How to store files with meta data in LoopBack?

问题

What I want to do: Have an html form, with a file input inside. When a file is chosen, the file input should upload the file, and get a file id, so when the form is submitted, the file id is posted with the form and written in the database.

Shorter version: I want to store meta data (id for example) with my files.

Sounds simple, yet I struggle to do that in LoopBack.

There has been a couple conversations ( 1, 2 ) about this topic, and neither seemed to lead to a solution, so I thought this might be a good place to find one once and for all.

The simplest solution would be to use model relations, but LoopBack doesn\'t support relations with the file storage service. Bump. So we have to go with a persistedmodel named File for example, and override default create, delete so it saves and deletes from the file store model I have - named Storage.

My setup so far:

I have a model /api/Storage which is connected to a loopback storage service and is saving file successfully to the local filesystem.
I have a PersistedModel connected to Mongo with file meta data: name,size, url and objectId
I have a remote hook set up beforecreate so the file can be saved first and then it\'s url can be injected into File.create()

I\'m there, and according to this LoopBack page, I have the ctx which should have the file inside:

File.beforeRemote(\'create\', function(ctx, affectedModelInstance, next) {})`

What\'s ctx?

ctx.req: Express Request object.
ctx.result: Express Response object.

Ok, so now I\'m at the Express page, pretty lost, and it sais something about a \'body-parsing middleware\' which I have no idea what it might be.

I feel like I\'m close to the solution, any help would be appreciated. Is this approach right?

回答1:

Here's the full solution for storing meta data with files in loopback.

You need a container model

common/models/container.json

{
  "name": "container",
  "base": "Model",
  "idInjection": true,
  "options": {
    "validateUpsert": true
  },
  "properties": {},
  "validations": [],
  "relations": {},
  "acls": [],
  "methods": []
}

Create the data source for your container in server/datasources.json. For example:

...
"storage": {
    "name": "storage",
    "connector": "loopback-component-storage",
    "provider": "filesystem", 
    "root": "/var/www/storage",
    "maxFileSize": "52428800"
}
...

You'll need to set the data source of this model in server/model-config.json to the loopback-component-storage you have:

...
"container": {
    "dataSource": "storage",
    "public": true
}
...

You'll also need a file model to store the meta data and handle container calls:

common/models/files.json

{
  "name": "files",
  "base": "PersistedModel",
  "idInjection": true,
  "options": {
    "validateUpsert": true
  },
  "properties": {
    "name": {
      "type": "string"
    },
    "type": {
      "type": "string"
    },
    "url": {
      "type": "string",
      "required": true
    }
  },
  "validations": [],
  "relations": {},
  "acls": [],
  "methods": []
}

And now connect files with container:

common/models/files.js

var CONTAINERS_URL = '/api/containers/';
module.exports = function(Files) {

    Files.upload = function (ctx,options,cb) {
        if(!options) options = {};
        ctx.req.params.container = 'common';
        Files.app.models.container.upload(ctx.req,ctx.result,options,function (err,fileObj) {
            if(err) {
                cb(err);
            } else {
                var fileInfo = fileObj.files.file[0];
                Files.create({
                    name: fileInfo.name,
                    type: fileInfo.type,
                    container: fileInfo.container,
                    url: CONTAINERS_URL+fileInfo.container+'/download/'+fileInfo.name
                },function (err,obj) {
                    if (err !== null) {
                        cb(err);
                    } else {
                        cb(null, obj);
                    }
                });
            }
        });
    };

    Files.remoteMethod(
        'upload',
        {
            description: 'Uploads a file',
            accepts: [
                { arg: 'ctx', type: 'object', http: { source:'context' } },
                { arg: 'options', type: 'object', http:{ source: 'query'} }
            ],
            returns: {
                arg: 'fileObject', type: 'object', root: true
            },
            http: {verb: 'post'}
        }
    );

};

For expose the files api add to the model-config.json file the files model, remember select your correct datasources:

...
"files": {
    "dataSource": "db",
    "public": true
}
...

Done! You can now call POST /api/files/upload with a file binary data in file form field. You'll get back id, name, type, and url in return.

回答2:

I had the same problem. I solved it by creating my own models to store meta data and my own upload methods.

