mtds/support-software-interview-exercise-01

OLCF Support Software Development Team - Coding Exercise 01

OLCF Support Software Team - Coding Exercise 01

Overview

The simplified exercise below is representative of the kind applications we build within the OLCF/NCCS Support Software Team. When you are finished with the coding exercise, please send your submission directly to us via email as a .tar.gz or .zip file. The exercise should take approximately (2) hours to complete.

Deliverable

Using a Ruby or Python based web application framework (e.g. Rails, Sinatra, Django, or Flask), write a small application that provides a single RESTful API endpoint which returns a filtered set of "batch job records" from the provided dataset. We have no preference between Ruby and Python. For the sake of this exercise, think of a "batch job record" as a concise summary of some work done on a supercomputer.

Your deliverable, at minimum:

• Should provide an informative README.md (i.e., run/installation instructions, what could be improved, etc.)
• Should ETL the provided CSV into a datastore of your choosing
• Should gracefully handle omissions within the example dataset
• Should provide a functional API endpoint at /batch_jobs
• Your API endpoint should respond with a JSON-API document (string)
• Your API endpoint should correctly filter the dataset based upon any combination of filter keywords (see "Filter Keywords" section below)
• Should provide basic test(s) as you see fit

Limited use of open-source third-party gems/packages beyond your chosen framework is fine, as long as your README describes the build/installation process thoroughly.

Specifications

Below are detailed specifications to which your deliverable should adhere.

Example Data

The example batch job record dataset is contained in this repo in the example_batch_records.csv file.

Each batch job record occupies one line of the file, and each line contains up to (3) comma-delimited attributes for the record, in this order:

• batch_number (integer) - A number assigned to the batch job by the batch scheduler
• submitted_at (ISO 8601 datetime) - The time the batch job was submitted
• nodes_used (integer) - The number of nodes the batch job used

The data are well-formatted, but some attribute values could be missing entirely.

API Endpoint:

GET http://localhost/batch_jobs

API Endpoint Filtering

GET http://localhost/batch_jobs?filter[submitted_after]='2018-02-28T15:00:00+00:00'&filter[submitted_before]='2018-03-01T15:00:00+00:00'&filter[min_nodes]=2&filter[max_nodes]=20

API Response

The expected response is a JSON-API document (string) representing an array of objects, one object per record. For example, parsing the API response string should yield something like:

{
  "links": {
    "self": "http://localhost/batch_jobs?filter[max_nodes]=20"
  },
  "data": [
    { "type": "batch_jobs",
      "id": "1",
      "attributes": {
        "batch_number": 945,
        "submitted_at": "2018-03-30T15:00:00+00:00",
        "nodes_used": 8
      }
    }, {
      "type": "batch_jobs",
      "id": "2",
      "attributes": {
        "batch_number": 946,
        "submitted_at": "2018-03-30T15:07:01+00:00",
        "nodes_used": 20
      }
    }, ...
  ]
}

All filter parameters are optional. All filters should be inclusive (e.g. filter[min_nodes]=2&filter[max_nodes]=4 should return records that used 2, 3, or 4 nodes).

Filter Keywords

• [submitted_after]= (ISO 8601 datetime) - return only records submitted on or after the given date
• [submitted_before]= (ISO 8601 datetime) - return only records submitted on or before the given date
• [min_nodes]= (integer) - return only records that used at least the given number of nodes
• [max_nodes]= (integer) - return only records that used at most the given number of nodes