided_audit",
} as const satisfies MappedEnum<KitType>;

This pattern allows me to generate, from a string array, a validator for an element in that array, a type for an element of the resulting string union type, and a type for a runtime object that lets me access the values analogous to an enum. And it's great! My question is, I have the following type definition:

type KitCounts = { [K in KitType]?: number };

type KitCounts = { [K in KitType]?: number };

Is there a way to write an ArkType function that will generate a validator for this type (and the type definition itself), given the kitTypes array? My initial attempt is:

const kitCountsValidator = type(
  Object.fromEntries(kitType.enumValues.map((k) => [`${k}?`, "number"])),
);

const kitCountsValidator = type(
  Object.fromEntries(kitType.enumValues.map((k) => [`${k}?`, "number"])),
);

Obviously, this doesn't work, since the type information is completely lost. It actually works fine at runtime but the type is useless. Is there a way to do this without writing it out explicitly? (I have 5+ other enums that I'd like to process in a similar way)

22 replies

Aarktype

•Created by francis on 8/2/2024 in #questions

How to accept a generic non-empty const string array for use in `type`?

Note: this may be a TS question, feel free to send me there instead. I have the following type: PgEnum<TValues extends [string, ...string[]]> { enumValues: TValues; } I'm attempting to create a function which accepts this object and returns an arktype instance of the union of the strings in the array, which is const with well-known values. If I call it directly from the const signature, e.g. type('===', obj.enumValues), it works fine. But I can't figure out how to accept this as a generic function parameter and use it in the body. I have tried:

function arktypeUnionFromEnum<A extends string, B extends string[]>(pgEnum: {
  enumValues: [A, ...B];
}) {
  return type("===", ...pgEnum.enumValues);
}

function arktypeUnionFromEnum<A extends string, B extends string[]>(pgEnum: {
  enumValues: [A, ...B];
}) {
  return type("===", ...pgEnum.enumValues);
}

This fails on the spread argument with

Argument of type 'A' is not assignable to parameter of type 'conform<A, unknown>'.
  Type 'string' is not assignable to type 'conform<A, unknown>'.ts(2345)

Argument of type 'A' is not assignable to parameter of type 'conform<A, unknown>'.
  Type 'string' is not assignable to type 'conform<A, unknown>'.ts(2345)

Unwinding the generic produces different errors:

function arktypeUnionFromEnum<T extends string[]>(pgEnum: { enumValues: T }) {
  return type("===", ...pgEnum.enumValues);
}

function arktypeUnionFromEnum<T extends string[]>(pgEnum: { enumValues: T }) {
  return type("===", ...pgEnum.enumValues);
}

leads to A spread argument must either have a tuple type or be passed to a rest parameter. Replacing string[] with [string, ...string[]] leads to the same issue. I suspect the two generic parameters is the closer approach, but I have no idea what this conform<> error is, or how to begin to address it.

3 replies

DTDrizzle Team

•Created by francis on 3/7/2024 in #help

New string mode date handling for postgres in 0.30.0 is potentially inconsistent

When selecting a value from a column of type timestamp with time zone, I receive the following string format: 2024-01-19 16:54:36.74526+00 When selecting a value from a column of type date, I receive the following string format: 1990-01-01T00:00:00.000Z It appears the serialization changes for the postgres driver may not be working - the ones here https://github.com/drizzle-team/drizzle-orm/blob/main/drizzle-orm/src/postgres-js/driver.ts#L27 If I define my own custom type as:

  types: {
    date_temp: {
      to: 1082,
      from: [1082],
      serialize: (val: any) => val,
      parse: (val: any) => val,
    }
  }

  types: {
    date_temp: {
      to: 1082,
      from: [1082],
      serialize: (val: any) => val,
      parse: (val: any) => val,
    }
  }

in the postgres-js options, then suddenly I receive the raw sql values I expect: 1990-01-01

2 replies

DTDrizzle Team

•Created by francis on 1/28/2024 in #help

Issuing multiple sql queries in order, but without waiting for round trips?

I need to run several commands to set up transaction variables for my use case. They need to be run in order, but I don't need the result of any of them. Is there a way to issue e.g. 3 sequential commands in a single query to the database so I don't have to wait for the round trip for each one, like happens now with await tx.execute(.....) 3 times in a row?

21 replies

DTDrizzle Team

•Created by francis on 1/16/2024 in #help

How do you create a scalar subquery in drizzle?

I want to create the drizzle analog of select * from table_a where id = (select a_id from table_b where id = <....>) This works in raw SQL, but I'm not sure how to construct the subquery to get only the a_id field to supply in the where condition.

drizzle.select().from(tableA).where(eq(tableA.id, drizzle.select(<WHAT GOES HERE>).from(tableB).where(eq(tableB.id, userInputBId))))

drizzle.select().from(tableA).where(eq(tableA.id, drizzle.select(<WHAT GOES HERE>).from(tableB).where(eq(tableB.id, userInputBId))))

11 replies

DTDrizzle Team

•Created by francis on 1/14/2024 in #help

Is there a way to get the update count resulting from a query without returning?

I would like to assert that a statement in a transaction updates only the number of rows I expect, and fail if that number is different. I can do it with result.length and returning() but that includes a ton of data I don't need.

