Support key exclusions

[Richienb]

Something like:

const filterObject = require('filter-obj');

filterObject({a: 1, b: 1, c: 1} {keys: ['!a']});
//=> {b: 1, c: 1}

Source: sindresorhus/query-string#282 (comment)

[sindresorhus]

Makes sense 👍🏻

[faizanu94]

@Richienb this feature is available through below function being passed as predicate, also has a flexibility for keys as well as values:

filterObject({a: 1, b: 1, c: 1}, (key, value) => key !== 'a');

[Richienb]

@faizanu94 This will be used for sindresorhus/query-string#282 (comment)

[faizanu94]

I would say then exclusions should be available for values as well, @sindresorhus your thoughts on this?

[sindresorhus]

@faizanu94 I don't think that is as common of a need. If someone needs that, they can propose it then or just use the filter. I'm mostly just interested in the API sugar for keys.

[ehmicky]

Please correct if I'm wrong, but I believe, since we are using string matching and not globbing patterns, it would not make sense to mix excluded keys with included keys, such as:

filterObject({a: 1, b: 1, c: 1}, ['!a', 'b']);

Since it would have the same behavior as passing only ['b'].

Based on this, if a user wants to exclude keys, they would probably want to prepend ! to all keys, which might be slightly tedious:

filterObject({a: 1, b: 1, c: 1}, ['!a', '!b', '!c', '!d', '!e', '!f']);

Also, behavior might be ambiguous when both including and excluding the same key. Either should have priority, which would need to be documented and might create some confusion among some users.

filterObject({a: 1, b: 1, c: 1}, ['!a', 'a']);

I am wondering whether it would be simpler instead to export a separate function negating the predicate? It would also be more performant since the logic would need to do fewer argument checks and could be optimized for each case. Finally, it would also work with predicate functions.

import { include, exclude } from 'filter-obj';

include({a: 1, b: 2, c: 3}, ['a']); // {a: 1}
exclude({a: 1, b: 2, c: 3}, ['a']); // {b: 2, c: 3}

include({a: 1, b: 2, c: 3}, (key, value) => value === 1); // {a: 1}
exclude({a: 1, b: 2, c: 3}, (key, value) => value === 1); // {b: 2, c: 3}

[Richienb]

Yes, that is a problem I immediately stumbled upon as soon as I began to think of how to implement it. If you're suggesting to create separate include and exclude functions then logically, the original function should lose that functionality for simplicity. The new API could be something like this:

• excludeObjectKeys
• filterObjectKeys to communicate that it is the most common use, includeObjectKeys to be more explicit and to show that it doesn't filter values or pickObjectKeys for a more TypeScript approach that is also analogous to query-string
• filterObject to be short and sweet or filterObjectPredicate to be explicit

[ehmicky]

I was actually going to suggest this too! Meaning: breaking it into three methods instead: two with array of keys (with/without negation) and one with the function predicate.

When it comes to the names, I like them too, with two additional questions:

• Should ObjectKeys be shortened to just Keys? Since keys are usually understood to be object keys in JavaScript.
• I like pick*() too as it is what Lodash and some other functional libraries use. However, the other method should probably be omit*() instead of exclude*() then, since those are more "contrary". Thoughts?

[Richienb]

Should ObjectKeys be shortened to just Keys?

The Object part is more about keeping the naming similar to the predicate function though I suppose this more pragmatic approach would be fine.

the other method should probably be omit*()

Sure, though I'm not sure whether the names need to conform with other Sindre modules.

@sindresorhus What do you think?

[sindresorhus]

I prefer includeKeys and excludeKeys.