• truncate Rest from OAuth
• first-level support for streams (both request + response)
• rethink mechanisms to fetch all pages (e.g.: let promises take care of that?)
• tiny wrapper around request module

E.g.:

client.products.queryString("limit=10&where=name(en = \"Foo\")&staged=true").fetch()

Same for products search.

see d10dbed

• fetch should be configured to run in batches

It would be nice to get the language settings etc.

Currently, the request body payload is internally stringified before calling node-connect functions, but it should also work passing it as JSON object.

• login
• password
• email

Useful for export

ws = fs.createWriteStream('./output.json', {encoding: 'utf-8'})
client.products.fetchAsStream(ws)

https://www.npmjs.com/package/request#streaming

To avoid saving lots of result pages in memory we can pass an option to avoid that.

https://github.com/sphereio/sphere-node-client/blob/master/src/coffee/services/base.coffee#L385

Moin

i'm using the client the get a channel by id client.channels.byId('OrderXmlFileExport'). The result GET request looks like that

https://api.sphere.io/kokon-staging-1/channels/OrderXmlFileExport?limit=100

why do we need a parameter limit when asking for a single resource using byId(...)?

http://dev.sphere.io/http-api-projects-products-search.html#suggest-query

Current:
@client.productTypes.perPage(0).fetch()

Better:
@client.productTypes.all().fetch()

client.carts.byCustomer(id).fetch()

See old task sphereio/sphere-node-connect#7

At the moment the process function just recursively goes through the pages by incrementing the offset on each iteration, assuming the total is fixed.

There are cases though when the processed function changes the data model and influences the total results of the current query.
Some examples:

• when deleting the resource on each recursion
• when querying for published products and unpublishing the product on each iteration

To fix that, the process method should always check for changes of total and increment the offset only if the total is the same as before.

See https://github.com/commercetools/sphere-order-xml-export/blob/master/src/coffee/channelservice.coffee#L19

• facets
• filters

"user_agent": "sphere-node-sdk"

See https://github.com/petkaantonov/bluebird/blob/master/API.md as a guideline

The sphere.io API has a "hidden" feature for retrieving project infos by sending a GET request https://api-v0.sphere.io/my-project. We at elastic.io use this API call to retrieve all the languages configured for a project.

It would be nice if this client would provide a support for this feature.

Before:

?lang=en&text=hello+world

After:

?text.en=hello+world

sphereio/sphere-node-sync#24

client.shippingMethods.byCart(id).fetch()
client.shippingMethods.byLocation(location).fetch()

When fetching something from SPHERE.IO and use page(0) I get the same results as page(1).
Yes, the documentation defines 1 as the default.
It would be very developer friendly when we could throw an exceptions here as most of us start thinking from 0 ;)

• repo name
• references
• travis
• coveralls
• npm

Example: client.products.parallel(5).fetch()

In many situations it is helpful to get only the changed resource within a specific time frame.
What about this API:

client.customers.last('7d').fetch()
client.inventoryEntries.last('2h').fetch()
client.messages.last('30s').fetch()

• s for seconds
• h for hours
• d for days

We could also add m for month and a for year.

It is not supported for orders. But i would check if it is the case for other items like products, customers etc. too.

• https://github.com/mikejihbe/metrics
• https://github.com/etsy/statsd (has also graphite support)

I didn't see it in readme documentation, can you add, please?

Would be nice to use a logging fromework where you can change the log configuration to another log level like trace/debug for debugging. This would really help when working with the client to see whats going on (like target url, request/response header, response body etc.).

I'm trying get get a channel resource using method byId() but the promise is never fullfilled, instead a timeout occurs. The only logged message is "start", so then(), fail(result) or fin() are never called. Any idea why this happens?

console.log "start"
@client.channels.byId(channelId).fetch()
.then(channel) ->
  console.log "then"
.fail(result) ->
  console.log "fail"
.fin ->
  console.log "finally"

i also can see that the underlying request seems to return with status code 404 (might be ok if the channel doesn't exist yet) but the proper fail(result) method is not triggered though (i would expect that).

{
  "name": "sphere-node-client",
  "hostname": "Miami-2.local",
  "pid": 10946,
  "widget_type": "sphere-node-connect",
  "level": 20,
  "request": {
    "type": "REQUEST",
    "uri": {
      "protocol": "https:",
      "slashes": true,
      "auth": null,
      "host": "api.sphere.io",
      "port": 443,
      "hostname": "api.sphere.io",
      "hash": null,
      "search": "?limit=100",
      "query": "limit=100",
      "pathname": "\/kokon-staging-1\/channels\/OrderXmlFileExport",
      "path": "\/kokon-staging-1\/channels\/OrderXmlFileExport?limit=100",
      "href": "https:\/\/api.sphere.io\/kokon-staging-1\/channels\/OrderXmlFileExport?limit=100"
    },
    "method": "GET",
    "headers": {
      "User-Agent": "sphere-node-connect",
      "Authorization": "Bearer ADRjkM3XFtNmp70uAcMXQQcFXYrzjmqf",
      "accept": "application\/json"
    }
  },
  "response": {
    "type": "RESPONSE",
    "status": 404,
    "headers": {
      "server": "nginx",
      "date": "Wed, 02 Apr 2014 13:52:18 GMT",
      "transfer-encoding": "chunked",
      "connection": "keep-alive",
      "x-served-by": "app1.sphere.prod.commercetools.de",
      "x-served-config": "sphere-projects-ws-1.0",
      "access-control-allow-origin": "*",
      "access-control-allow-headers": "Accept, Authorization, Content-Type, Origin",
      "access-control-allow-methods": "GET, POST, DELETE, OPTIONS"
    }
  },
  "msg": "Rest response",
  "time": "2014-04-02T13:52:21.384Z",
  "v": 0
}

• use debug module
• make it "logger-unaware", meaning there are no logger deps (e.g. bunyan, winston) but a logger object is assumed to be given and used like logger.info {foo: 'bar'}, 'This is foo bar' and let the application decide what logger to use

https://github.com/petkaantonov/bluebird

When you have a lot of thousands of resources to process, it's not a good idea to first fetch all elements as it would consume a lot or memory.
Thus, it would be nice to pass a process function that is called for each page resp each element of each page.
What about passing a function into fetch?

Currently some services don't support all methods inherited from BaseService (e.g.: the messages endpoint is a read-only endpoint).

A service class should be composed by only the pieces it needs and not inherits everything.