pre, post middleware are not executed on findByIdAndUpdate

Because a findAndUpdate method is presented to cut down the following code:

 Model.findById(_id, function (err, doc) {
      if (doc) {
          doc.field = 'value';
          doc.save(function (err) {
                 // do something;
          });
      }
 });

to this:

Model
   .findByIdAndUpdate(_id, {$set: {field: 'value'}}, function (err, doc) {
        // do something 
    });

We need to use pre, post middleware exactly the same. At now pre, post middleware are not executed when I make findByIdAndUpdate.

[albanm]

Hello,

I know the issue is closed, but I am not entirely satisfied with the answer. Shouldn’t there be at least a post-save or post-update middleware for findOneAndUpdate and other similar operations ? Is seems ok since the document is returned. Not feasible for pre middlewares or for Model.update, i agree.

It would greatly improve the capabilities of plugins such as mongoosastic that is currently blind to some operations it should be able to support.

If no middleware, does someone have an idea on how to manage some post update operations in a plugin ?

Thanks

[joerx]

While I sort of understand the reasoning, the lack of hooks on atomic updates IMHO makes Mongoose somewhat pointless in practice. When I use atomic updates any validations, defaults, etc. are not executed, so the entire purpose of using an ODM is defeated. Using find/save will do the job, but is there any guarantee this is always used?

Moreover, usually I would try to avoid find/save since it’s not an atomic operation. MongoDB compensates it’s lack of transaction support by providing powerful atomic query & update features. So I would use these atomic operations but w/o middleware support Mongoose won’t provide much value over the native MongoClient.

Even the examples in http://aaronheckmann.tumblr.com/post/48943525537/mongoose-v3-part-1-versioning would use update, hence bypass middleware. I can use versioning properly or middleware but not combine both? Really, where’s the point in having it?

I don’t even fully understand the technical reasons: If update & co. wrap around the database operations, why can’t we intercept the call and pass the query objects so we can do some validation/customization before we actually do the update?