I created a model File which will store info like name,type,url,userId ( same as yours)
I created my own upload remote method because I was unable to do it with the hooks. Container model is the model which is created by loopback-component-storage.

var fileInfo = fileObj.files.myFile[0]; Here myFile is the fieldname for file upload, so you will have to change it accordingly. If you don't specify any field, then it will come as fileObj.file.null[0]. This code lacks proper error checking, do it before deploying it in production.

 File.uploadFile = function (ctx,options,cb) {
  File.app.models.container.upload(ctx.req,ctx.result,options,function (err,fileObj) {
    if(err) cb(err);
    else{
            // Here myFile is the field name associated with upload. You should change it to something else if you
            var fileInfo = fileObj.files.myFile[0];
            File.create({
              name: fileInfo.name,
              type: fileInfo.type,
              container: fileInfo.container,
              userId: ctx.req.accessToken.userId,
              url: CONTAINERS_URL+fileInfo.container+'/download/'+fileInfo.name // This is a hack for creating links
            },function (err,obj) {
              if(err){
                console.log('Error in uploading' + err);
                cb(err);
              }
              else{
                cb(null,obj);
              }
            });
          }
        });
};

File.remoteMethod(
  'uploadFile',
  {
    description: 'Uploads a file',
    accepts: [
    { arg: 'ctx', type: 'object', http: { source:'context' } },
    { arg: 'options', type 'object', http:{ source: 'query'} }
    ],
    returns: {
      arg: 'fileObject', type: 'object', root: true
    },
    http: {verb: 'post'}
  }

);

回答3:

For those who are looking for an answer to the question "how to check file format before uploading a file".

Actual in this case we can use optional param allowedContentTypes.

In directory boot use example code:

module.exports = function(server) {
    server.dataSources.filestorage.connector.allowedContentTypes = ["image/jpg", "image/jpeg", "image/png"];
}

I hope it will help someone.

回答4:

Depending on your scenario, it may be worth looking at utilising signatures or similar allowing direct uploads to Amazon S3, TransloadIT (for image processing) or similar services.

Our first decision with this concept was that, as we are using GraphQL, we wanted to avoid multipart form uploads via GraphQL which in turn would need to transfer to our Loopback services behind it. Additionally we wanted to keep these servers efficient without potentially tying up resources with (large) uploads and associated file validation and processing.

Your workflow might look something like this:

Create database record
Return record ID and file upload signature data (includes S3 bucket or TransloadIT endpoint, plus any auth tokens)
Client uploads to endpoint

For cases where doing things like banner or avatar uploads, step 1 already exists so we skip that step.

Additionally you can then add SNS or SQS notifications to your S3 buckets to confirm in your database that the relevant object now has a file attached - effectively Step 4.

This is a multi-step process but can work well removing the need to handle file uploads within your core API. So far this is working well from our initial implementation (early days in this project) for things like user avatars and attaching PDFs to a record.

Example references:

http://docs.aws.amazon.com/AmazonS3/latest/dev/UsingHTTPPOST.html

https://transloadit.com/docs/#authentication

回答5:

For anyone else having that problem with loopback 3 and Postman that on POST, the connection hangs (or returns ERR_EMPTY_RESPONSE) (seen in some comments here)... The problem in this scenario is, that Postman uses as Content-Type "application/x-www-form-urlencoded"!

Please remove that header and add "Accept" = "multipart/form-data". I've already filed a bug at loopback for this behavior

回答6:

Just pass the data as "params" object and at server you can get it as ctx.req.query

For example

At client side

Upload.upload(
{
    url: '/api/containers/container_name/upload',
    file: file,
    //Additional data with file
    params:{
     orderId: 1, 
     customerId: 1,
     otherImageInfo:[]
    }
});

At Server side

Suppose your storage model name is container

Container.beforeRemote('upload', function(ctx,  modelInstance, next) {
    //OUPTUTS: {orderId:1, customerId:1, otherImageInfo:[]}
    console.log(ctx.req.query); 
    next();
})

回答7:

For AngularJS SDK users... In case you would like to use generated methods like Container.upload(), you might want to add a line to configure the method in lb-services.js to set Content-Type headers to undefined. This would allow client to set Content-Type headers and add boundary value automatically. Would look something like this:

 "upload": {
    url: urlBase + "/containers/:container/upload",
    method: "POST",
    headers: {"Content-Type": undefined}
 }

来源：https://stackoverflow.com/questions/28885282/how-to-store-files-with-meta-data-in-loopback

标签

loopbackjs

strongloop