Best way to store references to a link

[andreyvital]

As the search function is entirely based on hashtags we need to store them in a wise way that accommodates both performance and reuse.

As the first candidate implementation of index.Index is using BoltDB, I thought in two buckets:

• kb:links
• kb:tags2link

kb:links

Where all links will be stored. As a link always have an UUID the key will be this identifier and the value is a JSON snapshot of link.Link.

Something like that:

Key	Value
0178b3b5-97ed-48a5-87ec-b5fba6d2d469	{...}
e4843306-b686-4f06-b63b-36810de580ec	{...}
88b72ea8-a15f-49bc-9473-b948a5ac2e7d	{...}

kb:tags2link

Where all tags will be stored: the tag itself as the key and value will be all link identifiers that did reference to this tag. You can imagine:

Key	Value
#google	["0178b3b5-97ed-48a5-87ec-b5fba6d2d469"]
#search	["0178b3b5-97ed-48a5-87ec-b5fba6d2d469"]
#front-page-of-the-internet	["0178b3b5-97ed-48a5-87ec-b5fba6d2d469", "(...)", "(...)"]

Thoughts?

[txgruppi]

I was thinking we should use Redis. It can be used by several apps, not just kb. Other apps can access the kb data, etc.

In Redis there are the Set commands which would work for this project.

We would store the data in the same way you described with bolt and we would use SINTER to get links which have all the same categories.

[andreyvital]

👍