After internal discussion we came to the conclusion that this would be an inconsistent solution - it misleads people into thinking that it is fn that does the filtering when it is spread that does it, it introduces new complex concepts to solve a very small problem.

This problem is easier to solve with native methods, although it may be a slightly more verbose solution

sample({
  clock: trigger,
  fn: (upd) => ({
    foo: 0,
    ...(upd.both ? {bar: 0} : null),
    ...(upd.baz ? {baz: 0} : null),
  }),
  target: spread({
    foo,
    bar,
    baz,
  })
})

But the underlying issue remains opened:

1. A trigger from a single data source is required
2. For multiple units at once
3. With support for data transformation
4. With the ability to skip updates of any of them by arbitrary criteria

We'll continue to think about this idea, maybe go back to the option suggested above, maybe solve it some other way

Actually, it looks like someting that prepend operator (proposed quite a while ago) should handle

sample({
 clock: trigger,
 target: spread({
    first: $first,
    second: prepend({
      filter: $allowed,
      fn: mapData
      target: $second
    }),
  }),
});

It does check all boxes:

• Supports the ability to skip updates by arbitrary criteria
• Supports data transformation
• Can be applied to multiple units (one can combine it with another spread to make it multiple units at once)