8 replies

DTDrizzle Team

•Created by francis on 12/28/2023 in #help

Does calling `drizzle()` multiple times on the same db client instance consume more resources?

related to a question from: https://github.com/drizzle-team/drizzle-orm/issues/594 Would using the same postgres(...) client instance and creating multiple Drizzle instances from it, e.g. one for each request in a middleware, consume additional database connection resources compared to creating one drizzle instance for the whole server process? @Angelelz this is related to the question you asked in the github issue. I believe my current implementation doesn't have this problem since it uses proxies, but I was about to refactor to call drizzle() once per request and want to make sure I'm not about to blow up my database!

7 replies

DTDrizzle Team

•Created by francis on 12/28/2023 in #help

Is there a better way to extract Typescript types for use in function args than what I'm doing here?

export const client = drizzle(adminPgClient, { schema });
export type DrizzleClient = typeof client;
export type DrizzleTransaction = Parameters<Parameters<DrizzleClient["transaction"]>[0]>[0];

export const client = drizzle(adminPgClient, { schema });
export type DrizzleClient = typeof client;
export type DrizzleTransaction = Parameters<Parameters<DrizzleClient["transaction"]>[0]>[0];

It works well enough, but this seems like something which should be exposed on a library level, maybe? I need it in order to e.g. write a function which takes in a transaction and performs operations on it, such that multiple functions can be called on a single transaction and then the transaction either committed or rolled back as one unit (so the client cannot be passed in instead).

12 replies

DTDrizzle Team

•Created by francis on 12/28/2023 in #help

Has JSON aggregation been added yet (analogous to Kysely?)

I added some jsonAgg... functions to generate raw SQL to helper utilities a while back, in order to generate arrays of nested json objects in aggregation. There was discussion of adding these as first-class drizzle functions, did that ever happen? Helpers source: https://gist.github.com/rphlmr/0d1722a794ed5a16da0fdf6652902b15

4 replies

DTDrizzle Team

•Created by francis on 12/12/2023 in #help

Is there a way to explicitly type certain drizzle results as `Array<T | undefined>`?

For example, I am making a query which I know will sometimes return undefined, since I am accessing a row that may or may not be present. I'm not sure how to type this such that I have type safety through the rest of the code to ensure that I never access properties without undefined checks. I'm aware of Typescript noUncheckedIndexedAccess but that will change the behavior of all arrays everywhere in my codebase, which I do not want. I just want the ability to indicate that a certain Drizzle select result will sometimes have no items returned. e.g. const [maybeExistingEntity] = await drizzle.select(.....).limit(1) now strongly types maybeExistingEntity where I would like the type to be T | undefined.

64 replies

DTDrizzle Team

•Created by francis on 12/12/2023 in #help

Is there a way to configure pg introspect to generate date rather than string inference columns?

My workflow involves writing migrations by hand outside of drizzle and using schema introspection to pull in changes. This works well except for date columns, which are always generated as mode: 'string' by default.

16 replies

DTDrizzle Team

•Created by francis on 9/15/2023 in #help

Creating join table that is simultaneously many-to-many and one-to-many

got a data modeling / sql question for you all: I have an entity A with either a many-to-many relationship to B, or a one-to-many relationship to C. There must be a unique ordering of all relationships from A to B or C, such that we can define an ordering such as B1, B2, C1, B3, C2... etc B and C are different entities, with wildly different table structures, and therefore cannot be combined into a single table. I am solving this with: A, B, and C all have id primary key fields. I create a table a_joins as:

create table a_joins (
    id uuid primary key default uuid_generate_v4(),
    a_id uuid not null references a (id) on delete cascade,
    b_id uuid references b (id) on delete restrict,
    c_id uuid unique references c (id) on delete restrict,
    index integer not null,
    check ((b_id is not null) != (c_id is not null)),
    unique (a_id, b_id),
    unique (a_id, index)
);

create table a_joins (
    id uuid primary key default uuid_generate_v4(),
    a_id uuid not null references a (id) on delete cascade,
    b_id uuid references b (id) on delete restrict,
    c_id uuid unique references c (id) on delete restrict,
    index integer not null,
    check ((b_id is not null) != (c_id is not null)),
    unique (a_id, b_id),
    unique (a_id, index)
);

This allows me to ensure an ordering for all a_id entries, such that each entry in the ordering points to either a b (shared by many a) or a c (for this a only). The problem I have is that I want cascade delete behavior, such that on deletion of a, the a_joins entries are deleted (which works due to cascade), and any c associated with a is deleted as well. How would you recommend I accomplish this? I attempted to add a foreign key relationship to c to perform the cascade:

alter table c add constraint c_a_fkey
foreign key (id, a_id)
references a_joins (c_id, id) on delete cascade;

alter table c add constraint c_a_fkey
foreign key (id, a_id)
references a_joins (c_id, id) on delete cascade;

This should ensure both that a c entry only ever points to the single unique a_joins entry that references it, and that the c entry is deleted when the corresponding a_joins entry is removed. However, this fails due to the lack of a unique index on a_joins (c_id, id), even though both of those columns are individually marked as either unique or primary key. How is this possible?

32 replies

Gaming

Programming