see assets/task.pdf
Please take a look at deployed demo.
Everything is packaged with docker / docker-compose.
To build project from scratch:
make build
To get running solution locally please run the following command:
docker-compose up or make run
Note: to build and run solution solution you should have docker and docker compose installed.
Projects consists of the following modules:
backend
API is implemented with Golang.
/
returns index page
---/hc
returns sample health check page
---/version
returns page containing current running application commit hash and build time
---/clients
returns page containing clients data
API module contains integrations tests which are run before every build.
frontend
UI is implemented using reactjs, d3.
To keep it simple simple and due to time limitations no ready-to-go framework for charting was used.
Current implementation is using directly d3 apis to graph two histograms for actual and predicted usage.
x axis corresponds to time period: 12 months numerated from 1 to 12.
y axis corresponds to usage data.
As far as usage data are heterogeneous we need to scale the y axis to have a consistent view independently of the clients data.
Possible improvements:
-
Currently due to time limitations we load data all at once. This can be improved by lazy loading only the data we need for visualization.
-
User experience can be improved consistently. Current interface is pretty basic due to time limitations.
-
Periods where actual usage differs consistenly from predicted one we can put markers to easily identify the differences
-
Have more relevant and complex visualizations to easily compare the differences between actual and predicted usage.
Instead of showing histogram we can use different visualizations:
-
stacked barcharts example
-
calendar heatmap example
-
show percentage increase / decrease instead of charts example
database
For this given task MongoDB database was chosen for its flexible schema approach.
To import data into mongodb we run dedicated service with docker-compose (see mongo-seed)
which runs mongo-seed/import.sh script.
To go faster we have manually extracted two collections from input data sophia-test.json:
cat sophia-test.json | jq '.domainY' > clients.json
cat sophia-test.json | jq '.data' > usages.json
The above mentioned script import.sh is using mongoimport flag --upsertFields which lets you merge two collections by providing merge fields.
After import in database we will have only one collection clients where each given record will contain client info as well predicted and actual usage data :
{
"salesforceId": 22699,
"country": "Morocco",
"owner": "Paul Piovene",
"manager": "Aristide Mullane",
"predictedUsage": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
1,
0,
0
],
"actualUsage": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
3,
0,
0
]
}
Given record is used directly by page comparing predicted / real usage for the client example.
assets
task definition, test data