Added maxBytes support.

[dmedina2015]

(1) Added support to maxBytes option, using similar logic from skipper-disk. Behavior:

• An error is thrown when upload stream exceeds bytes defined in maxBytes parameter.
• Limit is applied for all upload stream (all files in the same request, not for each file individually)
• Files uploaded before limit is reached are saved in GridFS. Only the one that exceeds and the subsequent are not saved.
• Garbage of unfinished upload is removed using GridFSBucketWriteStream.abort() function

(2) Added support to onProgress using same logic from skipper-disk

@mikermcneil or @willhuang85 , could you review it?

[willhuang85]

Hey @dmedina2015, awesome work! I’ll try to review your PRs over the weekend as I haven’t really had time to work with SailsJS in awhile.

[dmedina2015]

Hey @willhuang85, thank you very much for your message. I'm happy to help the project. Below are some attention points for your analysis:

(1) I commented the listeners __newFile.once('close') and __newFile.once('done') since they were being called together with its outs__ correspondents. As both emits events in receiver__, receiver__.once('error') was being called twice every time an error (like E_EXCEEDS_UPLOAD_LIMIT for instance).

(2) I used the exact same logic of skipper-disk for the onProgress feature. However, I'm a little bit concerned about this piece of code:

module.exports = function buildProgressStream (options, __newFile, receiver__, outs__) {
    options = options || {};
    var log = options.log || function noOpLog(){};
 
    // Generate a progress stream and unique id for this file
    // then pipe the bytes down to the outs___ stream
    // We will pipe the incoming file stream to this, which will
    var localID = _.uniqueId();
    var guessedTotal = 0;
    var writtenSoFar = 0;
    var __progress__ = new TransformStream();
    __progress__._transform = function(chunk, enctype, next) {

        // Update the guessedTotal to make % estimate
        // more accurate:
        guessedTotal += chunk.length;
        writtenSoFar += chunk.length;

I cannot understand why guessedTotal is initialized with 0 and added by chunk.length every time a new chunk arrives. This is causing guessedTotal and writtenSoFar to be exactly equal every time and the percent is always 100%.
My understanding is that it must be initialized with the total size of the stream: guessedTotal=__newfile.stream.byteCount and not be added by chunk.length() as today. If you are ok with my logic, I can update my branch.

(3) I strongly recommend you to take a look at PR #44 . It fix a bug identified in issue #39 and I think it is important to get in master due to a callback being called twice.

Greetings from Brazil