Feature Request: Random Document Retrieval in GROQ

[ghost]

**Is your feature request related to a problem?
Currently, GROQ lacks a built-in mechanism to fetch documents in a random order.
This limitation complicates scenarios where displaying randomised content is essential, such as in creating dynamic, engaging user experiences. Developers must resort to custom, often cumbersome, workarounds to achieve this functionality, which can increase development time and complexity.

Describe the solution you'd like
Introduce a straightforward method for random sorting within GROQ queries.
This enhancement could be implemented with simple function calls within the order clause, examples of which might include:

order(random()): Orders the results based on a random seed.
order(shuffle()): Shuffles the results to return them in a non-deterministic order.

Describe alternatives you've considered

While custom workarounds exist, they often involve fetching more documents than necessary and shuffling them client-side or periodically updating a "random" field in each document to sort by. These methods are not as efficient or straightforward as having native support for randomness in GROQ queries.

Additional context
Random document retrieval is a common requirement for many applications, including content sites looking to showcase a "random article of the day," e-commerce platforms featuring random products, or any service aiming to enhance user engagement by providing a non-static experience. Native support for this feature in GROQ would significantly simplify development workflows and enable more dynamic content delivery strategies.

[nicolopadovan]

This was briefly discussed already here.
I will tag @kmelve since he was already involved in the discussion.

He makes a good point in this comment: how would random sorting affect caching?

[judofyr]

The simplest way of providing some randomness while still supporting caching: Add support for a seed=<num> parameter in the REST endpoint. The same seed always generates the same random number/sequence. The client would then have to do seed=${Math.floor(Math.random() * 100)}. By choosing a large number they will get "more" variants, but less cache hit rate. By choosing a smaller value (e.g. 10) it's "less random", but performance will be better.

[nicolopadovan]

The simplest way of providing some randomness while still supporting caching: Add support for a seed=<num> parameter in the REST endpoint. The same seed always generates the same random number/sequence. The client would then have to do seed=${Math.floor(Math.random() * 100)}. By choosing a large number they will get "more" variants, but less cache hit rate. By choosing a smaller value (e.g. 10) it's "less random", but performance will be better.

This could work but basically delegates the responsibility to the final user. I can totally see new developers using big numbers just for the sake of it, resulting in less cache hit rate.

I don't know whether or not this aligns with the team's view as they might prefer a more consistant strategy, even if that came with less cache hit rates.

However I think that having a built-in function for this purpose would be more efficient than using client-side solutions: a client-side workaround I have found is, for example, to have a random index value (possibly normally distributed) in each document, then generating a random number client-side and fetching the first document that presents a larger index.
This comes with:

• Unpredictable caching
• Multiple fetches to achieve "true" randomness (you could fetch the first N numbers after the randomly generated number, but those would always be in the same order: you would need to run the same query with a different number N times)
• Additional properties within each object

{
  "_id": "doc1",
  "_type": "document",
  "title": "Document 1",
  "randomIndex": 0.42
},
{
  "_id": "doc2",
  "_type": "document",
  "title": "Document 2",
  "randomIndex": 0.78
},
{
  "_id": "doc3",
  "_type": "document",
  "title": "Document 3",
  "randomIndex": 0.15
},
{
  "_id": "doc4",
  "_type": "document",
  "title": "Document 4",
  "randomIndex": 0.63
}

*[_type == "document" && randomIndex > $randomNumber] | order(randomIndex asc)[0]