Thanks to Alex Mckay I had a resolve for dynamic setting a props:

  for(let prop in filter)
      (state.filter as Record<string, any>)[prop] = filter[prop];

When we do something like this obj[key] Typescript can't know for sure if that key exists in that object. What I did:

Object.entries(data).forEach(item => {
    formData.append(item[0], item[1]);
});

With out typescript error

    const formData = new FormData();
    Object.keys(newCategory).forEach((k,i)=>{  
        var d =Object.values(newCategory)[i];
        formData.append(k,d) 
    })

I have made a simulation of the problem. looks like the issue is how we should Access Object Properties Dynamically Using Bracket Notation in Typescript

interface IUserProps {
  name: string;
  age: number;
}

export default class User {
  constructor(private data: IUserProps) {}

  get(propName: string): string | number {
    return this.data[propName as keyof IUserProps];
  }
}

I found a blog that might be helpful to understand this better.

here is a link https://www.nadershamma.dev/blog/2019/how-to-access-object-properties-dynamically-using-bracket-notation-in-typescript/

I use this:

interface IObjectKeys {
  [key: string]: string | number;
}

interface IDevice extends IObjectKeys {
  id: number;
  room_id: number;
  name: string;
  type: string;
  description: string;
}

NOTE: "[key: string]" what is it? An object in JavaScript is primarily just a collection of properties made up of key-value pairs. Moreover, the key can only be a string (even for array elements), but the value can be any data type.

If you use the optional property in your object:

interface IDevice extends IObjectKeys {
  id: number;
  room_id?: number;
  name?: string;
  type?: string;
  description?: string;
}

... you should add 'undefined' value into the IObjectKeys interface:

interface IObjectKeys {
  [key: string]: string | number | undefined;
}

I made some small changes to Alex McKay's function/usage that I think make it a little easier to follow why it works and also adheres to the no-use-before-define rule.

First, define this function to use:

const getKeyValue = function<T extends object, U extends keyof T> (obj: T, key: U) { return obj[key] }

In the way I've written it, the generic for the function lists the object first, then the property on the object second (these can occur in any order, but if you specify U extends key of T before T extends object you break the no-use-before-define rule, and also it just makes sense to have the object first and its' property second. Finally, I've used the more common function syntax instead of the arrow operators (=>).

Anyways, with those modifications you can just use it like this:

interface User {
  name: string;
  age: number;
}

const user: User = {
  name: "John Smith",
  age: 20
};

getKeyValue(user, "name")

Which, again, I find to be a bit more readable.

When using Object.keys, the following works:

Object.keys(this)
    .forEach(key => {
      console.log(this[key as keyof MyClass]);
    });

public getUserName(): string {

const accessToken = this.getAccessToken();
const claims:any = this.getUserClaims();
console.log('access token ',accessToken);
this.getUserInfo();
return claims['sub'].split('@')[0];

}

//give any type to the variable

I know it's a little too late, but all it's needed is to add a little type conversion, I wrote a static function that safely-returns the array of keys with the correct typing. All you need is to define the type and pass the object as a parameter:

export class ObjectUtil {
  public static getObjectKeys<T>(obj: Object) {
    if (!obj) {
      return [];
    }

    return Object.keys(obj).map((key: string) => key as keyof T);
  }
}

Below is a simple example:

ObjectUtil.getObjectKeys<Address>(address).forEach((key) => {
  console.log(address[key]);
});

This is not a answer to the original question, but a generic work around to this problem.

Original problem: person[cr.field] causes this error

I'm doing a generic advanced search form where user can select a field, comparator and the desired value. When trying to read the value from the object based on the key, I get this error (altought the field value is type of string and I think it should be just fine)

So what I do is I extract the [key, value] like this

const x: [string, any] = Object.entries(person).find(([key, _]) => key === cr.field);

For example if my criterion (cr) is { field: 'name', value: 'John' } and field name actually exists in a person obj., it should return the field name and the value as tuple (x is [string, any] or undef). If not found, undefined.

For anyone who stumbles upon this in the future:

If you're getting the TypeScript error

'...expression of type string cannot be used to index...'

then simply specify that the 'expression of type string' is a key of the type of that object. For example,

const someObj:ObjectType = data;
const field = 'username';

// This gives an error
const temp = someObj[field];

// Solution 1: When the type of the object is known
const temp = someObj[field as keyof ObjectType]

// Solution 2: When the type of the object is not known
const temp = someObj[field as keyof typeof someObj]

This is what it worked for me. The tsconfig.json has an option noImplicitAny that it was set to true, I just simply set it to false and now I can access properties in objects using strings.

// bad
const _getKeyValue = (key: string) => (obj: object) => obj[key];

// better
const _getKeyValue_ = (key: string) => (obj: Record<string, any>) => obj[key];

// best
const getKeyValue = <T extends object, U extends keyof T>(key: U) => (obj: T) =>
  obj[key];

Bad - the reason for the error is the object type is just an empty object by default. Therefore it isn't possible to use a string type to index {}.

Better - the reason the error disappears is because now we are telling the compiler the obj argument will be a collection of string/value (string/any) pairs. However, we are using the any type, so we can do better.

Best - T extends empty object. U extends the keys of T. Therefore U will always exist on T, therefore it can be used as a look up value.

Here is a full example:

I have switched the order of the generics (U extends keyof T now comes before T extends object) to highlight that order of generics is not important and you should select an order that makes the most sense for your function.

const getKeyValue = <U extends keyof T, T extends object>(key: U) => (obj: T) =>
  obj[key];

interface User {
  name: string;
  age: number;
}

const user: User = {
  name: "John Smith",
  age: 20
};

const getUserName = getKeyValue<keyof User, User>("name")(user);

// => 'John Smith'

Alternative Syntax

const getKeyValue = <T, K extends keyof T>(obj: T, key: K): T[K] => obj[key];

This happens because you try to access plotOptions property using string name. TypeScript understands that name may have any value, not only property name from plotOptions. So TypeScript requires to add index signature to plotOptions, so it knows that you can use any property name in plotOptions. But I suggest to change type of name, so it can only be one of plotOptions properties.

interface trainInfo {
    name: keyof typeof plotOptions;
    x: Array<number>;
    y: Array<number>;
    type: string;
    mode: string;
}

Now you'll be able to use only property names that exist in plotOptions.

You also have to slightly change your code.

First assign array to some temp variable, so TS knows array type:

const plotDataTemp: Array<trainInfo> = [
    {
      name: "train_1",
      x: data.filtrationData.map((i: any) => i["1-CumVol"]),
      y: data.filtrationData.map((i: any) => i["1-PressureA"]),
      type: "scatter",
      mode: "lines"
    },
    // ...
}

Then filter:

const plotData = plotDataTemp.filter(({ name }) => plotOptions[name]);

If you're getting data from API and have no way to type check props at compile time the only way is to add index signature to your plotOptions:

type tplotOptions = {
    [key: string]: boolean
}

const plotOptions: tplotOptions = {
    train_1: true,
    train_2: true,
    train_3: true,
    train_4: true
}