interview-tests

Welcome!

Setup

You will need Visual Studio Code and Live Share extension.

Repository

nodejs/0001 contains sample application. Your task is to create implementation in nodejs/0001/src/job.ts. Rest of files is not important for you, it's just testing/reviewing framework.

What you have

You don't need to import anything else. Your job has everything prepared.

Mongo

In job.ts you have MongoClient class available. Just instantiate it like ordinary MongoClient. You don't need any connection string. It has some restricted set of methods analogous to real Mongo driver.

Your working Mongo DB is test_0001. There is collection with source data prepared for you. Its name is source_data :) It contains 100.000 documents which you should process in the method.

Sink

Second library is sink. It consumes transformed data.

function initMultipartUpload(): string creates new Multipart upload and returns its identifier. You will feed the data into one Multipart upload.

function uploadPart(multipartIdentifier: string, partNumber: number, data: string): string uploads one part of the data. The partNumber is ordering of your part. If you upload data to same partNumber twice the latter will overwrite data of the former upload. But it's up to you whether you will number the parts like 1, 2, 3 of 10, 15, 100. Just the ordering is important. The method returns ETag which is needed to check upload consistency.

function completeMultipartUpload(multipartIdentifier: string, partDescriptors: PartDescriptor[]): void completes the Multipart upload. It takes descriptors of all parts which should be in the resulting upload (typically all of them).

export interface PartDescriptor {
    number: number
    etag: string
}

The number is number of part and etag is value returned from UploadPart() function. etag is used to ensure consistency. If it doesn't match the value returned by uploadPart() the completeMultipartUpload() call fails.

Job

Your job has restricted lifetime. Your process() function has shouldEnd callback. Whenever the callback returns true your job is about to die (timeout). That means you have to save your current state and resume when your job is called again. If shouldEnd() returns true, it means you can do just less than 1.200 Mongo operations left (retrieving 1 document from cursor also counts as operation). Your method should save its state and end ASAP with result ProcessResult.NotYetDone.

The task

You need to:

• Take all data from source data Mongo collection.
• Order them by order field; the order field value is unique among all documents; its values are greater than 0.
• Transform them to output format.
• Send the output data to sink.

Source data look like this:

/* 1 */
{
    "_id" : ObjectId("6426bc7a43b2bdaf4efe65b8"),
    "first_name" : "Exciting",
    "last_name" : "Randene",
    "index" : 1,
    "order" : 100
}

/* 2 */
{
    "_id" : ObjectId("6426bc7a43b2bdaf4efe668c"),
    "first_name" : "Promising",
    "last_name" : "Ranna",
    "index" : 213,
    "order" : 288
}

/* 3 */
{
    "_id" : ObjectId("6426bc7a43b2bdaf4efe66e1"),
    "first_name" : "Continuing",
    "last_name" : "Hedi",
    "index" : 298,
    "order" : 203
}

/* 4 */
{
    "_id" : ObjectId("6426bc7a43b2bdaf4efe69d7"),
    "first_name" : "Printed",
    "last_name" : "Aurelie",
    "index" : 1056,
    "order" : 1045
}

The resulting data is composed of source data in this format:

{order},{index},{first_name},{last_name}

In our case the records will be:

100,1,Exciting,Randene
203,298,Continuing,Hedi
288,213,Promising,Ranna
1045,1056,Printed,Aurelie

The 4 records will be sent to sink as 4 separate parts of one multipart upload.

When your job returns ProcessResult.NotYetDone it means its task is not done yet and it needs to be called again to continue when it left off. If it returns ProcessResult.Finished it means the task is done and the output data is ready to be checked.

At this time all temporary data (documents) in Mongo should be cleaned up. Remaining empty collections are OK.