[Feature] Schema#removeIndex(...)

[AbdelrahmanHafez]

I came across a case where I cloned a schema, and would like to reuse it in another model without the indexes.
The issue is that the cloned schema includes its indexes, and there is no official way to remove indexes, I had to access schema._indexes which doesn't work with TS.

const cartSchema = new Schema({ customerId: String, products: [productSchema]  });
cartSchema.index({ customerId: 1 }, { unique: true });

const cartSchemaClone = cartSchema.clone();

cartSchemaClone._indexes = []; // there is no other way to do it.

const checkoutSchema = new Schema({ cart: cartSchemaClone });

[AbdelrahmanHafez]

Turns out cartSchemaClone._indexes = []; doesn't really remove all indexes, it only removes the top-level paths indexes and ignores nested paths indexes.

The workaround that I found for removing all indexes:

const cartSchemaClone = cartSchema.clone();
cartSchemaClone.set('excludeIndexes', true); // ignore _all_ indexes
const checkoutSchema = new Schema({ cart: cartSchemaClone });

This fixes my issue, but would become a problem once I need to add some indexes. I think we need to provide:
Schema#clearIndexes and Schema#removeIndex.

[AbdelrahmanHafez]

@vkarpov15
I can imagine the API for Schema#clearIndexes();
How do you think the API for Schema#removeIndex(): should look like?
One path can have multiple indexes, one option would be to make const indexId = schema.index({ status: 1 }); return an indexId which is a randomoy generated string, we can assign that id as metadata in schema._indexes, and allow users to remove using those ids.

Admittedly, this is a hacky solution that I can see causing issues for users thinking that this identifier representing something on MongoDB server, but I am not sure what other way we can reliably identify indexes in our schemas.

Another way would be to identify using the object reference by strict equality, that way we make it the dev's responsibility to find the desired index from Schema#indexes();

[vkarpov15]

We should support both schema.removeIndex('my-index-name') and schema.removeIndex({ status: 1 }) . The latter should remove any index that is deep equal to the given spec.

Also schema.clearIndexes() should remove all indexes defined on the schema. Shouldn't have any impact on indexes in MongoDB

[AbdelrahmanHafez]

I agree that removing an index from schema shouldn't touch MongoDB.

For deep equality check, I think we should check both the index definition and the options. Also, deep equality check should treat objects with the same properties in different order as equal.

For string index names, I am not sure if they're unique or not, one other concern, AFAIK people can set a custom index name, if we wanna match with the index name, does that mean we'll need to fetch indexes from the database? Needing to fetch indexes means we'll have to tie the index removal with database connections.

I think supporting string index names, knowing that users can set custom indexes names, will introduce more issues than the feature requires, deep equality check should do the trick.