bdkiran / lyr-service Goto Github PK

Create a readme to contain:

• Use proper markdown syntax

• Brief explanation about the repository

• How to download

• Dependencies

Utilizing gothic, authentication with google, facebook and apple should be implemented.

• Watch out for concerns with xss, website may need to be consolidated on one url.

Create a query to return the "best" lyrics

• Return lyrics with the greatest scores (# of upvotes - # of downvotes)

Tests should be provided covering all functions, these unit tests should be small if scope but provide sufficient converage. These should exist in their own package :

• Determine what to do with api calls functions to elastic-search

• Determine how to test http handler functions

Upvote should reflect an upvote for a single user

• Users id gets added to the upvoted users array in the lyric.

• User can only appear once in this array.

• If already present sent 400

• If user is currently in the downvote array, remove them.

• Endpoint: /upvote/{document_id}

Some Request require that the caller be authenticated/authorized to complete that action. We need to add a check.

Routes:

• /upvote/{doc_id}

• /downvote/{doc_id}

• Check if request has auth bearer token, if not return 401

• Check if token is valid (secret is kept in this service) if not return 401

• Check if user id matches with token user sent

All functions should have all possible errors handled:

• no _ for errors

• Generic functions for error, to bubble up to the stack

• Preventing errors from completely overwhelming the service

Message are able to be sent from the front end to the backend. However, information is not stored.

• Add honeypot to reduce spam

• Store the messages

• ^Decide where to store the messages? Mongo, Elastic-search, Postgres?

Upvote should reflect a downvote for a single user

• Users id gets added to the downvoted users array in the lyric.

• User can only appear once in this array.

• If user is already present sent 400

• If user is currently in the upvote array, remove them.

• Endpoint: /downvote/{document_id}

The exact matching on keywords in the song_lyrics index will cause an issue with the following scenario:

1. In the elastic-search document, there are songs by the artist roddy ricch

2. User searches for roddy

3. The user does not get any results back

This is because the keyword fields of artist and title are looking for an exact match, this needs to be adjusted so that these fields search is not so strict

• Allow for words contained in a keyword field to be returned.

Resources:
https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-simple-query-string-query.html#simple-query-string-syntax
https://stackoverflow.com/questions/54058630/multiple-words-act-as-single-word-in-search-elasticsearch
https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-simple-query-string-query.html
https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-query-string-query.html

Create a new package that will handle all logging for the service
Include

• Type of log: debug, info, warn, error

• Timestamp

• Package/File it was called

Change the / endpoint to a health check endpoint

• Query elastic-search to get the health of the cluster

• Send back a message on the status

Return 15-20 results instead of 10, or instead look at using a "scroll" function

The last data structure change in ES alleviated issues concerning duplicate lyrics.
When this was done the artist and title fields in our elastic-search documents were changed from text fields to Keyword fields. This change was made for the future in mind(for aggregations and such).
However, this has left us with an issue since the query performs and "exact" match search on these keyword fields.
For Example take the following document:

/song_lryics/doc1
{
    'artist': 'drake'
    'title': 'cool song'
    ...
}

When a match query is performed using Drake this document would not be returned.
This is because the case difference is and issue with the exact match.

• Fix case sensitivity for artist field in query
• Fix case sensitivity for title field in query

Come up with a way to store environment variables for:

• Elastic-search Connection string

Sometimes lyrics have duplicates such as the chorus. We only want one instance of each lyric to be returned to the user for a better experience

• No 2 identical lyrics are returned

Sources

• https://stackoverflow.com/questions/25448186/remove-duplicate-documents-from-a-search-in-elasticsearch
• https://www.elastic.co/guide/en/elasticsearch/reference/current/search-aggregations-metrics-top-hits-aggregation.html
• https://stackoverflow.com/questions/32622309/elasticsearch-remove-duplicates-from-search-results-of-analyzed-fields